Nine new deployments landed today — a German grocer selling inside ChatGPT, a Brazilian telecom's first ChatGPT app, a Japanese prefecture's AI concierge, and more — plus twelve follow-up reviews, including one where a bank's own efficiency numbers run headlong into a real Trustpilot complaint.

New launches

Enterprise tier

Germany's Knuspr becomes one of Europe's first grocers to sell inside ChatGPT.
Knuspr is the German online supermarket arm of Rohlik Group, a major European online grocery delivery company.
Knuspr launched an official ChatGPT integration letting customers search products, discover recipes, manage their basket, and plan weekly groceries through natural conversation, including uploading photos of handwritten shopping notes. It also built its own MCP server so the same catalog can be used by other AI agents.
How to try it: Open the Knuspr app inside ChatGPT; the catalog is also available via Knuspr's MCP server at mcp.knuspr.de.
Sources: retail-news.de, Lebensmittel Zeitung

France's largest estate agency network, Orpi, plugs its full property catalog into ChatGPT and Claude.
Orpi is France's largest real estate agency network, a 1,250+ agency cooperative.
Orpi and AI vendor Kleio launched a conversational AI platform live on Orpi.com and all 1,250 agency sites, then extended the same property catalog as a native, queryable source inside ChatGPT, Claude, and Google AI Overview, letting buyers and renters describe what they want in plain language instead of using filter-based search.
How to try it: Visit orpi.com or any Orpi agency site and use the AI search assistant.
Sources: Kleio, Usine Digitale, Inman

Spanish insurtech Tuio becomes the first insurer to sell inside ChatGPT, OpenAI says.
Tuio is a Spanish digital-native home insurance provider founded in 2021.
Tuio and AI-distribution startup WaniWani built the first insurer-created app approved by OpenAI to run natively inside ChatGPT — customers converse naturally to get a personalized, regulated home-insurance quote without leaving the chat, with in-chat purchase planned next.
How to try it: Search for and add the Tuio app within ChatGPT.
Sources: INESE, Fintech Global, Reinsurance News

TIM becomes the first Brazilian telecom with a native app inside ChatGPT.
TIM Brasil is a major Brazilian mobile carrier, part of the Telecom Italia group.
TIM launched an app inside ChatGPT that answers natural-language questions about plans, services and partnerships and links out to purchase, no fixed menus, joining a global cohort (Spotify, Canva, Booking.com, Expedia, Zillow) building on OpenAI's Apps SDK.
Sources: Exame, TIM, TI Inside

Mexico's largest department store, Liverpool, wires its catalog into ChatGPT, Gemini and Copilot.
Liverpool is Mexico's largest department store chain.
Liverpool selected commercetools' Agentic Commerce offering to expose accurate product info, availability and pricing to AI shopping assistants, and to power conversational agents that resolve order-tracking, returns and credit-card questions around the clock.
Sources: PR Newswire, Fashion Network Mexico, commercetools

A Japanese prefecture launches a 24/7 generative-AI concierge, retiring its old FAQ-menu bot.
Tochigi Prefecture is a Japanese prefectural government.
Tochigi's official website AI concierge replaces a fixed-scenario/FAQ-matching chatbot, answering free-form questions conversationally across all prefecture services, 24/7, with a patented "ask-back" function that clarifies ambiguous questions before answering.
How to try it: Visit Tochigi Prefecture's official website and use the AI concierge widget (とちぎAIコンシェルジュ).
Sources: PR TIMES, EnterpriseZine

Japanese cosmetics brand Atenia launches an AI avatar beauty counselor — a cosmetics-industry first, its vendor says.
Atenia is a Japanese cosmetics D2C brand.
Atenia's AI Beauty Advisor uses a dialogue-based AI avatar built on ZEALS' "Omakase AI Avatar" platform to reproduce a beauty counselor's knowledge and personality for consultation-style interaction.
How to try it: Use Atenia's online AI Beauty Advisor on their e-commerce site.
Sources: Predge

Platform tier

Brazilian payments giant Cielo launches an AI shopping assistant with in-chat checkout.
Cielo is one of Brazil's largest card-payment acquirers and processors.
Cielo's new assistant lets shoppers search products, get recommendations, compare specs, and complete payment inside a chat window embedded in merchant sites, using Google's open AP2 (Agent Payments Protocol) standard — Cielo claims up to 60% faster checkout and roughly 30% less cart abandonment.
Sources: Mobile Time, Exame, Cielo

Paywhere lets banks and credit unions serve customers directly inside ChatGPT and Claude.
Paywhere is a fintech infrastructure vendor connecting bank accounts to AI assistants.
Paywhere announced commercial availability of a "financial-grade connective layer" that brings a bank's brand, accounts and controls into AI assistants customers already use, so retail and small-business banking customers can check balances, move money and manage accounts conversationally inside ChatGPT, Claude or Copilot.
Sources: Finextra, FF News, PR Newswire

Reception & review findings

Launched 1 day ago (September 2026): Gorgias's own case study on Arc'teryx (company-reported) quotes the brand's CX team calling setup "super easy" with minimal config. Functionally the CX Advisor looks genuinely agentic within scope — auto-translating sizing guidance across regions and linking to fit reviews in conversational form — but there's no evidence yet of order-modification actions, no independent reviews, and no escalation-handling or app-store/social sentiment (too new). Sentiment: positive (company-reported only).
Sources: Gorgias

Launched 1 day ago (September 2026): Decagon's own case study on Hunter Douglas (company-reported) claims over $1M in revenue from conversations fully resolved by AI with zero human escalation, and 85% higher average order value on AI-assisted orders — implying genuinely agentic, purchase-assisting capability rather than pure FAQ. Localized AI personas (Archie/Roman in the UK, Buddy in Australia) adjust tone/accent by market. No independent review or verified escalation mechanics beyond the vendor's claim. Sentiment: positive (company-reported only).
Sources: Decagon

Launched 1 day ago (September 2026): Gorgias's customer story on Nude Project (company-reported) cites ticket-volume/deflection figures that appear to predate this record's own detection, suggesting a longer-running Gorgias relationship rather than a brand-new rollout. Gorgias's platform generally supports genuinely agentic actions (editing orders, processing refunds inside the ticket), but the story itself doesn't confirm those actions are live for this Spanish-market deployment specifically. Sentiment: positive (company-reported only, evidence somewhat dated).
Sources: Gorgias

Launched 1 day ago (September 2026): Decagon's case study on Whop (company-reported) claims 65-70% ticket deflection with persona-tailored responses that pull account/usage metadata — reportedly "many customers don't realize they're talking to AI." No detail is given on what agentic actions it can actually execute, nor its escalation path, and no independent coverage was found. Sentiment: positive (company-reported only).
Sources: Decagon

Launched 2 days ago (September 2026): genuine independent hospitality trade-press coverage exists for Accor's ALL Concierge — Hotel Dive, Business Travel News Europe, Hospitality & Catering News, and Travel PR News all confirm the rollout across 5,800 hotels following a 1M-conversation pilot, describing it as an "end-to-end conversational travel companion" spanning discovery, booking, and in-stay concierge. Coverage is descriptive rather than evaluative — no hands-on review or escalation-handling detail found. General Accor customer-service sentiment on Trustpilot skews very negative, but that reflects Accor's broader support reputation, not this tool specifically. Sentiment: mixed.
Sources: Hotel Dive, Business Travel News Europe, Ru Sustainable Tourism News

Launched 2 days ago (September 2026): Ada's case study on Deezer (company-reported) claims about 30% automated resolution on day one, scaling toward 85% of contacts handled by AI, with humans explicitly kept in the loop for "sensitive inquiries." The underlying Ada platform has generally positive but non-unanimous G2/Capterra reviews across its broader customer base. No Deezer-specific independent review or complaint found. Sentiment: positive/mixed.
Sources: Ada, G2

Launched 2 days ago (September 2026): Ada's case study on Life360's Vida assistant (company-reported) says it let Life360 avoid roughly 400 hires while shifting from a 90%+ email-based reactive model to a more accessible AI-first one; self-reported inquiry volume rose about 10x post-launch while containment reportedly stayed high. No detail on whether Vida can take account actions versus just answer questions, and no independent or user-level coverage found. Sentiment: positive (company-reported only).
Sources: Ada

Launched 2 days ago (September 2026): Ada's case study on Neo Financial (company-reported) claims 5x faster response times and a 50% cut in agent workload. But Trustpilot reviews of Neo Financial describe real frustration — phone calls redirected back into the same AI-driven app flow, and difficulty reaching a human agent for complex issues — a genuine escalation-handling complaint that cuts against the vendor's efficiency claims. Sentiment: mixed.
Sources: Ada, Trustpilot

Launched 2 days ago (September 2026): independent Brazilian press (Fast Company Brasil, Olhar Digital, O Povo) covered TSE's ChatVote launch, descriptively rather than evaluatively. Its capability is explicitly FAQ-scripted rather than agentic, but its escalation/trust design is unusually well disclosed: the tool identifies itself as AI and explicitly defers to official TSE channels on any conflicting answer. No criticism or controversy coverage found. Sentiment: positive/neutral.
Sources: Fast Company Brasil, Olhar Digital

Launched 4 days ago (September 2026): independent German tech press (PC-WELT, ad-hoc-news) and Italy's Il Sole 24 Ore covered Vorwerk's Cookidoo Assistant, descriptively rather than as hands-on reviews. Notably, it's confirmed built on Google Gemini — a rare concrete disclosure in this batch. Capability reads as genuinely conversational (finds recipes, compares options, plans meals, understands spoken natural language) though scoped to meal-planning rather than transactional actions. Sentiment: positive/neutral.
Sources: PC-WELT, Il Sole 24 Ore

Launched 10 days ago (mid-September 2026): coverage of Bank of Baroda's bob World 2.0 confirms the launch and lists AI/voice features, but these are launch announcements, not hands-on reviews. The bank's high general complaint volume reflects a pre-existing app and a well-documented 2023 "illegal onboarding" scandal, not this new AI feature specifically, so it shouldn't be conflated. No independent evaluation of the new AI capability itself was found.
Sources: APAC News Network, Retail Intel

Launched 44 days ago (mid-August 2026): despite extra runway, VoiceBit AI's ChatGPT restaurant-ordering integration still has no independent hands-on review or user complaint on record — coverage remains PR-wire pieces, including a fresh promotional piece just yesterday. By the vendor's own description it's genuinely agentic (completes a transaction inside the chat), built on OpenAI's Apps SDK, but no escalation-handling detail specific to VoiceBit was found. Worth rechecking again given the continued promotional interest. Sentiment: positive (company-reported/PR only).
Sources: AI Journal, The American Reporter

Platform Watch

• Gupshup had a rough stretch financially: Fidelity marked down its stake, implying roughly an 80% valuation drop from prior "unicorn" status, and its India unit (about 60% of revenue) posted a 5% revenue decline and 52% net profit decline for fiscal 2025, with roughly 300 jobs cut. Product messaging stays upbeat regardless — it launched an autonomous "Superagent" in April. Sentiment: negative. Sources: Business Standard.
• Twilio had the opposite week: shares up roughly 19% on bullish AI sentiment, Wall Street price targets raised into the $200–250 range after its SIGNAL 2026 conference, and Gartner again named it a CPaaS Leader for 2026. Sentiment: positive. Sources: Twilio, StocksToTrade.
• Infobip reported 169.3% year-over-year global RCS traffic growth (the US leg alone up nearly 13x), 7,300 unique brands onboarded in 2026, and a top-25 spot on Fortune's Europe's Most Innovative Companies list. Sentiment: positive. Sources: Yahoo Finance.
• Sinch published its own "AI Production Paradox" research this month, finding that 74% of enterprises have rolled back or shut down a live AI customer-communications agent after deployment over governance failures — even as 88% expect to have agents live by year-end. Its own Q2 results were mixed (Americas/API growth vs. EMEA softness). Sentiment: mixed. Sources: Sinch, UC Today.
• Kore.ai held steady: 500+ large enterprises served, 20B+ interactions/year, a 4.6/5 G2 rating, and another year as a Gartner Magic Quadrant Leader for Conversational AI. Sentiment: positive. Sources: G2.

This Week

15 new launches this week (5 enterprise, 10 platform-tier) across the US (5), Global/Multi-market products (6), and one each in Germany, Canada, Brazil and Spain. Reception-highlight of the week: Liberty Mutual's move to sell auto-insurance quotes conversationally inside ChatGPT drew a pointedly skeptical trade-press piece questioning who's liable if the AI under-insures a buyer on a platform the insurer doesn't control. Sentiment: mixed. Source: Live Insurance News.

• Agoda's new AI trip-booking assistant got a trust-focused independent piece asking whether travelers are actually comfortable letting AI complete a purchase on their behalf. Sentiment: mixed. Source: TTR Weekly.
• HubSpot's Customer Agent drew praise for its native CRM integration and smart escalation, but independent reviews flagged recurring accuracy complaints and AI features gated behind a paid tier. Sentiment: mixed. Source: eesel AI.
• Fresha's ChatGPT/Claude booking feature got one genuine independent hands-on test — a UK beauty blogger's real search came back empty on the first try but surfaced real, bookable slots once broadened. Sentiment: positive, on thin evidence. Source: Mama Bella.
• Revolut's AIR assistant expanded to Spain; an independent UX review of its earlier UK rollout found it completes some tasks end-to-end but "behaves more like a guide than an execution engine" for others, like currency exchange. Sentiment: mixed. Source: Craft Innovations.

Full record: bombycine.github.io/claude-routines/index.html

Consumer AI Agent Tracker tracks how real businesses are putting AI chat/agents in front of customers. New here? Read the About page — or browse the full record.