Three small but genuinely conversational AI deployments surfaced today via a vendor case-study sweep, and two follow-up reviews landed — one with a real regulatory wrinkle.

New launches

Platform tier

Music streamer Deezer automates most support contacts with Ada AI agent.
Deezer is the French-founded global music streaming service with 9M+ paying subscribers.
Deezer deployed Ada's AI agent to handle customer support. The agent touches about 85% of contacts and reached roughly 30% fully automated resolution on day one, with humans kept in the loop for sensitive inquiries.
Sources: ada.cx

Canadian fintech Neo Financial shifts support from voice-heavy to AI chat with Ada.
Neo Financial is a Canadian digital banking/fintech challenger.
Neo added Ada's AI chat agent to its app and web portal with instant in-chat authentication for real-time account resolution, shifting its support mix from roughly 80% voice to a 50/50 chat/voice split. Neo reports 5x faster response times, 50% fewer inquiries reaching a human, and 600,000+ chat interactions per year.
How to try it: Open chat inside the Neo Financial app or web portal — instant in-chat authentication resolves most account questions without switching to a phone call.
Sources: ada.cx

Family safety app Life360 automates most support chats with AI agent 'Vida'.
Life360 is a family safety and location-sharing app used by tens of millions of families in the US.
Life360 deployed Ada's AI agent 'Vida' to automate customer support. Vida resolves roughly 70% of support inquiries — described as the workload of about 400 full-time support agents — including handling a support-volume surge roughly 10x higher than expected without losing containment.
Sources: ada.cx

Reception & review findings

Launched 1 day ago (per detection; the underlying rollout actually dates to late May/June 2026): Liberty Mutual's ChatGPT-based auto-insurance quoting tool is getting a real look from trade press. Most coverage just syndicates Liberty Mutual's own press release (7-state rollout inside ChatGPT via the Apps SDK, plans for 40+ states by year-end), but two independent pieces push back: Intelligent Insurer frames Liberty Mutual as "joining the scramble" of carriers racing into ChatGPT quoting, and Live Insurance News raises a pointed, unresolved question — who's liable if the AI gives bad guidance or under-insures a buyer on a platform (ChatGPT) the insurer doesn't fully control? On capability, it's confirmed genuinely functional rather than scripted (it runs on Liberty Mutual's own rating engine for a real personalized quote), but it isn't fully agentic end-to-end — it still hands off to libertymutual.com to actually bind the policy, and no human-escalation path is clearly disclosed. Sentiment: mixed.
Sources: Intelligent Insurer, Live Insurance News, Live Insurance News, Fintech Global, Fintech Global, eMarketer, PropertyCasualty360, ReinsuranceNe.ws, IIReporter, International Business Magazine, InsureMojo

Launched 2 days ago (September 2026): Agoda's new AI travel assistant, which plans and books entire trips conversationally, is drawing a more skeptical trust-focused take alongside the usual launch-release syndication. TTR Weekly's piece, "Would you trust AI to book and buy a trip?", questions whether travelers are actually comfortable letting an AI complete a purchase on their behalf rather than just recommend one. Per Agoda's own (not yet independently verified) claims, the assistant is genuinely agentic — connecting to live inventory of 6M+ accommodations plus flights and activities, retaining session context, and using Google's "Grounding with Google Maps" for complex multi-factor requests like a family trip with teens. It's built on OpenAI plus Google Cloud/Gemini grounding. No hands-on reviewer test or human-escalation path has surfaced yet, consistent with this still being a limited public beta. Sentiment: mixed/cautious.
Sources: TTR Weekly, RU Tourism News, PR Newswire, ANI News, Out of Town Blog

Platform Watch

• Visa pushed real developer infrastructure this week: general availability of a Visa MCP server (lets AI agents call Visa's Intelligent Commerce APIs directly, cutting integration from weeks to hours) and a no-code Visa Acceptance Agent Toolkit for invoicing and Pay by Link. Incremental, but it's Visa building out the plumbing behind the agentic-commerce push rather than just talking about it. Sources: Digital Transactions, Techstrong.ai.
• Manhattan Associates launched "Editions," a new three-tier packaging model for its Manhattan Active suite meant to let smaller, growing retailers adopt the platform (including its AI customer-service layer, Active Maven) without a full re-implementation later. A go-to-market change, not new adoption data on the chatbot itself. Sources: Manhattan Associates IR, StockTitan.

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.