The 'AI Employee' 62% Hoped For Was Shipped on July 6, 2026 by Survey Publisher INGAGE Itself; Effect Numbers Emerged Within Two Weeks -- Chiku Takubin Automated About 2,000 First-Response Calls per Month (All Corporate Releases, Not Independently Verified by This Site)

ANK-Doc ID: ANK-2026-07-19-007 Version: v1.0.0 Publication date: 2026-07-19 Author: 竹之內凜（全球總編輯） Category: Generative AI / Workplace adoption / Customer support & AI phone response / Customer harassment (kasuhara) countermeasures / AI cost governance Extended from: This card is an extension of ANK-2026-07-06-010 (not an independent new topic) -- target card in three languages: Traditional Chinese / Japanese / English Covered articles: 18 new items = PRTIMES#1336163 (INGAGE "AI Employee" release, 2026-07-06), #1364414 (INGAGE July 16 webinar notice, 2026-07-08), #1360037 (Helpfeel "autopilot call center" concept plus AI-Ready consulting, 2026-07-08), #1360730 (Verbex voice-AI market forecast, 2026-07-08), #1372747 (RightTouch "Web Assist", 2026-07-09), #1372915 (Quarq x Elephancube kasuhara-countermeasure e-learning, 2026-07-09), #1373002 (Altius Link x Chiku Takubin voicebot adoption case, 2026-07-09), #1462577 (Softfront Japan commubo contextualization feature, 2026-07-13), #1465644 (INGAGE N-ASSET adoption case, 2026-07-13), #1481873 (Channel Corporation two surveys, 2026-07-14), #1506471 (Verbex x TSK San-in Chuo Television PoC, 2026-07-15), #1508549 (JIIT CCDX Guideline FY2026 edition, 2026-07-15), #1539860 (Upsell Technologies x pluszero "miraio AI Main-Line Phone", 2026-07-16), #1565138 (VALTEC MOT/CallCenter Lite IVR statistics upgrade, 2026-07-17), #1566705 (Locaop Desk self-serve sign-up, 2026-07-17), #1568297 (pickupon DX human-resource training, 2026-07-17), #1579406 (INGAGE BPO Japan adoption case, 2026-07-17), #1600742 (Wonderful Fly token-cost explainer, 2026-07-19); 5 base items carried from the target card = PRTIMES#1337119 (ielove survey, 2026-07-06), #1336094 (INGAGE survey, 2026-07-06), #1281884 (GMO Pepabo, 2026-07-01), #1308694 (pickupon subsidy registration, 2026-07-03), #1337255 (C&T tel-trans, 2026-07-06) (all detailed in the target card) Selection method: The daily extension selector (ANKRUN-20260719213705) designated target card ANK-2026-07-06-010 (extension lane). The writing desk searched the full ainews library along the target card's "AI for first response, humans for complex cases" thesis over 2026-07-06 to 07-19 and adopted 18 new items (per-item dates under "Covered articles"). Doc-id note: this desk was originally assigned ANK-2026-07-19-006, but during production that number was taken first by another desk per the COORDINATION registry (allocated 2026-07-19 22:01); this desk re-allocated to ANK-2026-07-19-007 via the three-source procedure (local max=ANK-2026-07-19-006 / COORDINATION max=ANK-2026-07-19-006 / online HTTP404; DB not queried because REVIEW_ONLY desks may not ssh) and registered the reservation. Rejected materials and reasons: PRTIMES#1539863 (duplicate text of #1539860, counted once); #1506454 (TSK-side release of the same PoC as #1506471, counted as one event; the fact that both parties issued releases is noted in the Scope section); #1600594 (Optyino MCP connection = AI-search visibility, off-axis); #1579288 / #1463597 / #1463585 / #1464911 (customer-support award announcements = weak relevance, dropped); #1361645 (Virtualex Amazon Connect methodology = concept release, no numbers); #1411609 (ALICE outbound staffing calls = outbound, off the target card's inbound axis); #1373395 (FIXER x ElevenLabs development verification start = no numbers). Taiwan line: a full-library search (keywords such as AI customer service / smart customer service / voice customer service) returned zero relevant hits -- following the target card's principle of "honest comparison, no forced linkage", this card does not connect a Taiwan line and states only this. Total materials: 23 items = 18 items newly adopted + 5 items carried from the target card (owner directive of 2026-07-19: no cap on item count; quality is the only constraint). Other desks' drafts today (ANK-2026-07-19-001 through 006) were unpublished at press time; per the "link only published cards" rule no links are set, stated here only. Zero verified official anchors across the card (see Scope); this card attaches no Wikidata Q identifiers (no registry verification performed by this desk; absence over falsity). Publication date 2026-07-19 >= latest source date 2026-07-19 (PRTIMES#1600742); apart from year-notation transcriptions such as "previous year = 2023" (marked where used), this card generates no new numbers, computes no ratios, and totals no amounts. Regarding the customer-harassment countermeasure mandate, this card records only the facts and dates as stated in the releases and provides no legal commentary of any kind.

TL;DR

The target card (ANK-2026-07-06-010, published July 6, 2026) fixed the division-of-labor thesis "AI for first response, humans for complex cases": in the INGAGE survey, 62% of respondents hoped for an autonomously responding "AI employee" and 76% valued a "smooth handoff to humans". [F-024] Looking back two weeks later, that thesis advanced along five lines. First, from hope to product: on the very same day (July 6, 2026), the very same publisher INGAGE shipped a Re:lation feature named "AI Employee" -- feed it your existing manuals and it autonomously judges and handles customer contacts, escalating complex cases to a human with a report attached (staffed escalation) -- survey publisher = product vendor, survey and product released the same day; this card records that publication structure as such (PRTIMES#1336163). [F-001] Second, effect numbers emerged: per the adoption case published by Altius Link, with inbound calls at Chiku Takubin reaching about 200% of the prior year (specific year not stated in the source), the voicebot automated about 2,000 first-response inquiries per month (equivalent to about 1.5 operators) and cut per-call handling time from an average of 150 seconds to 120 seconds (about 20% shorter) (PRTIMES#1373002) [F-003][F-004]; INGAGE's N-ASSET case records email drafting time cut 50% versus pre-adoption via an AI mail agent (PRTIMES#1465644) [F-005]; the BPO Japan case records team-based handling of tens of thousands of emails per month (PRTIMES#1579406) [F-006] -- all three are single cases published by the service providers themselves and cannot be extrapolated to average effects (see Scope). Third, a third independent survey added the "kasuhara" axis entirely absent from the target card: Channel Corporation's two surveys (497 service-counter workers plus 514 general consumers) record that 51.7% of surveyed companies have no customer-harassment countermeasures; among consumers who actually encountered a store-side mistake (re-tabulated on 348 people), 52.3% gave up making a legitimate complaint for fear of "being seen as a harasser"; 65.2% are positive about an AI-mediated system; about 60% of non-face-to-face counter staff answered that "40% or more of the workload could be lightened by AI" (PRTIMES#1481873) [F-007][F-008][F-009]; Quarq's release states that the amended Comprehensive Labor Policy Promotion Act enacted in June 2025 makes kasuhara countermeasures mandatory for all employers from October 1, 2026 (as stated in the release; official statute not checked by this site), with a countermeasure e-learning package on sale July 15, 2026 (PRTIMES#1372915). [F-010] Fourth, a supply wave plus market and institutional coordinates: within two weeks came the launch of miraio AI Main-Line Phone (PRTIMES#1539860) [F-011], commubo's contextualization feature (PRTIMES#1462577) [F-012], RightTouch's "Web Assist" (PRTIMES#1372747) [F-013], Locaop Desk's web-complete self-serve sign-up (PRTIMES#1566705) [F-014], and Helpfeel's "autopilot call center" concept with the Tohoku Electric Power lead case (adopted March 2025; ten months in, 160,000 monthly page views and inbound call volume down about 20% = per the release) (PRTIMES#1360037) [F-015][F-016]; Verbex cited a METI survey (Japan's BtoC EC market at 26.1 trillion yen in 2024, up 5.1% from the previous year, 2023) and, by its own estimate, projected a voice-AI TAM of about 19.1 billion yen for EC and mail order by 2030 (PRTIMES#1360730) [F-017], and launched a first-response phone PoC with TSK San-in Chuo Television -- instant FAQ answers, individual and specialized cases routed to humans: the field version of the target card's thesis (PRTIMES#1506471) [F-018]; VALTEC upgraded IVR statistics and cited the Persol Research Institute estimate of "a 6.44 million-person labor shortage in 2030" (PRTIMES#1565138) [F-019]; JIIT published the "Contact Center DX Promotion Guideline, FY2026 edition" (PRTIMES#1508549) [F-020]; and pickupon advanced the subsidy axis recorded in the target card into a training product (PRTIMES#1568297). [F-021] Fifth, today's cost antithesis: on July 19, 2026, a Wonderful Fly explainer framed "token cost" as an emerging management issue, citing reports that Uber burned through its annual AI-coding-tool budget in four months and that Microsoft consolidated Claude Code direct licenses into GitHub Copilot CLI (both second-hand mentions of press reports inside the company's explainer) -- the next gate for the division-of-labor thesis is not "can it be done" but "does it pay" (PRTIMES#1600742). [F-022] Scope (details in the Scope section): all 23 source items are corporate or organizational self-published PR TIMES releases with zero verified official anchors; each survey's percentages represent only its own respondent population; a product release is not an adoption record, a PoC is not formal adoption, and "coming soon" is not "available now".

Body

Overview: what is extended, and how

This card extends ANK-2026-07-06-010. The target card, snapshotted at July 6, 2026, recorded what AI phone response delivers and where it stops, plus the training limits of customer-support sites, converging on the collaborative scheme "AI for first response, humans for complex cases". This card implements 19 explicitly labeled layers from 18 new materials (typed actions and sources mapped in J-003) to answer one question: how, in the following two weeks (July 6 to July 19, 2026), did that scheme grow from survey numbers into products, effect numbers, institutional coordinates, and a cost antithesis? The five base facts carried from the target card (F-023 to F-027) are marked "carried from target card, not new" and do not count toward significance.

Three processing principles first (details in Scope): first, all 23 source items are corporate or organizational self-published releases on PR TIMES, and the card has zero verified official anchors; second, adoption-case effect numbers are single cases published by the service providers themselves (scope details in the Scope section); third, the surveys differ in population and method, so this card presents scopes side by side without summing or adjudicating.

Layer 1 CONNECT + CHALLENGE: from hope to product -- the same-day shipping of the "AI Employee"

The target card recorded: in the INGAGE survey (June 3-11, 2026; main survey of 484), 62% of respondents hoped for the ideal division of labor "AI handles first response, humans take over complex cases", which the publisher also phrased as 62% hoping for an autonomously responding "AI employee". [F-024] Two weeks on, "AI employee" did not remain a survey term -- on the same day (July 6, 2026), the same publisher, INGAGE, shipped a Re:lation feature named "AI Employee": feed it the existing manuals written for new operators and it autonomously judges and handles customer contacts; select "AI Employee" in the case-owner field and it starts working; sliders adjust personality parameters such as politeness, reply speed, and emotional attentiveness. The release describes the development trigger as an internal experiment -- deploying an "AI that can answer anything about Re:lation" at the internal help window reached "a level where almost no human response was needed" (the publisher's own account, no quantitative figures) -- and frames the background problem as the "maintenance wall" of dictionary-type (scenario-type) chatbots: narrow coverage premised on continuously maintaining a vast knowledge base (PRTIMES#1336163). [F-001]

The CHALLENGE in this layer concerns not the product but the publication structure: the survey (PRTIMES#1336094, July 6, 2026) and the product (PRTIMES#1336163, July 6, 2026) were released the same day by the same publisher, and the product release directly re-quotes the survey's numbers (52% understaffed, 55% needing six months or more for new hires to work independently, 63% over-dependence on individuals) as development background. [F-001] That both releases are dated July 6, 2026 is a checkable fact; that the survey publisher is the product vendor is a checkable fact -- this card records that structure and interest relationship so readers can place "62% hope for an AI employee" back into its product context, and asserts nothing about publication intent (see Scope). Worth noting alongside: the product design itself embeds the direction that 76% of the target card's respondents valued -- "complex contacts it cannot resolve are immediately handed to a human, with a report based on the interaction attached (staffed escalation)" (PRTIMES#1336163). [F-001]

Layer 2 CORROBORATE: a second release pins the ship date

On July 8, 2026, INGAGE issued a separate webinar notice stating that "AI Employee" is "the Re:lation feature released on July 6, 2026", with a free webinar held online on July 16, 2026, 12:00-12:30, and the same notice claims over 6,000 companies using Re:lation (including trials) (PRTIMES#1364414). [F-002] This second release by the publisher itself pins the ship date at July 6, 2026 -- the same-day-as-survey fact can be cross-checked across the publisher's own two documents.

Layers 3-4 ADD_EVIDENCE: effect numbers, part one -- Chiku Takubin's three hard figures

The target card's supply side had product releases only, no effect numbers. Within two weeks the hardest set on this theme so far appeared: per Altius Link Co., Ltd.'s adoption case published July 9, 2026, after the customer center of Chiku Takubin Co., Ltd. deployed the voicebot "Virtual Agent Voice", first-response automation reached about 2,000 inquiries per month (equivalent to about 1.5 operators), per-call handling time fell from an average of 150 seconds to 120 seconds (about 20% shorter), and a completion rate of about 80% was maintained (the release carries a "*1" annotation whose definition is not expanded in the captured text) (PRTIMES#1373002). [F-003]

Layer 4 adds background and method: the adoption background was inbound calls reaching about 200% of the prior year (baseline is the prior year; specific year not stated in the source) amid surging parcel volumes and chronic understaffing; the method was to sort inquiry types by direct-mail category, phase in from the easier inbound patterns, refine hearing items and voice guidance, and tune the "pauses" of the machine voice for audibility (PRTIMES#1373002). [F-004] Pin the scope: this is a single adoption case published by the service provider (Altius Link), not a third-party audit, not extrapolatable (see Scope). But its structure is precisely the quantified version of the target card's thesis -- first response to AI (about 2,000 calls per month), complex cases to humans ("operators can focus more on complex inquiries" = the release's wording).

Layer Five ADD_EVIDENCE: effect numbers, part two -- two ledgers, real estate and BPO

The target card's user-side survey took the real-estate industry as its population (the ielove survey). [F-023] Within two weeks the same industry produced an effect case: per INGAGE's July 13, 2026 release, N-ASSET Co., Ltd. (Kawasaki, Kanagawa; the release calls it No. 1 in rental brokerage transactions in Kawasaki's Takatsu ward -- vendor wording, survey body and time point not stated) used Re:lation's "AI mail agent" (type the gist of a reply and it auto-generates appropriate text) to cut email drafting time by 50% versus pre-adoption; it also removed personal email addresses from business cards entirely, unifying contact through shared addresses, and the release says complaint occurrence "plummeted" (qualitative, no figure) (PRTIMES#1465644). [F-005]

The other ledger is in BPO: per INGAGE's July 17, 2026 release, BPO Japan Co., Ltd. (centered on recruitment process outsourcing, RPO) moved tens of thousands of emails per month to a team-based handling regime with Re:lation -- several hundred to over 1,000 emails a day processed with labels, auto-routing, and owner assignment; at a scale of over 100 full-time employees, with multiple sites (Osaka, Sendai, Fukuoka and others) plus remote members, response status is visualized in real time; the title's "toward zero mistakes" and the body's "zero-mistake regime" are the publisher's claims (PRTIMES#1579406). [F-006] Both cases are adoption stories published by Re:lation's vendor INGAGE itself (see Scope); read together with Layer 1, the value is this: the vendor of the "AI Employee" is accumulating an "effect numbers" narrative layer from existing-customer cases -- which is also the baseline against which readers can check future quantitative disclosures.

Layer 6 CORROBORATE: a third independent survey -- the "kasuhara" axis joins the division-of-labor thesis

The target card's two surveys (ielove, INGAGE) both asked about efficiency and training; the target card had no "customer harassment (kasuhara)" axis at all. A third independent publisher supplied it: Channel Corporation (provider of the AI agent "Channel Talk" -- the survey publisher is simultaneously a vendor of related products, see Scope) published two surveys on July 14, 2026. Survey 1 (497 workers in service counters, customer support and similar roles; internet survey June 18-22, 2026; main contact settings = in-person store/counter 67%, phone/email/chat 29%, other 4%): to "countermeasures your company has adopted against kasuhara" (options drawn from the Tokyo Metropolitan customer-harassment prevention program's recommended measures), the most common answer was "none" at 51.7%; the top adopted measures were "publicizing a basic policy internally and externally" 27.4% and "creating and publicizing a kasuhara response manual" 24.9%, while technical and environmental measures such as recording or AI systems were under 10% (PRTIMES#1481873). [F-007]

Worker expectations of AI also have structure: if an AI chat with first-response and complaint-filtering functions were introduced, the top expectation was "fewer occasions of directly receiving emotional complaints, easing the mental burden" at 28%; the answer "first response would be reduced, letting me face the customers who really matter" was chosen by 21.4% of email/inquiry-form staff and 10.7% of call-center/CS-counter staff (the release states both are about twice the overall average); about 60% of non-face-to-face counter staff answered that "40% or more of the workload could be lightened by AI" (PRTIMES#1481873). [F-008]

Survey 2 (514 general consumers in their 20s to 60s nationwide, June 12-14, 2026) illuminates the consumer side's silence: 35.4% of all respondents (about 1 in 3.5) have given up pointing out a store's clear mistake (wrong orders, billing errors, damaged goods) or requesting a refund for fear of being seen as a "harasser" or "complainer" (frequently 7.8% + sometimes 17.1% + once or twice 10.5%); excluding the 27% who "have never encountered a store-side mistake" and re-tabulating on the 348 who actually did, 52.3% have this "swallowed-it" experience; if an AI could sit in between and deliver "legitimate complaints" to the store in organized form, "definitely want to use it" 22.6% + "would use it depending on the content" 42.6% = 65.2% positive (PRTIMES#1481873). [F-009] This set of numbers extends the target card's division-of-labor thesis from "efficiency" to "worker protection and the voice of the customer": first response to AI here gains a third rationale -- AI as a buffer layer for emotional conflict (the publisher calls it "AI-mediated communication"). Each percentage represents only its respondent population (see Scope).

Layer 7 CONTEXTUALIZE: the mandate date and the commodification of countermeasures

The kasuhara axis has an institutional clock: per Quarq Co., Ltd.'s release of July 9, 2026, the amended Comprehensive Labor Policy Promotion Act enacted in June 2025 makes customer-harassment countermeasures mandatory for all employers from October 1, 2026, with no grace period by company size (as stated in the release; this site has not checked the official statute -- treated as a risk note, not an official endorsement); the "Customer Harassment Countermeasure E-Learning Package", a tie-up between Quarq and Elephancube (cumulative e-learning production record of over 2,000 titles = as stated in the release), went on sale July 15, 2026 -- eight chapters, total running time 1 hour 13 minutes, plus a ten-question test (PRTIMES#1372915). [F-010] Channel Corporation's survey release likewise frames itself explicitly against the "October 2026 mandate" (PRTIMES#1481873). [F-007] This card records only the facts and dates as stated in these releases and offers no legal commentary (see Scope). The checkable structural fact is this: the mandate date has entered the product and survey narratives of two different publishers -- the countermeasures themselves are already a product.

Layer Eight ADD_EVIDENCE: the supply wave, part one -- miraio AI Main-Line Phone

The supply side logged at least four product moves in two weeks. First: on July 16, 2026, Upsell Technologies Co., Ltd. and pluszero Inc. announced the launch of their co-developed AI operator "miraio AI Main-Line Phone" (miraio AI 代表電話) -- combining Upsell Technologies' self-claimed 20-plus years and cumulative 8,200-plus companies of call-center operating know-how with the next-generation AI engine "PUT" built on pluszero's proprietary concept "AEI (Artificial Elastic Intelligence)"; its pitch is "it does not lie" = a dual-process AEI model that never returns LLM output directly, suppressing hallucination risk to the limit (vendor claim), with cloned voices of real human operators for natural dialogue; functions include real-time transcription and summarization of incoming calls with automatic notification (Slack, Teams, Lark, email and others), automatic screening and blocking of sales calls, autonomous FAQ answers, and emergency escalation -- however, "first response and transfer (24/7/365 capable)" is explicitly marked "coming soon" = not yet available at publication (PRTIMES#1539860). [F-011]

Layer Nine ADD_EVIDENCE: the supply wave, part two -- commubo turns the "handoff" into structured data

76% of the target card's respondents valued a "smooth handoff to humans". [F-024] The supply side's direct answer: on July 13, 2026, Softfront Japan (a subsidiary of Softfront Holdings) announced the generative-AI-type "contextualization feature" for the AI voicebot "commubo" -- call content is extracted and structured in real time into pre-set fields (name, address, inquiry content, preferred dates and the like), cutting the workload of data shaping, operator handoff, and content review (after-call work, ACW); it supports slot filling (auto-organizing out-of-order utterances and asking follow-ups for missing items), call summarization, and kanji/kana spelling confirmation; it ships as a standard feature that users configure themselves (PRTIMES#1462577). [F-012] The feature already existed for scenario-type bots and has now been extended to the generative-AI type -- the release also states the design philosophy: concentrate generative AI on "hearing work", where affinity is high, to contain wrong-answer risk.

Layer Ten ADD_EVIDENCE: the supply wave, part three -- RightTouch lets the AI operator see the customer's web actions

On July 9, 2026, RightTouch Inc. (carved out of PLAID Inc., TSE Growth 4165) announced the launch of "Web Assist", a new feature of QANT Speak: the AI operator grasps the customer's web-behavior data (past and real-time) during a call and supports on-screen operations by voice -- aimed at preventing mid-procedure drop-off, raising self-resolution rates, and reducing repeat calls (vendor description); it works in both directions (phone-to-web guidance = URL sent by SMS while the voice connection is maintained; web-to-phone = pre-entered questions and web-behavior data handed to the AI); when the AI cannot resolve, the conversation log and web-behavior data are consolidated on the operator dashboard "Connect Board" and handed to a human (PRTIMES#1372747). [F-013] The release also states that the QANT series recorded an intent-based routing accuracy of 99.3% in a verification case (a single verification case, per the release; see Scope). The target card's "smooth handoff to humans" is here engineered as data handoff -- same direction as Layer 9, different implementation.

Layer Eleven ADD_EVIDENCE: the supply wave, part four -- Locaop cuts away the adoption friction itself

On July 17, 2026, Locaop Inc. announced self-serve sign-up for its AI phone reception service "Locaop Desk": the adoption process, previously premised on a sales meeting, now completes entirely on the web -- application, online identity verification, and credit-card payment (Stripe; monthly or annual) -- with service starting as soon as the same day once the prescribed checks and setup are complete; the entry point is a "14-day free trial", and if not upgraded within the trial the issued phone number is decommissioned (PRTIMES#1566705). [F-014] The adoption barriers the target card recorded were "IT staff" and "accuracy anxiety" (INGAGE bottleneck survey, 34% and 32%) [F-024]; this move attacks a third barrier -- the friction of meetings and paperwork itself.

Layer 12 CONTEXTUALIZE: Helpfeel's "autopilot" concept and the Tohoku Electric Power lead case

On July 8, 2026, Helpfeel Inc. announced its "autopilot call center (AI autonomous operation)" concept and the start of an AI-Ready consulting business: a six-stage roadmap, Level 0 to 5, by AI maturity (Level 5 = an autonomous AI agent handles all work including discretionary judgments and exceptions, with humans supervising and managing the AI); in the lead case, Tohoku Electric Power Co., Inc. adopted the self-resolution AI "Helpfeel" in March 2025, and the release records 160,000 monthly page views ten months after adoption, inbound call volume down about 20% (the baseline time point is not stated verbatim in the source; the context is a pre/post-adoption comparison), and a substantial improvement in answer rates; Tohoku Electric Power has completed Level 1 and is now in Level 2 (PRTIMES#1360037). [F-015]

The same release re-cites three sets of market statistics (all second-hand citations inside the release, original surveys not checked by this site; see Scope): Japan's call-center market exceeds 1 trillion yen (Yano Research Institute, 2025 survey); on operator hiring, "very difficult at all sites" 35% + "very difficult at some sites" 41% = about 80% of companies struggling to secure staff, and answer rates falling for the third consecutive year (both from Ric Telecom's "Call Center White Paper 2025"; covered years not expanded in the release); the share of Japanese companies saying generative-AI results "exceeded expectations" is one quarter that of Western countries (PwC Japan Group, "Generative AI Survey Spring 2025, five-country comparison") (PRTIMES#1360037). [F-016] The "gap between adoption and results" -- this is the quantitative counterpart of the "adopted but unused" problem that GMO Pepabo's release flagged on the target card's supply side. [F-025]

Layer 13 CONTEXTUALIZE: Verbex posts the first market coordinate -- a TAM of about 19.1 billion yen (own estimate)

A market-size number appeared for the first time in the two weeks: on July 8, 2026, Verbex Inc. published a voice-AI market forecast -- citing METI's "FY2024 (Reiwa 6) Digital Transaction Environment Improvement Project (e-commerce market survey)", it states Japan's BtoC EC market at 26.1 trillion yen in 2024, up 5.1% from the previous year (2023) (Verbex citing the METI survey; original survey not checked by this site); on top of that, by its own estimate, it puts the order-taking and customer-support market attached to EC and mail order at about 300 billion yen, of which the voice-response segment (phone orders and inquiries) is about 150 billion yen, and forecasts that by 2030 voice AI will be used in 17% of that segment, expanding the voice-AI TAM in EC and mail order to about 19.1 billion yen (the release title says "about 19.0 billion yen") (PRTIMES#1360730). [F-017] Pin the scope: the 26.1 trillion yen is a re-citation of the METI survey, while the 300 billion, 150 billion, 17%, and roughly 19.1 billion yen figures are all Verbex's own estimates and forecasts, not official statistics (see Scope). The release's three diffusion steps (first response, callbacks and FAQ; then structured order-taking and repeat orders; then hybrid operation with AI-human division of roles) are isomorphic to the target card's thesis.

Layer 14 CORROBORATE: the TSK San-in Chuo Television PoC -- the field version of "first response to AI"

Verbex further announced on July 15, 2026 the start of a proof of concept (PoC) with San-in Chuo Television Broadcasting Co., Ltd. (TSK, Matsue, Shimane; President Tanabe Choemon): for inquiries to TSK's new business launched in May 2026, the "TSK Brand Item Home-Pickup Purchase Service", Verbex's voice AI handles first-response phone reception -- routine inquiries answerable from FAQs (application method, shipping method, service content) are answered instantly, while matters involving individual circumstances or requiring specialized judgment are not completed by the AI but guided to the appropriate contact; operation starts as internal verification, moving in stages to customer-facing use after reviewing actual inquiries and operational issues; the two companies have partnered from FY2026 to promote voice-AI adoption among municipalities and regional companies centered on the San-in region (PRTIMES#1506471). [F-018] This is an instance of the target card's "AI for first response, humans for complex cases" scheme written verbatim into a verification design -- though PoC does not equal formal adoption and this is the internal-verification stage (see Scope). The same PoC was also announced separately by the TSK side (PRTIMES#1506454, counted as one event, not adopted -- see Selection method).

Layer 15 CONTEXTUALIZE: VALTEC's IVR statistics and the 6.44-million citation

On July 17, 2026, VALTEC Co., Ltd. announced an upgrade to the IVR (interactive voice response) statistics and reporting of MOT/CallCenter Lite: added aggregation of IVR abandoned calls (abandonment counts and rates), elimination of "IVR unknown" labels so all calls are correctly classified, accurate counting of unanswered extension transfers, and improved call-time aggregation (accurate AHT, average handling time); the release also re-cites the Persol Research Institute estimate of "a 6.44 million-person labor shortage in 2030" (second-hand citation, original estimate not checked by this site); it further records, as a general IVR adoption example, average daily incoming calls falling from 347.8 to 272.5 (about a 20% cut; baseline is the pre-adoption 347.8 calls/day) (PRTIMES#1565138). [F-019] Outside the AI-agent conversation, the measurement infrastructure of existing IVR is being reinforced in parallel -- abandoned calls and AHT are precisely the denominator data for judging how "first response to AI" performs.

Layer 16 CONTEXTUALIZE: JIIT codifies the "common language"

The Japan Institute of Information Technology (JIIT; Minato, Tokyo; chairman Akasaka Yuji, chairman of the board of Japan Airlines; founded July 16, 1981; 241 member companies) announced on July 15, 2026 the free publication of the "Contact Center DX Promotion Guideline, FY2026 edition": DX maturity is assessed on 12 diagnostic axes composed from four perspectives (customer, employee, management, society), organizing roadmap design, KPI setting, and promotion points by industry and scale; the goal is to turn contact centers into "strategic bases for customer value creation" and serve as a common DX language from executives to front-line staff; FOXeey Consulting Inc. (founded April 1, 2025) served as lead organizer, with the cooperation of the CCDX Guideline Drafting Committee (PRTIMES#1508549). [F-020] Scope: the publisher is a public-interest incorporated association, but this card's capture is the PR TIMES release, not the association's original document (see Scope).

Layer 17 CONNECT: the subsidy axis, same entity, next step -- pickupon from "registration" to "training"

The target card's policy side recorded: pickupon's AI phone tool was registered, effective June 12, 2026, as eligible under the "Digitalization & AI Adoption Subsidy 2026" (regular track 50,000 yen to under 1.5 million yen; subsidy in principle up to one half). [F-026] The same entity's follow-up two weeks later: on July 17, 2026, pickupon Inc. announced a hands-on "DX Human Resource Training" -- using the client company's own call, meeting and negotiation data as the teaching material, participants build AI summarization rules, CRM/SFA recording methods, and team review flows in practice, checking results in numbers (deal-conversion rate, win rate, handling time, missed-record counts, per company-set targets); four courses (AI article DX / product & service improvement DX / risk analysis & crisis management DX / internal feedback reinforcement DX), each about 13 hours (e-learning plus live online sessions), minimum three participants per course; the release notes the possibility of applying the Human Resource Development Support Grant to training costs and the Digitalization & AI Adoption Subsidy 2026 to new pickupon adoption (regular track 50,000 yen to under 1.5 million yen, subsidy in principle up to one half; applicability varies by company size, business, training content, and application timing = stated in the release), with a partnered labor and social security attorney supporting requirement checks through document preparation (PRTIMES#1568297). [F-021] The policy-money axis has stretched from "tool registration" to "human-resource training" -- the subsidy narrative is beginning to cover the whole chain around adoption.

Layer 18 CHALLENGE + CONNECT: today's cost antithesis -- token cost

The four lines above all point to expansion. Today's (July 19, 2026) fifth line is the antithesis: Wonderful Fly Co., Ltd.'s explainer frames "token cost" as an emerging management issue -- an AI agent calls generative AI multiple times to complete a single task (reading the inquiry, judging intent, searching internal information, drafting the answer, checking the answer), so what looks like one operation to the user consumes more tokens than expected behind the scenes; cost drivers include feeding long documents and past conversation history every time, using high-end models even for simple classification and extraction, uncontrolled retry counts on errors, and not tracking consumption per agent; it proposes four management points (setting token usage caps, visualizing usage and cost, per-agent usage analysis, and dividing work with workflows and rules) (PRTIMES#1600742). [F-022]

Citing press reports, the same piece states: in 2026 Uber burned through its annual budget for AI coding tools in just four months and subsequently set monthly caps per tool; Microsoft terminated many direct licenses for Anthropic's "Claude Code", which had spread rapidly internally, consolidating usage into GitHub Copilot CLI -- both examples are second-hand mentions of press reports inside Wonderful Fly's explainer (the source says "it was reported that"), not independent facts of this site, and this card makes no factual assertion of them (PRTIMES#1600742). [F-022] The "AI cost becoming a management issue" recorded in our Workplace AI Chapter One (ANK-2026-07-01-004) here connects to the customer contact point: every "first response by AI" is token consumption -- the next gate for the division-of-labor thesis is not "can it be done" but "does it pay" (this card's editorial reading, see J-002).

Layer 19 CONNECT: pinning back to this site's existing record

Two published cards are cited back: ANK-2026-07-01-004 (Workplace AI Chapter One: the "Big Three era" in tool share and AI cost becoming a "management issue" -- numbers in that card) -- this card's Layer 18 token-cost explainer is the extension of that card's cost axis into the customer-contact domain (the two cards differ in materials and populations; directional linkage only, no number comparison). ANK-2026-07-03-018 (Workplace AI Chapter Two: the "ledger of realized gains" -- the Kokuyo Group/Kaunet employee survey recording an average of about five hours saved per respondent per month; numbers in that card) -- this card's Layers 3-5 adoption-case effect numbers (about 2,000 calls per month, about 1.5 operators, about 20% shorter, 50% cut) are the next page of the same "realized-gains ledger" methodology at customer-support sites (that card is a single organization's employee survey while this card is vendor adoption cases; different scopes, no summing, no comparison).

Scope (§口徑, single consolidated block)

• Zero verified official anchors across the card: official_count_verified=0, official_count_raw=0, official_attribution_unverified_count=0. All 23 source items are corporate or organizational self-published releases on the PR TIMES platform (including the release by the public-interest association JIIT). Figures and dates appearing inside releases and attributed to the METI survey (2024 BtoC EC market 26.1 trillion yen, up 5.1% from the previous year, 2023), the Persol Research Institute estimate (6.44 million-person labor shortage in 2030), Yano Research Institute (market over 1 trillion yen), Ric Telecom's "Call Center White Paper 2025" (about 80% struggling with hiring; answer rates down three consecutive years), the PwC Japan survey (results-exceeding-expectations share one quarter of Western countries), Ministry of Health, Labour and Welfare-related passages, the Tokyo Metropolitan recommended-measure options, and the amended Comprehensive Labor Policy Promotion Act with the October 1, 2026 mandate date are all second-hand citations or mentions inside the releases; this site has not checked the original documents or surveys and has not independently verified them. They are recorded as facts-with-risk-notes, not treated as official endorsement. This card attaches no Wikidata Q identifiers (no registry verification performed; absence over falsity).
• Adoption cases = single cases published by the vendors themselves: the effect figures for Chiku Takubin (published by Altius Link), N-ASSET and BPO Japan (published by INGAGE), and Tohoku Electric Power (published by Helpfeel) are adoption cases published by the service providers, not third-party audits; single cases cannot be extrapolated to average effects or guarantees. The "*1" definition behind Chiku Takubin's "about 80% completion rate" is not expanded in the captured text; N-ASSET's "No. 1 in rental brokerage in Takatsu ward, Kawasaki" is the release's wording with survey body and time point unstated; BPO Japan's "zero mistakes" is the title's and publisher's claim.
• Publication structure and interest relationships: INGAGE released its survey (#1336094) and the "AI Employee" product (#1336163) on the same day, July 6, 2026, with the survey's numbers re-quoted inside the product release -- this card records that structure and asserts nothing about intent. Channel Corporation is the provider of the AI agent "Channel Talk", so its surveys (#1481873) are interested-party surveys. The surveys (ielove 1,108 responses; INGAGE 484; Channel 497 + 514) differ in population, method and questions; percentages represent only their own respondent populations and cannot be added to or netted against each other.
• Own estimates and vendor-claimed figures: Verbex's roughly 300 billion yen, 150 billion yen, 17% in 2030, and TAM of about 19.1 billion yen (title: about 19.0 billion yen) are all company estimates and forecasts, not official statistics; "over 6,000 companies (including trials)" (INGAGE), "cumulative 8,200-plus companies" (Upsell Technologies), "over 230,000 companies globally" (Channel), "over 900 sites (as of April 2026)" (Helpfeel), "56 patents in 25 countries" (Verbex), "over 2,000 production credits" (Elephancube), and "99.3% intent-based routing accuracy" (RightTouch, single verification case) are the publishers' own claims, not independently verified by this site.
• Product release is not adoption record; PoC is not formal adoption; planned is not available: miraio's "first response and transfer" is marked "coming soon" = not available at publication; Verbex x TSK is a proof of concept starting with internal verification; Locaop, commubo, Web Assist and others are feature-launch releases with no adoption counts disclosed; the same PoC was announced separately by Verbex (#1506471) and TSK (#1506454, not adopted), treated by this card as one event.
• Scope of the mandate date: "the amended Comprehensive Labor Policy Promotion Act enacted in June 2025 makes kasuhara countermeasures mandatory for all employers from October 1, 2026, with no size-based grace period" is as stated in Quarq's release (#1372915), and Channel's release (#1481873) also premises the same mandate; this site has not checked the official statute. This card records only the stated facts and dates and provides no legal analysis, no explanation of the obligations' content, and no compliance advice.
• Second-hand press mentions: Uber's four-month budget burn and subsequent monthly caps, and Microsoft's termination of Claude Code direct licenses with consolidation into GitHub Copilot CLI, are both press reports cited inside Wonderful Fly's explainer (#1600742) (the source says "it was reported that"); this card always attaches the attribution and asserts no facts.
• Time point of carried figures: F-023 to F-027 are snapshots as recorded by the target card (published July 6, 2026); not new to this card, time points not updated.
• Taiwan line: a full-library search returned zero relevant hits, so this card connects no Taiwan line (see Selection method).
• Dead-link note: the original PR TIMES link for PRTIMES #1600742 (the Wonderful Fly explainer) is dead (checked 2026-07-27); the Sources list carries the web.archive.org snapshot (archived 2026-07-20, verified to open and to match the original title and publisher).

Risk factors

• Durability of effect numbers unverified: all effect figures in this card are single vendor-published adoption cases; whether third parties or the companies themselves will publish quantitative follow-ups is a tracking item (see P-001).
• Countermeasure rate around the mandate unknown: the 51.7% "no countermeasures" is a single-shot survey value as of June 2026; how it moves before and after the mandate date (October 1, 2026, per the release) awaits follow-up surveys (see P-002).
• Other risks (publication structure and interests / own-estimate scope / PoC stage / second-hand press mentions / carried time points): see Scope, not repeated.

FAQ

Q: What is the "AI Employee"? Who shipped it and when? What should one note about the publication structure?

The "AI Employee" is a Re:lation feature shipped by INGAGE Inc. on July 6, 2026: feed it the existing manuals written for new staff and it autonomously judges and handles customer contacts; select "AI Employee" in the case-owner field and it starts working, escalating complex cases to a human with a report attached (staffed escalation). What to note is the publication structure: the survey carrying "62% hope for an AI employee" (PRTIMES#1336094) and the product release (PRTIMES#1336163) came the same day from the same publisher, with the survey's numbers quoted inside the product release as development background.

That both releases are dated July 6, 2026 is a checkable fact; this card records the structure and interest relationship without asserting intent. The ship date is cross-pinned by the July 8, 2026 webinar notice (PRTIMES#1364414) (see Scope).

Q: What "effect numbers" exist so far for AI phone and customer support, and on what scope?

Three sets emerged within two weeks: Chiku Takubin's voicebot automated about 2,000 first-response inquiries per month (equivalent to about 1.5 operators), with per-call handling time down from an average of 150 seconds to 120 seconds (about 20% shorter) (PRTIMES#1373002); N-ASSET cut email drafting time by 50% versus pre-adoption with an AI mail agent (PRTIMES#1465644); Tohoku Electric Power, ten months after adopting a self-resolution AI, saw inbound call volume down about 20% per the release (PRTIMES#1360037). Scope: all are adoption cases published by the service providers themselves, not third-party audits; single cases cannot be extrapolated.

There is also BPO Japan's team-based handling of tens of thousands of emails per month (no ratio figure) (PRTIMES#1579406). The Chiku Takubin backdrop was inbound calls at about 200% of the prior year (specific year not stated in the source) (see Scope).

Q: What does kasuhara have to do with AI customer service, and when does the mandate start?

Channel Corporation's two surveys (497 workers + 514 consumers) record: 51.7% of surveyed companies have no kasuhara countermeasures; among consumers who actually encountered a store-side mistake (re-tabulated on 348), 52.3% gave up making a legitimate complaint for fear of "being seen as a harasser"; 65.2% are positive about a system in which AI delivers legitimate complaints on their behalf -- AI first response here gains the rationale of an "emotional-conflict buffer layer" (PRTIMES#1481873). The mandate date: per Quarq's release, kasuhara countermeasures become mandatory for all employers from October 1, 2026 (under the amended Comprehensive Labor Policy Promotion Act enacted in June 2025; official statute not checked by this site).

Channel is the provider of "Channel Talk" = an interested-party survey; this card records only the stated facts and dates and offers no legal commentary (PRTIMES#1372915; see Scope).

Q: How big is this market? What is the scope of the "about 19.1 billion yen"?

About 19.1 billion yen = Verbex's own estimate of the "voice-AI TAM in EC and mail order in 2030" (release title says about 19.0 billion yen): against the backdrop of the METI-survey figure it cites (Japan's BtoC EC market at 26.1 trillion yen in 2024, up 5.1% from the previous year, 2023), Verbex estimates the attached order-taking and customer-support market at about 300 billion yen, the voice-response segment at about 150 billion yen, and 17% voice-AI usage in 2030 -- not official statistics (PRTIMES#1360730).

Two more cited coordinates: Japan's call-center market above 1 trillion yen (Yano Research Institute 2025 survey, cited in the Helpfeel release) (PRTIMES#1360037); a 6.44 million-person labor shortage in 2030 (Persol Research Institute estimate, cited in the VALTEC release) (PRTIMES#1565138). Neither independently verified by this site (see Scope).

Q: What did the supply side do in the two weeks, and does it point one way?

At least six moves: the launch of miraio AI Main-Line Phone (hallucination suppression = vendor claim; first response and transfer marked "coming soon") (PRTIMES#1539860); commubo's generative-AI contextualization feature (real-time structuring of calls, cutting handoff workload) (PRTIMES#1462577); RightTouch "Web Assist" (AI operator grasps web behavior in real time, with data consolidated at handoff) (PRTIMES#1372747); Locaop Desk's web-complete self-serve sign-up (start as soon as same day) (PRTIMES#1566705); Helpfeel's autopilot concept plus AI-Ready consulting (Level 0-5 roadmap) (PRTIMES#1360037); the Verbex x TSK first-response phone PoC (instant FAQ answers, individual/specialized to humans) (PRTIMES#1506471). The direction is highly consistent: all embed the division "AI first response plus handoff to humans" -- isomorphic to the target card's thesis.

On the institutional side there are also JIIT's CCDX Guideline FY2026 edition (12 diagnostic axes) (PRTIMES#1508549) and VALTEC's IVR statistics upgrade (abandoned calls, AHT measurement) (PRTIMES#1565138) (see Scope).

Q: What is the cost antithesis, and how does it relate to Chapter One on this site?

On July 19, 2026, Wonderful Fly's explainer framed "token cost" as a new management issue: an AI agent calls generative AI multiple times to complete one task, so consumption easily exceeds expectations; it also cites reports that Uber burned through its annual AI-coding-tool budget in four months and that Microsoft consolidated Claude Code direct licenses into GitHub Copilot CLI (both second-hand mentions of press reports inside the company's explainer, not independent facts of this site) (PRTIMES#1600742).

Our Chapter One (ANK-2026-07-01-004) recorded "AI cost becoming a management issue"; this card connects it to the customer contact point: every "first response by AI" is token consumption, and the next gate for the division-of-labor thesis is cost governance (see Scope; J-002).

F-Units

F-001: INGAGE Inc. (headquarters: Osaka; President Wada Tetsuya) announced on July 6, 2026 the release of the Re:lation feature "AI Employee": feed it the existing manuals written for new operators and it autonomously judges and handles customer contacts; it starts working when "AI Employee" is selected in the case-owner field; sliders adjust personality settings such as politeness, reply speed, and emotional attentiveness; complex contacts it cannot resolve are immediately handed to a human with a report based on the interaction attached (staffed escalation); the release describes the development trigger as an internal experiment (deployment at the internal help window reached "a level where almost no human response was needed" = the publisher's own account, no quantitative figures) and frames the background problem as the maintenance burden of dictionary-type (scenario-type) chatbots; the product release re-quotes the company's own survey numbers (52% understaffed; 55% needing six months or more for new-hire independence; 63% over-dependence on individuals; about 1 in 5 facing 500-plus inquiries per month -- survey June 3-11, 2026, main survey of 484)

• source: PRTIMES #1336163
• source_url: https://prtimes.jp/main/html/rd/p/000000319.000029485.html
• confidence: high
• basis: official_statement
• period: published 2026-07-06 (product released same day)
• caveat: corporate self-published release; forms the "survey publisher = product vendor, same-day release" structure with the same company's survey (#1336094) published the same day (see Scope); "almost no human response needed" is the publisher's own account

F-002: INGAGE Inc. published a free-webinar notice on July 8, 2026: it states that "AI Employee" is the Re:lation feature released on July 6, 2026; the webinar was held online (Zoom) on Thursday, July 16, 2026, 12:00-12:30; the same notice claims over 6,000 companies using Re:lation (including trial use)

• source: PRTIMES #1364414
• source_url: https://prtimes.jp/main/html/rd/p/000000326.000029485.html
• confidence: high
• basis: official_statement
• period: published 2026-07-08; webinar 2026-07-16 (scheduled)
• caveat: corporate self-published release; "over 6,000 companies" is the vendor's own claim (including trials); the webinar was an announcement

F-003: Altius Link Co., Ltd. (Shibuya, Tokyo; President Naya Masatoshi) published on July 9, 2026 an adoption case at the customer center of Chiku Takubin Co., Ltd. (Nerima, Tokyo) for the voicebot "Virtual Agent Voice": first-response automation of about 2,000 inquiries per month (equivalent to about 1.5 operators); per-call handling time down from an average of 150 seconds to 120 seconds (about 20% shorter); a completion rate of about 80% maintained (the release carries a "*1" annotation whose definition is not expanded in the captured text); the release states that from the early adoption phase, work equivalent to about 1.5 operators was automated and answer rates improved

• source: PRTIMES #1373002
• source_url: https://prtimes.jp/main/html/rd/p/000000511.000019180.html
• confidence: medium
• basis: official_statement
• period: published 2026-07-09 (adoption start/end dates not stated in the source)
• caveat: a single adoption case published by the service provider (Altius Link) itself, not a third-party audit, not extrapolatable to average effects (see Scope)

F-004: The same release's background and method: the adoption backdrop was inbound calls reaching about 200% of the prior year (baseline is the prior year; specific year not stated) amid surging parcel volumes, plus chronic understaffing raising on-site load; the method was to sort inquiry content by direct-mail type, phase in from the comparatively easy inbound patterns, refine hearing items and voice guidance, and tune the machine voice's reading "pauses" for audibility

• source: PRTIMES #1373002
• source_url: https://prtimes.jp/main/html/rd/p/000000511.000019180.html
• confidence: medium
• basis: official_statement
• period: published 2026-07-09
• caveat: "about 200% of the prior year" is the release's wording for the adoption backdrop, specific year unstated; vendor adoption case (see Scope)

F-005: INGAGE Inc. published on July 13, 2026 an adoption case at N-ASSET Co., Ltd. (Kawasaki, Kanagawa; President Yoshimura Takashi): using Re:lation's "AI mail agent" (type the gist of a reply and appropriate text is auto-generated), email drafting time was cut 50% versus pre-adoption; personal email addresses were removed from business cards entirely and contact unified through Re:lation shared addresses; the release says complaint occurrence "plummeted" (qualitative, no figure); the release calls N-ASSET No. 1 in rental brokerage transactions in Takatsu ward, Kawasaki (survey body and time point not stated in the source)

• source: PRTIMES #1465644
• source_url: https://prtimes.jp/main/html/rd/p/000000321.000029485.html
• confidence: medium
• basis: official_statement
• period: published 2026-07-13 (adoption period not stated)
• caveat: an adoption case published by Re:lation's vendor INGAGE itself; the "50% cut" baseline is pre-adoption ("half the previous time"); "No. 1" is the release's wording (see Scope)

F-006: INGAGE Inc. published on July 17, 2026 an adoption case at BPO Japan Co., Ltd. (Shinjuku, Tokyo; President An Toshiaki; centered on recruitment process outsourcing, RPO): with Re:lation, tens of thousands of emails per month were moved to a team-based handling regime -- several hundred to over 1,000 emails a day processed with labels, auto-routing and owner assignment; at a scale of over 100 full-time employees, with multiple sites (Osaka, Sendai, Fukuoka and others) and remote members, response status is visualized in real time; personal-address usage was eliminated; the title's "toward zero mistakes" and the body's "achieved a zero-mistake regime" are the publisher's claims; the company voiced strong expectations for AI features (outlook)

• source: PRTIMES #1579406
• source_url: https://prtimes.jp/main/html/rd/p/000000322.000029485.html
• confidence: medium
• basis: official_statement
• period: published 2026-07-17 (adoption period not stated)
• caveat: vendor-published adoption case; "zero mistakes" is the publisher's claim, not an audit result (see Scope)

F-007: Channel Corporation (Chiyoda, Tokyo; provider of the AI agent "Channel Talk") published Survey 1 on July 14, 2026 (497 workers in service counters, customer support and similar; internet survey June 18-22, 2026; main contact settings = in-person store/counter 67%, phone/email/chat 29%, other 4%): the most common answer on "countermeasures your company has adopted against kasuhara" (options from the Tokyo Metropolitan customer-harassment prevention program's recommended measures) was "none" at 51.7%; top adopted measures were "publicizing a basic policy internally and externally" 27.4% and "creating and publicizing a kasuhara response manual" 24.9%; technical and environmental measures such as recording/filming or AI systems were under 10%; the release explicitly frames the October 2026 mandate as context

• source: PRTIMES #1481873
• source_url: https://prtimes.jp/main/html/rd/p/000000173.000029184.html
• confidence: high
• basis: official_statement
• period: survey 2026-06-18 to 22, published 2026-07-14
• caveat: the survey publisher is a vendor of related AI products (interested party); percentages represent only the 497-respondent population; the mandate date is as stated in the release (see Scope)

F-008: Same Survey 1, AI expectations: if an AI chat with first-response and complaint-filtering functions were introduced, the top expectation was "fewer occasions of directly receiving emotional complaints, easing the mental burden" at 28%; "first response reduced, letting me give attentive service to the customers who really matter" was chosen by 21.4% of email/inquiry-form staff and 10.7% of call-center/CS-counter staff (the release states both are about twice the overall average); about 60% of non-face-to-face counter staff answered "40% or more of the workload could be lightened by AI"

• source: PRTIMES #1481873
• source_url: https://prtimes.jp/main/html/rd/p/000000173.000029184.html
• confidence: medium
• basis: official_statement
• period: survey 2026-06-18 to 22, published 2026-07-14
• caveat: "about twice" is the release's own comparison against the overall average; "about 60%" and "40% or more" are the release summary's wording; percentages represent only the respondent population (see Scope)

F-009: The company's Survey 2 (514 general consumers in their 20s to 60s nationwide; internet survey June 12-14, 2026): 35.4% of all respondents (about 1 in 3.5) have given up pointing out a store's clear mistake (wrong orders, billing errors, damaged goods) or requesting a refund for fear of being seen as a "harasser" or "complainer" (frequently 7.8% + sometimes 17.1% + once or twice 10.5%); excluding the 27% who have never encountered a store-side mistake and re-tabulating on the 348 who actually did, 52.3% have this experience; the most common reason for not speaking up is "fear of sour faces or backlash" (release summary); for a system in which AI sits in between and delivers "legitimate complaints" to the store, "definitely want to use it" 22.6% + "would use it depending on content" 42.6% = 65.2% positive

• source: PRTIMES #1481873
• source_url: https://prtimes.jp/main/html/rd/p/000000173.000029184.html
• confidence: high
• basis: official_statement
• period: survey 2026-06-12 to 14, published 2026-07-14
• caveat: 348 is the publisher's re-tabulated population of "those who actually encountered a mistake" (the denominator of 52.3%, stated in the source); percentages represent only the 514-respondent population (see Scope)

F-010: Quarq Co., Ltd. (Shibuya, Tokyo; President Maekawa Hideyuki) announced on July 9, 2026: the "Customer Harassment Countermeasure E-Learning Package", in partnership with Elephancube Inc. (Shinjuku, Tokyo; cumulative e-learning production record of over 2,000 titles, 200-plus per year = per the release), goes on sale July 15, 2026; the course is eight chapters with a total running time of 1 hour 13 minutes plus a ten-question comprehension test, with progress and completion-history management on the LMS "Qualif"; the release states that the amended Comprehensive Labor Policy Promotion Act enacted in June 2025 makes customer-harassment countermeasures mandatory for all employers from October 1, 2026, with no grace period by company size

• source: PRTIMES #1372915
• source_url: https://prtimes.jp/main/html/rd/p/000000012.000145914.html
• confidence: medium
• basis: official_statement
• period: published 2026-07-09; on sale 2026-07-15; mandate date 2026-10-01 (as stated in the release)
• caveat: the mandate and law-related passages are as stated in the release; this site has not checked the official statute; this card records facts and dates only and provides no legal commentary (see Scope)

F-011: Upsell Technologies Co., Ltd. (Toshima, Tokyo; CEO and Group Chairman Takahashi Ryota) and pluszero Inc. (Setagaya, Tokyo) announced on July 16, 2026 the launch of their co-developed AI operator "miraio AI Main-Line Phone": built on Upsell Technologies' self-claimed 20-plus years and cumulative 8,200-plus companies of call-center operating know-how and the next-generation AI engine "PUT" based on pluszero's proprietary concept "AEI (Artificial Elastic Intelligence)"; its pitch is suppressing hallucination risk to the limit with a dual-process AEI model that never returns LLM output directly (vendor claim), and natural dialogue via cloned voices of real human operators; functions include real-time transcription and summarization of incoming calls with automatic notification (Slack, Teams, Lark, email and others), automatic screening and blocking of sales calls, autonomous FAQ answers, and emergency escalation; "first response and transfer (24/7/365 capable)" is explicitly marked "coming soon"

• source: PRTIMES #1539860
• source_url: https://prtimes.jp/main/html/rd/p/000000070.000030394.html
• confidence: medium
• basis: official_statement
• period: published 2026-07-16 (launch; first response/transfer marked coming soon)
• caveat: "8,200-plus companies" and hallucination suppression are vendor self-claims; "first response and transfer" was not yet available at publication (see Scope); duplicate release #1539863 counted once

F-012: Softfront Japan (Chiyoda, Tokyo; subsidiary of Softfront Holdings) announced on July 13, 2026 the release of the generative-AI-type "contextualization feature" for the AI voicebot "commubo": call content is extracted and structured in real time into pre-set fields (name, address, inquiry content, preferred dates and the like), cutting the workload of data shaping, operator handoff, and content review (after-call work, ACW); supports slot filling (auto-organizing out-of-order utterances, follow-up questions for missing items, handling corrections and restatements), call summarization, and kanji/kana spelling confirmation; provided as a standard feature that users configure themselves with immediate effect; the feature existed for scenario-type bots and was extended this time to the generative-AI type

• source: PRTIMES #1462577
• source_url: https://prtimes.jp/main/html/rd/p/000000121.000045699.html
• confidence: high
• basis: official_statement
• period: published 2026-07-13
• caveat: corporate self-published release; feature descriptions are the publisher's own account, with no adoption counts or quantitative effect figures

F-013: RightTouch Inc. (Shinagawa, Tokyo; carved out of PLAID Inc., TSE Growth 4165) announced on July 9, 2026 the launch of "Web Assist", a new feature of QANT Speak: the AI operator grasps the customer's web-behavior data (past and real-time) during a call and supports on-screen operations by voice, aiming to prevent mid-procedure drop-off, raise self-resolution rates, and reduce repeat calls (vendor description); supports both directions -- phone-to-web guidance (URL sent by SMS with the voice connection maintained) and web-to-phone (pre-entered questions and web-behavior data handed to the AI); when the AI cannot resolve, the conversation log and web-behavior data are consolidated on the operator dashboard "Connect Board" for handoff to a human; the release also states the QANT series recorded intent-based routing accuracy of 99.3% in a verification case

• source: PRTIMES #1372747
• source_url: https://prtimes.jp/main/html/rd/p/000000114.000098678.html
• confidence: medium
• basis: official_statement
• period: published 2026-07-09
• caveat: the feature's effects (drop-off prevention, self-resolution gains, fewer repeat calls) are the vendor's stated aims, not measured disclosures; 99.3% is a release-stated figure from a single verification case (see Scope)

F-014: Locaop Inc. (Osaka; President Agata Masataka) announced on July 17, 2026 self-serve sign-up for the AI phone reception service "Locaop Desk": the adoption process, previously premised on a sales meeting, now completes on the web -- application, online identity verification, and credit-card payment (Stripe; monthly or annual) -- with use starting as soon as the same day once prescribed checks and setup are complete; the entry point is a "14-day free trial", and if not upgraded to the standard plan within the trial the issued phone number is decommissioned

• source: PRTIMES #1566705
• source_url: https://prtimes.jp/main/html/rd/p/000000122.000040797.html
• confidence: high
• basis: official_statement
• period: published 2026-07-17
• caveat: corporate self-published release; "as soon as same day" is conditional on completing prescribed checks and setup; no adoption counts disclosed

F-015: Helpfeel Inc. (Kyoto; CEO Rakusai Isshu) announced on July 8, 2026 its "autopilot call center (AI autonomous operation)" concept and the start of an AI-Ready consulting business: a six-stage Level 0-5 roadmap by AI maturity (Level 5 = an autonomous AI agent handles all work including discretionary judgments and exceptions, with humans supervising and managing the AI); in the lead case, Tohoku Electric Power Co., Inc. (Sendai, Miyagi) adopted the self-resolution AI "Helpfeel" in March 2025, and the release records 160,000 monthly page views ten months after adoption, inbound call volume down about 20% (baseline time point not stated verbatim; context is a pre/post-adoption comparison), and substantially improved answer rates; Tohoku Electric Power has completed Level 1 (building a knowledge base used by people and AI) and entered Level 2 (AI that assists people); Helpfeel claims over 900 sites adopted (as of April 2026)

• source: PRTIMES #1360037
• source_url: https://prtimes.jp/main/html/rd/p/000000506.000027275.html
• confidence: medium
• basis: official_statement
• period: published 2026-07-08; Tohoku Electric Power adoption 2025-03, figures for the ten months after
• caveat: the Tohoku Electric Power figures are as stated in Helpfeel's (the service provider's) release, not Tohoku Electric Power's own publication; "over 900 sites" is a vendor self-claim (see Scope)

F-016: Three sets of market statistics re-cited in the same release (all second-hand citations inside the release): Japan's call-center market exceeds 1 trillion yen (Yano Research Institute, "Call Center Services Market / Contact Center Solutions Market Survey (2025)"); operator hiring "very difficult at all sites" 35% + "very difficult at some sites" 41% = about 80% of companies struggling to secure staff, and call-center answer rates down for the third consecutive year (both from Ric Telecom, "Call Center White Paper 2025"; covered years not expanded in the release); the share of Japanese companies saying generative-AI results "exceeded expectations" is one quarter that of Western countries (PwC Japan Group, "Generative AI Survey Spring 2025, five-country comparison")

• source: PRTIMES #1360037
• source_url: https://prtimes.jp/main/html/rd/p/000000506.000027275.html
• confidence: medium
• basis: official_statement
• period: survey years as noted (2025); re-cited in the 2026-07-08 release
• caveat: all are third-party survey figures re-cited inside the release; this site has not checked the original surveys (see Scope); the "one quarter" comparison is against Western countries (stated in the same sentence)

F-017: Verbex Inc. announced a voice-AI market forecast on July 8, 2026: citing METI's "FY2024 (Reiwa 6) Digital Transaction Environment Improvement Project (e-commerce market survey)", it states Japan's BtoC EC market at 26.1 trillion yen in 2024, up 5.1% from the previous year (2023) (Verbex citing the METI survey; original survey not checked by this site); by its own estimate, Verbex puts the order-taking and customer-support market attached to EC and mail order at about 300 billion yen, of which the voice-response segment (phone orders and inquiries) is about 150 billion yen; it forecasts that by 2030 voice AI will be used in 17% of that segment, expanding the EC/mail-order voice-AI TAM to about 19.1 billion yen (the release title says "about 19.0 billion yen"); it presents three diffusion steps (first response, callbacks and FAQ; structured order-taking and repeat orders; hybrid operation with AI-human division of roles); Verbex claims 56 patents in 25 countries

• source: PRTIMES #1360730
• source_url: https://prtimes.jp/main/html/rd/p/000000017.000172103.html
• confidence: medium
• basis: official_statement
• period: METI survey covers 2024; TAM forecast targets 2030; published 2026-07-08
• caveat: the roughly 300 billion, 150 billion, 17%, and about 19.1 billion yen figures are all company estimates/forecasts, not official statistics; "previous year = 2023" is a year-notation transcription (the source says "the previous year") (see Scope)

F-018: Verbex Inc. announced on July 15, 2026 the start of a proof of concept (PoC) with San-in Chuo Television Broadcasting Co., Ltd. (TSK, Matsue, Shimane; President Tanabe Choemon): for inquiries to TSK's new business launched in May 2026, the "TSK Brand Item Home-Pickup Purchase Service", Verbex's voice AI handles first-response phone reception -- routine inquiries answerable from FAQs (application method, shipping method, service content) are answered instantly, while matters involving individual circumstances or requiring specialized judgment are not completed by the AI but guided to the appropriate contact; operation begins as internal verification, moving in stages to customer-facing use after reviewing actual inquiries and operational issues; the two companies have partnered from FY2026 to promote voice-AI adoption among municipalities and regional companies centered on the San-in region

• source: PRTIMES #1506471
• source_url: https://prtimes.jp/main/html/rd/p/000000018.000172103.html
• confidence: high
• basis: official_statement
• period: published 2026-07-15; target service launched May 2026
• caveat: PoC does not equal formal adoption; starts with internal verification; the same PoC was announced separately by the TSK side (#1506454, counted as one event, not adopted) (see Scope)

F-019: VALTEC Co., Ltd. (Shinjuku, Tokyo) announced on July 17, 2026 an upgrade to the IVR (interactive voice response) statistics and reporting of MOT/CallCenter Lite: added aggregation of IVR abandoned calls (abandonment counts and rates), elimination of "IVR unknown" labels so all calls are classified and counted correctly, accurate counting of unanswered extension transfers, and improved call-time aggregation (accurate AHT, average handling time); the release re-cites the Persol Research Institute estimate of "a 6.44 million-person labor shortage in 2030"; it also records, as a general IVR adoption example, average daily incoming calls falling from 347.8 to 272.5 (about a 20% cut; baseline is the pre-adoption 347.8 calls/day)

• source: PRTIMES #1565138
• source_url: https://prtimes.jp/main/html/rd/p/000000893.000008585.html
• confidence: medium
• basis: official_statement
• period: published 2026-07-17; the 6.44 million figure is a 2030 estimate (Persol Research Institute, cited)
• caveat: the 6.44 million figure is a second-hand citation inside the release, original estimate not checked by this site; 347.8 to 272.5 is a vendor-stated adoption example (see Scope)

F-020: The Japan Institute of Information Technology (JIIT; Minato, Tokyo; chairman Akasaka Yuji, chairman of the board of Japan Airlines; founded July 16, 1981; 241 member companies) announced on July 15, 2026 the free publication of the "Contact Center DX Promotion Guideline, FY2026 edition": DX maturity is assessed on 12 diagnostic axes composed from the four perspectives of customer, employee, management, and society, organizing roadmap design, KPI setting, and industry- and scale-specific promotion points; the goal is to transform contact centers into "strategic bases for customer value creation" and to serve as a common DX language from executives to front-line staff; FOXeey Consulting Inc. (founded April 1, 2025) was lead organizer, drafted with the cooperation of the CCDX Guideline Drafting Committee (composed of contact-center operators, experts and others)

• source: PRTIMES #1508549
• source_url: https://prtimes.jp/main/html/rd/p/000000119.000006381.html
• confidence: high
• basis: official_statement
• period: published 2026-07-15
• caveat: the publisher is a public-interest incorporated association, but this card's capture is the PR TIMES release, not the association's original document; official anchor unverified (see Scope)

F-021: pickupon Inc. (President Obata Yoichi) announced on July 17, 2026 the launch of a hands-on "DX Human Resource Training": using the client company's own phone, negotiation, and meeting conversation data as teaching material, participants build in practice the design of AI summarization rules, CRM/SFA recording methods, team review flows, and numerical results checks (deal-conversion rate, win rate, handling time, missed-record counts, per company-set targets); four courses (AI article DX / product & service improvement DX / risk analysis & crisis management DX / internal feedback reinforcement DX), each about 13 hours (e-learning plus live online sessions), minimum three participants per course; the release notes the possibility of applying the Human Resource Development Support Grant to training costs and the Digitalization & AI Adoption Subsidy 2026 to new adoption of the AI cloud IP phone "pickupon" (regular track 50,000 yen to under 1.5 million yen, subsidy in principle up to one half; applicability varies by company size, business, training content, application timing and more = stated in the release); a partnered labor and social security attorney supports requirement checks, training-plan organization, and document preparation

• source: PRTIMES #1568297
• source_url: https://prtimes.jp/main/html/rd/p/000000171.000033268.html
• confidence: high
• basis: official_statement
• period: published 2026-07-17
• caveat: grants and subsidies are described as "possibilities" under the scheme; individual applicability depends on screening; the same-entity follow-up to target card F-026 (the company's subsidy-tool registration effective 2026-06-12)

F-022: Wonderful Fly Co., Ltd. (Chuo, Tokyo; President Hirata Masako) published an explainer on July 19, 2026: it frames "token cost" as an emerging management issue -- an AI agent calls generative AI multiple times to complete one task (reading the inquiry, judging intent, searching internal information, drafting the answer, checking the answer), so what looks like a single operation to the user consumes more tokens than expected across multiple AI processes; cost drivers include feeding long documents and past conversation history every time, using high-performance models even for simple classification and extraction, uncontrolled retry counts on errors, and not tracking consumption per agent; it proposes four management points (setting token usage caps, visualizing usage and cost, per-agent usage analysis, dividing work with workflows and rules); citing press reports, it states that in 2026 Uber used up its annual AI-coding-tool budget in just four months and later set monthly caps per tool, and that Microsoft terminated many direct licenses for Anthropic's "Claude Code", which had spread rapidly internally, consolidating usage into GitHub Copilot CLI (both second-hand mentions of press reports in the company's explainer; the source says "it was reported that")

• source: PRTIMES #1600742
• confidence: medium
• basis: official_statement
• period: published 2026-07-19 (today)
• caveat: the Uber and Microsoft examples are second-hand mentions of press reports inside the explainer, not independent facts of this site; this card makes no factual assertion of them (see Scope); the explainer's publisher sells AI-agent construction support services

F-023: [Carried from target card, not new] ielove GROUP "Survey on AI Phone Response" (online questionnaire May 25 to June 8, 2026; 1,108 valid responses = 67 real-estate companies + 1,041 end users): 60.8% of surveyed users have had a call answered by AI (automated voice) ("no" 31.3%, "not sure" 7.9%), and 41.6% feel uneasy about AI response; among experienced users (n=633, multiple answers) the top merit is "24-hour availability" 59.7%; complaints are led by "cannot handle detailed consultations" 61.5% and "the conversation doesn't get through" 57.2% (the complaint question's population is not restated verbatim in the source); 80.6% of real-estate companies still handle phones with their own staff, and 70.4% answered that AI phone response will spread (the question's target is not stated verbatim in the source)

• source: PRTIMES #1337119
• source_url: https://prtimes.jp/main/html/rd/p/000000805.000008550.html
• confidence: high
• basis: official_statement
• period: survey 2026-05-25 to 06-08, published 2026-07-06 (target-card snapshot)
• caveat: the population is readers of the company's own media, SNS followers, and employees of firms using its services -- not a national random sample; details in the target card (see Scope)

F-024: [Carried from target card, not new] INGAGE "Survey on AI Use in Inquiry Handling" (online, June 3-11, 2026; screening 5,000 -> main survey 484): 55% of sites need "six months or more" for new hires to work independently; 64% feel veterans' and leaders' time is eaten by training; 63% see over-dependence on individuals as an issue; 52% feel understaffed; 22% of surveyed sites face 500-plus inquiries per month; attitudes to AI adoption split positive 51% vs. cautious 49%; adoption barriers are "no IT-savvy staff" 34% and "fear of AI giving wrong answers" 32%, with adoption cost 26% in third place; 50% do not use AI at work at all; 76% value a "smooth handoff to humans" function; expectation for the ideal division "AI handles first response, humans take over complex cases" is 62% (the publisher also phrases this as 62% hoping for an autonomously responding "AI employee")

• source: PRTIMES #1336094
• source_url: https://prtimes.jp/main/html/rd/p/000000320.000029485.html
• confidence: high
• basis: official_statement
• period: survey 2026-06-03 to 11, published 2026-07-06 (target-card snapshot)
• caveat: the publisher is the vendor of the CS platform Re:lation; percentages represent only the 484-respondent population; "AI employee" is the publisher's framing term; details in the target card (see Scope)

F-025: [Carried from target card, not new] GMO Pepabo, Inc. began offering "GMO Sokuresu AI for CS" on July 1, 2026: one-stop support including FAQ auto-generation, an operations dashboard/analytics, and seamless connection to human response; the background narrative flags the on-site problems of "adopted but unused" and "operations that don't run", with the focus shifting from "adoption" to the phase of "keeping operations running" (the publisher's background description, not statistics)

• source: PRTIMES #1281884
• source_url: https://prtimes.jp/main/html/rd/p/000005424.000000136.html
• confidence: high
• basis: official_statement
• period: offering began 2026-07-01 (published same day) (target-card snapshot)
• caveat: a product release; functions and background problems are the publisher's own description; details in the target card

F-026: [Carried from target card, not new] pickupon Inc. was selected, effective June 12, 2026, as an IT adoption support operator under the "Digitalization & AI Adoption Subsidy 2026" (formerly the IT Adoption Subsidy), and the AI phone service "pickupon" was registered as an eligible tool: qualifying SMEs and small businesses can receive, under the regular track (50,000 yen to under 1.5 million yen), a subsidy of in principle up to one half of adoption cost; from 2026 the subsidy was revised to put more weight on AI-enabled tools

• source: PRTIMES #1308694
• source_url: https://prtimes.jp/main/html/rd/p/000000170.000033268.html
• confidence: high
• basis: official_statement
• period: selection effective 2026-06-12, published 2026-07-03 (target-card snapshot)
• caveat: the subsidy rate and range describe the regular track; individual application depends on screening; registration is not an adoption record; details in the target card

F-027: [Carried from target card, not new] C&T Co., Ltd. announced it would exhibit and demonstrate, at the "International Modern Hospital Show 2026" (July 8-10, 2026, Tokyo Big Sight), the Google Maps-linked AI interpreting phone "tel-trans" (the release title says "first public showing") and the medical-facility "tel-trans QR": foreign patients can pick a hospital on Google Maps and start an AI-interpreted call in their native language, with 149 languages supported and browser-only use

• source: PRTIMES #1337255
• source_url: https://prtimes.jp/main/html/rd/p/000000007.000178687.html
• confidence: high
• basis: official_statement
• period: published 2026-07-06, exhibition 2026-07-08 to 10 (planned) (target-card snapshot)
• caveat: the exhibition/demonstration was a planned event; specifications such as 149 languages are the vendor's own claims; details in the target card

J-Units

J-001: Official-anchor scope (single consolidated block is the Scope section; this unit only adds source grading, without repetition): official_count_verified=0, official_count_raw=0, official_attribution_unverified_count=0. All 23 source items are corporate or organizational self-published releases on the PR TIMES platform (platform-reviewed corporate-release grade); #1508549's publisher is a public-interest incorporated association (JIIT), but the capture is the PR TIMES release, not the association's original document. All figures and dates inside releases attributed to METI, Persol Research Institute, Yano Research Institute, Ric Telecom, PwC Japan, MHLW, the Tokyo Metropolitan Government, or the amended Comprehensive Labor Policy Promotion Act are second-hand citations or mentions; this site has not checked the originals. This card attaches no Wikidata Q identifiers.

• confidence: high
• basis: official_statement

J-002: Five lines of advance for the "AI for first response, humans for complex cases" scheme (structural synthesis only; no new numbers generated, no causality built, confined to cited facts): (1) hope to product -- the "AI employee" that 62% hoped for was shipped the same day by the survey's publisher, with staffed escalation built into the product design (F-001, F-024); (2) effect numbers -- all three vendor adoption cases are scoped to "first response" (about 2,000 calls/month, about 1.5 operators, about 20% shorter; 50% cut; inbound volume down about 20%) (F-003, F-005, F-015); (3) the kasuhara axis -- a third survey extends the division-of-labor thesis from "efficiency" to "worker protection and the voice of the customer" (51.7% no countermeasures; 52.3% swallowed complaints; 65.2% positive on AI mediation) (F-007 to F-009), with the institutional clock of the October 1, 2026 mandate (as stated in releases) (F-010); (4) supply and institutional coordinates -- all six product moves embed the "AI first response plus handoff to humans" division (F-011 to F-015, F-018), and market coordinates appeared (TAM about 19.1 billion yen own estimate; 6.44 million cited) (F-017, F-019); (5) the cost antithesis -- every "first response by AI" is token consumption, connecting the division-of-labor thesis to Chapter One's cost-governance axis (F-022). The five lines are mutually independent; this card asserts no causality among them; the above is this card's editorial reading.

• confidence: medium
• basis: official_statement

J-003: Layer labels and boundaries: Layer 1 CONNECT + CHALLENGE = PRTIMES#1336163 (62% hope to same-day product; publication structure recorded without imputing intent); Layer 2 CORROBORATE = #1364414 (second release pins the July 6 ship date); Layers 3-4 ADD_EVIDENCE = #1373002 (Chiku Takubin effect numbers and background); Layer 5 ADD_EVIDENCE = #1465644, #1579406 (N-ASSET 50% cut; BPO Japan tens of thousands monthly); Layer 6 CORROBORATE = #1481873 (third independent survey, kasuhara axis); Layer 7 CONTEXTUALIZE = #1372915 (mandate date recorded; countermeasures commodified); Layers 8-11 ADD_EVIDENCE = #1539860, #1462577, #1372747, #1566705 (supply wave, four products); Layer 12 CONTEXTUALIZE = #1360037 (autopilot concept, Tohoku Electric Power, three cited statistics); Layer 13 CONTEXTUALIZE = #1360730 (TAM about 19.1 billion yen own estimate); Layer 14 CORROBORATE = #1506471 (TSK PoC = the thesis in the field); Layer 15 CONTEXTUALIZE = #1565138 (IVR statistics, 6.44 million cited); Layer 16 CONTEXTUALIZE = #1508549 (CCDX Guideline); Layer 17 CONNECT = #1568297 (same-entity follow-up to target card F-026); Layer 18 CHALLENGE + CONNECT = #1600742 (token-cost antithesis, Chapter One link-back); Layer 19 CONNECT = two published cards. F-023 to F-027 are the carried base, marked "carried from target card, not new", excluded from significance. Other desks' drafts today (ANK-2026-07-19-001 through 006) were unpublished at press time; per the "link only published cards" rule no links are set, stated here only.

• confidence: high
• basis: official_statement

P-Units

P-001: Adoption record and quantified effects of the "AI Employee" itself -- currently only the publisher PM's qualitative wording ("visible, major effects") and the vendor's cases for other products; tracking the feature's own quantitative disclosure (open)

P-002: How the 51.7% "no countermeasures" figure moves around the mandate date (October 1, 2026, per the release) -- verifiable by follow-up surveys; this card does not forecast (open)

P-003: The gap between Verbex's TAM of about 19.1 billion yen (2030, own estimate) and actual adoption -- whether the TSK PoC, starting from internal verification, moves to formal customer-facing use and formal adoption is verifiable in future releases (open)

P-004: The formal availability date of miraio's "first response and transfer" (marked "coming soon") -- verifiable by official announcement (open)

P-005: Whether token-cost governance produces quantified cases in customer-support AI (per-contact cost, cap-operation records) -- this card's sources are an explainer and second-hand press mentions only; awaiting disclosure by the companies concerned (open)

Significance self-check (extension-lane hard check)

From 18 new materials this card adds 22 entirely new traceable F-Units (F-001 to F-022), plus 5 base F-Units carried from the target card (F-023 to F-027, marked carried, not counted as new), for a total of 27 (target card: 12); article sources are 23 = 18 new + 5 carried (target card: 5); the internal citation chain counts 3 cards including the target card, and still 2 published cards after excluding the target card (ANK-2026-07-01-004, ANK-2026-07-03-018) (target card: 2) -- new F-Units 22>12, sources 23>5, internal links 3>2: all three exceed the target card and none falls below it; this is not a rewrite of the target card.

同事件・三視角 / Three Perspectives on the Same Event / 同一イベント・三つの視点

• 繁體中文: ANK-2026-07-19-007
• 日本語: ANK-2026-07-19-007-ja
• English: ANK-2026-07-19-007-en

Internal citation chain

Published ANK-Docs cited by this article:

• ANK-2026-07-06-010 (Workplace AI, Chapter Three -- The Customer Contact Point: What AI Phone Response Delivers and Where It Stops -- 60.8% of Surveyed Users Have Been Answered by AI (Automated Voice); Top Merit '24-Hour Availability' at 59.7% and Top Complaints 'Cannot Handle Detailed Consultations' at 61.5%; 55% of Support Sites Say New Hires Need 'Six Months or More' to Work Independently, 76% Value 'Smooth Handoff to Humans', 62% Expect an 'AI Employee' -- the 'AI for First Response, Humans for Complex Cases' Division of Labor) -- this card is its extension (target card): Layers 1-2 advance its "62% expect an AI employee" to the same-day product shipped by the survey's publisher; Layers 3-5 add the effect numbers its supply side lacked; Layers 6-7 add the kasuhara axis and mandate date it entirely lacked; Layers 8-16 add the supply wave and market/institutional coordinates; Layer 17 follows the same entity as its F-011 subsidy axis; Layer 18 raises the cost antithesis with today's token-cost explainer. Links: zh / ja / en
• ANK-2026-07-01-004 (Workplace AI Chapter One: the "Big Three era" in tool share and AI cost becoming a "management issue" -- numbers in that card) -- this card's Layer 18 token-cost explainer (PRTIMES#1600742) extends that card's cost axis into the customer-contact domain; the two cards differ in materials and populations, directional linkage only, no number comparison. https://ainews.washinmura.jp/ainews/en/ank/ANK-2026-07-01-004
• ANK-2026-07-03-018 (Workplace AI Chapter Two: the "ledger of realized gains" -- the Kokuyo Group/Kaunet employee survey recording an average of about five hours saved per respondent per month; numbers in that card) -- this card's Layers 3-5 adoption-case effect numbers are the next page of the same realized-gains-ledger methodology at customer-support sites; that card is a single organization's employee survey while this card is vendor adoption cases; different scopes, no summing, no comparison. https://ainews.washinmura.jp/ainews/en/ank/ANK-2026-07-03-018

Sources

1. [PRTIMES #1336163] INGAGE Inc., English translated title: "INGAGE releases 'AI Employee', a Re:lation feature that responds autonomously just by being handed your manuals" (corporate release, in Japanese), 2026-07-06. https://prtimes.jp/main/html/rd/p/000000319.000029485.html
2. [PRTIMES #1364414] INGAGE Inc., English translated title: "[July 16] INGAGE holds a next-generation customer-response webinar using the new 'AI Employee' feature" (corporate release, in Japanese), 2026-07-08. https://prtimes.jp/main/html/rd/p/000000326.000029485.html
3. [PRTIMES #1373002] Altius Link Co., Ltd., English translated title: "Altius Link deploys a voicebot at Chiku Takubin's customer center" (corporate release / adoption case, in Japanese), 2026-07-09. https://prtimes.jp/main/html/rd/p/000000511.000019180.html
4. [PRTIMES #1465644] INGAGE Inc., English translated title: "[Adoption case] N-ASSET cuts email drafting time by 50% with Re:lation's AI" (corporate release / adoption case, in Japanese), 2026-07-13. https://prtimes.jp/main/html/rd/p/000000321.000029485.html
5. [PRTIMES #1579406] INGAGE Inc., English translated title: "[Adoption case] BPO Japan moves tens of thousands of monthly emails to team handling with Re:lation, toward zero mistakes" (corporate release / adoption case, in Japanese), 2026-07-17. https://prtimes.jp/main/html/rd/p/000000322.000029485.html
6. [PRTIMES #1481873] Channel Corporation, English translated title: "51.7% of companies have no kasuhara countermeasures; 52.3% of consumers who experienced store mistakes gave up complaining for fear of 'being seen as harassers' -- two surveys ahead of the October mandate" (corporate release / surveys, in Japanese), 2026-07-14. https://prtimes.jp/main/html/rd/p/000000173.000029184.html
7. [PRTIMES #1372915] Quarq Co., Ltd., English translated title: "Kasuhara countermeasures become mandatory for all companies from October 1, 2026. Quarq partners with Elephancube to sell a countermeasure e-learning package" (corporate release, in Japanese), 2026-07-09. https://prtimes.jp/main/html/rd/p/000000012.000145914.html
8. [PRTIMES #1539860] Upsell Technologies Co., Ltd., English translated title: "AI operator 'miraio AI Main-Line Phone' launched: hallucinations suppressed to the limit, reception automated with natural human-like dialogue" (corporate release, in Japanese), 2026-07-16. https://prtimes.jp/main/html/rd/p/000000070.000030394.html
9. [PRTIMES #1462577] Softfront Holdings (Softfront Japan), English translated title: "AI voicebot 'commubo' releases a generative-AI contextualization feature converting call content into structured data in real time" (corporate release, in Japanese), 2026-07-13. https://prtimes.jp/main/html/rd/p/000000121.000045699.html
10. [PRTIMES #1372747] RightTouch Inc., English translated title: "RightTouch launches 'Web Assist' on QANT Speak, letting the AI operator guide customers while grasping their web behavior" (corporate release, in Japanese), 2026-07-09. https://prtimes.jp/main/html/rd/p/000000114.000098678.html
11. [PRTIMES #1566705] Locaop Inc., English translated title: "AI phone reception 'Locaop Desk' launches self-serve sign-up completing everything from application to start of use on the web" (corporate release, in Japanese), 2026-07-17. https://prtimes.jp/main/html/rd/p/000000122.000040797.html
12. [PRTIMES #1360037] Helpfeel Inc., English translated title: "Launching a new initiative to support organizational AI-Readiness, indispensable to realizing the 'autopilot call center (AI autonomous operation)'" (corporate release, in Japanese), 2026-07-08. https://prtimes.jp/main/html/rd/p/000000506.000027275.html
13. [PRTIMES #1360730] Verbex Inc., English translated title: "[Voice-AI market forecast report] Japan's EC/mail-order voice-AI market to reach about 19.0 billion yen by 2030" (corporate release / own estimate, in Japanese), 2026-07-08. https://prtimes.jp/main/html/rd/p/000000017.000172103.html
14. [PRTIMES #1506471] Verbex Inc., English translated title: "Verbex provides voice AI for first-response phone reception at TSK San-in Chuo Television, starting a proof of concept" (corporate release, in Japanese), 2026-07-15. https://prtimes.jp/main/html/rd/p/000000018.000172103.html
15. [PRTIMES #1565138] VALTEC Co., Ltd., English translated title: "Accelerating call-center DX in the era of labor shortage: IVR statistics and reporting functions reinforced" (corporate release, in Japanese), 2026-07-17. https://prtimes.jp/main/html/rd/p/000000893.000008585.html
16. [PRTIMES #1508549] Japan Institute of Information Technology (JIIT), English translated title: "'Contact Center DX Promotion Guideline, FY2026 edition' published" (organizational release, in Japanese), 2026-07-15. https://prtimes.jp/main/html/rd/p/000000119.000006381.html
17. [PRTIMES #1568297] pickupon Inc., English translated title: "Turning customers' raw voices into on-site improvement: pickupon launches hands-on 'DX Human Resource Training'" (corporate release, in Japanese), 2026-07-17. https://prtimes.jp/main/html/rd/p/000000171.000033268.html
18. [PRTIMES #1600742] Wonderful Fly Co., Ltd., English translated title: "What costs should you check before adopting an AI agent?" (corporate release / explainer, in Japanese), 2026-07-19. https://web.archive.org/web/20260720045117/https://prtimes.jp/main/html/rd/p/000000041.000159668.html
19. [PRTIMES #1337119] ielove GROUP Co., Ltd., English translated title: "AI phone response: top merit is '24-hour availability', while 'cannot handle detailed consultations' remains an issue | ielove survey" (carried from target card; corporate release / survey, in Japanese), 2026-07-06. https://prtimes.jp/main/html/rd/p/000000805.000008550.html
20. [PRTIMES #1336094] INGAGE Inc., English translated title: "A majority say new hires need 'six months or more' to work independently: surveying the training-cost limits hitting understaffed customer-support sites and the real expectations for 'AI use'" (carried from target card; corporate release / survey, in Japanese), 2026-07-06. https://prtimes.jp/main/html/rd/p/000000320.000029485.html
21. [PRTIMES #1281884] GMO Pepabo, Inc., English translated title: "New customer-support/helpdesk solution 'GMO Sokuresu AI for CS' launched July 1" (carried from target card; corporate release, in Japanese), 2026-07-01. https://prtimes.jp/main/html/rd/p/000005424.000000136.html
22. [PRTIMES #1308694] pickupon Inc., English translated title: "'pickupon' registered as an eligible tool under the Digitalization & AI Adoption Subsidy 2026" (carried from target card; corporate release, in Japanese), 2026-07-03. https://prtimes.jp/main/html/rd/p/000000170.000033268.html
23. [PRTIMES #1337255] C&T Co., Ltd., English translated title: "First public showing of 'tel-trans', an AI interpreting phone that calls hospitals in your native language from Google Maps" (carried from target card; corporate release, in Japanese), 2026-07-06. https://prtimes.jp/main/html/rd/p/000000007.000178687.html