Agent Beat for November 2, 2026

This is Agent Beat, the daily briefing on how AI finds brands and how agents choose what to buy. It's Monday, November 2, 2026.

TechCrunch reports TikTok is building toward an app people use for most of their digital activities, and it is adding restaurant guides and payments.

A Metric Sherpa study produced with Sprinklr found 92% of organizations with customer-facing AI in regular use said it met or exceeded expectations, but only 27% said they could prove it. Of those who said AI made work faster, 67% also said it cut costs, and only 32% of that group had tracked costs before launch. The study found 99% could name an intended outcome, 26% defined success metrics and 10% recorded a baseline. MarTech.org reported that 85% of those surveyed reported improvement in at least one metric category they had not tracked before launch, and that only 17% said their organization names an owner for AI business outcomes, the least common of the eight practices the study measured.

About 16% of shoppers begin product discovery in ChatGPT, while only 5% buy on the spot after an AI recommendation. After an AI recommendation, consumers check reviews, price, delivery and return terms.

Amazon started offering Alexa+, its generative AI assistant, in Japan at 3,000 yen a month, with no extra charge for Prime members. Early access is free, and the assistant works with Tabelog, Uber Eats, Demae-can and Ticket Pia.

CNN International is building in-house infrastructure to sell advertising through AI agents, targeting first revenue in the first quarter of 2027. It is working with outside vendors including demand-side platforms. Faisal Karmali, vice president of digital operations, said the company had not begun trading but wants to see how buy-side agents read its inventory, what they ask for and how they get there. DIGIDAY Japan reported that, under the original plan, CNN International was to lock down protocol specifications by the end of the second quarter, test in the third quarter how one or two outlets are recognized by large language models, and in the fourth quarter examine buyer behavior and budget allocation, and that the first use cases will center on performance campaigns, with the scope widening later.

At MarkeZine Day 2026 Autumn, Ryo Taoka of suswork introduced a gap between what people recall and what AI recommends, first made public in joint research with X-Ops. Taoka, who has run more than 100 category entry point studies, also listed 11 pitfalls that can arise before execution.

Tohoku Electric Power is working on generative engine optimization because the utility, well known in its home region, does not show up in answers from AI search tools. Its case is that local recognition does not carry over to AI assistants.

Search optimization is not dead, but the ground has shifted from traditional engines to AI search, and the shift happened quickly. The advice is to optimize for AI search, including ChatGPT and Google.

Content that can be summarized is a losing proposition, while content built around a distinct point of view has value. The advice is to keep a human expert involved, and large language models are agreeable and without authority.

TechCrunch reports that Big Tech is pouring billions into AI infrastructure while debt mounts, cash flow runs negative and investors grow skeptical about the boom's sustainability.

Google's Gary Illyes said on the Odys podcast, in an interview with Aneta Samkoff, that bot impersonation has become more of a problem because of AI, with rogue bots pretending to be Googlebot.

AMD agreed to acquire World Labs, the spatial intelligence company founded by Fei-Fei Li, in an all-stock deal worth $8.2 billion. It is AMD's second-largest acquisition after Xilinx, which it bought for about $50 billion in 2022. AI科技评论, or Leiphone, reported that World Labs raised a $1 billion C round on February 18, 2026, led by Autodesk, with AMD, Nvidia, Emerson Collective and Fidelity among the participants, at a post-money valuation of about $5.4 billion. It reported that World Labs closed a seed round in the same month it was founded, led by a16z and Radical Ventures, at a valuation of about $200 million.

Meituan released LongCat-2.5, a native multimodal model with 1.6 trillion parameters built for long-horizon agent tasks. Meituan says it trained the model entirely on Chinese-made compute, the largest training run completed on that hardware.

Anthropic says Claude carried out 26% of the AI research activity Anthropic measured in August 2026, up from 1% in March, under human supervision.

That's today's briefing. Thank you for listening. For the full written brief, go to agent-beat.com.