No brand-new consumer AI agent launches cleared the bar today — but three agents that launched earlier this month just got their first reception data, and two AI providers in the platform tracker had a real week.

New reception findings

Chewy's beta bot Cai resolves a third of support issues

Leading US online pet products retailer (NASDAQ: CHWY)

Launched 9 days ago (September 2026): Chewy CEO Sumit Singh told CX Dive that the beta AI assistant Cai — still reaching under 15% of app traffic — is already resolving roughly a third of common issue categories (orders, returns, account/membership questions) without human involvement, with a seamless handoff to a live care team member when a customer needs or prefers one. Chewy says it's holding Cai to a high bar for tone and accuracy given its early stage. Sentiment: mixed (promising resolution rate, but still a small-scale beta with no independent, non-company-sourced testing yet).

Sources: customerexperiencedive.com

Instacart's Clementine builds bigger carts — with caveats

Leading North American online grocery delivery marketplace

Launched 9 days ago (September 2026): early company-reported performance data shows orders placed through Clementine carry more items on average than a typical Instacart basket. But Instacart's own consumer-facing materials caveat that Clementine can return incomplete, incorrect, or outdated information — specifically flagged around allergen, nutrition, and food-safety details — and tell customers to treat every generated cart as a draft needing review rather than a finished answer. Sentiment: mixed (a real engagement signal, tempered by the company's own accuracy caveats).

Sources: techcrunch.com, pymnts.com

Shipt's Ask Shipt gets cautious optimism, not a rave

Target-owned same-day delivery marketplace operating across the US

Launched 9 days ago (September 2026): independent tech-press commentary is cautiously optimistic rather than a clean rave — reviewers note AI shopping assistants have a track record of looking good in press releases and then hallucinating substitutions or missing obvious items in real use, but flag that Ask Shipt's tie-in to Target's large product catalog and Shipt's existing same-day-delivery strength give it a real structural edge if it holds up in practice. Sentiment: mixed (conditional optimism, not yet independently stress-tested).

Sources: intomobile.com

Platform Watch

• OpenAI: a new simultaneous ~90-minute outage hit ChatGPT, Claude, and Grok together on Sept 3, drawing fresh scrutiny of shared cloud-infrastructure risk across AI providers, not just OpenAI's own reliability; separately, OpenAI ended its $1/year pilot pricing for US government agencies on Sept 10, moving federal users to metered pricing at a 50% discount. (source, source)
• Perplexity: legal exposure keeps widening — 11 lawsuits across 3 countries are now tracked against the company, up from 9 previously, and a separate class action filed in April alleges Perplexity routes users' chat content to Meta and Google, a privacy claim distinct from the publishers' copyright suits already on file. (source, source)
• Salesforce Agentforce: on Sept 11 Salesforce introduced seven named Agentforce agents (Casey, Paige, Carter, Hunter, Marshall, Piper, Fin) for specific business functions, and at Dreamforce on Sept 15 expanded its Google Cloud partnership to connect Salesforce's architecture with Gemini Enterprise — product momentum that doesn't yet answer the partner-revenue skepticism already on record. (source, source)

This Week

4 new launches this week: 2 enterprise (Bank of Baroda's conversational-AI-core "bob World 2.0" relaunch in India; Revolut's AIR money assistant expanding to Spain) and 2 platform-tier (WHOOP and Topstep, both surfaced via an Intercom/Fin vendor case-study sweep, in the US). 10 entries picked up new reception findings this week. Highlights:

• Kroger's AI Shopping Assistant is drawing privacy pushback, not usability complaints. EPIC's Sara Geoghegan questioned whether the chatbot's product recommendations serve the shopper or Kroger's own commercial interests, as part of EPIC's broader criticism of Kroger's data practices. Sentiment: mixed. Source
• Sephora's ChatGPT app nails discovery, but checkout still isn't native. A launch-day analyst review called the conversational, personalized product discovery genuinely strong — but in-app purchase remains a "future update," so it's recommendation-first, not a complete shopping flow yet. Sentiment: mixed. Source
• Salesforce's new "Casey" agent claims real scale fast. Salesforce reports its named customer-service agent has handled close to five million conversations across seven languages since its Sept 11 launch, autonomously resolving 68% without a human for a reported $100M in annualized savings — company-reported, no independent user assessment yet. Sentiment: positive. Source
• India's PVR INOX cinema chain built a genuinely agentic WhatsApp bot, per Haptik. "Movie Jockey" books tickets, orders snacks, and gives real-time recommendations by language, location, and showtime across six Indian languages, plus accessibility info like wheelchair-friendly screenings. Vendor-sourced, not independently reviewed. Sentiment: positive. Source
• Chewy's Cai and Instacart's Clementine both posted mixed early reception — see above for the full picture. Source

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.