Newsletter, 20 November 2026
The UK ad regulator now has AI scanning gambling ads
This week: the ASA's AI ad monitoring catches a Casumo promotion, Mark Ritson on why agents shop like people, what Zapier learned watching Claude read its website, and Spotify's three different versions of "the AI product manager".
Our four views on how AI is changing iGaming are on the blog: Acquisition is changing, A new front door, How products get built and businesses run and Regulation. This week's news sits underneath them.
The UK ad regulator now has AI scanning gambling ads
Our view: RegulationSarah MacDonald and Matthew Plampton of law firm Wiggin report that the UK Advertising Standards Authority (ASA) upheld a ruling against a Casumo promotion that was flagged by the ASA's AI-driven Active Ad Monitoring system. The ad, seen in August 2026, offered up to 100 spins through an opt-in "daily mission": deposit £50 and wager £200 to earn a wheel spin, up to five spins a day, between 6 July and 2 August. The ASA accepted the offer was optional, but said each mission unlocked another chance, which directly incentivised repeat play, and it also questioned the "mission" and "quest" wording (Wiggin, 18 November).
Wiggin points out that since 19 January 2026 the Gambling Commission's rules cap wagering requirements on bonus funds at ten times, and that its January changes removed the rule on rewards tied to a pre-determined frequency. Its read is that the ASA judged the promotion by its most extreme possible use. Its closing line: "With AI now scanning ads, the absence of complaints offers little comfort" (same article).
Mark Ritson: agents shop like people, so build the brand
Our view: Acquisition is changingMark Ritson, marketing consultant and founder of the MiniMBA, argues in The Drum (16 November) that AI agents copy human decision-making closely enough that brands do not need a separate strategy for them. His advice is to build a well-known, well-rated brand and the agents follow, with clean product data, reviews and visible prices as the plumbing underneath. He draws on Columbia's mock store that agents shopped thousands of times, MIT Media Lab's 80,000 experiments on 17 agents and an Ahrefs study of 75,000 brands (The Drum).
He also notes how quickly the sources an assistant leans on can move: ChatGPT's citations of Reddit fell from close to 60% of responses in early August to around 10% by mid-September (same article).
Zapier watched Claude read its website, and the page gets summarised first
Our view: A new front doorZapier engineers Ben McAdams, Alex Black, Tim Hey and Andrew Kordampalos built a tool called Agent Arm Runner that runs the same Claude Code task against a live page and a test version of it, so they can see what an agent does with each. They report (6 November) that markdown versions of pages are already part of the agent's request path, and that when the agent fetches a page the content first goes to a separate, smaller model with its own instructions, including a strict 125-character limit on quotes, before the main model sees it (Zapier).
Spotify: "the AI product manager" is three different jobs
Our view: How products get built and businesses runThree Principal Product Managers at Spotify, Henrik Lindström, Divita Vohra and Humberto Corona, wrote on the Spotify Engineering blog (9 November) that "the AI PM" is not one role. It depends on what is being built, and they split it into three layers: research, where product is technical leverage; a recommendation foundation model, where the customers are internal teams; and the listener-facing experience. Core product fundamentals still apply, with AI adding a layer of specialisation on top (Spotify Engineering).
Keshav Parthasarathy makes a related point in his Substack on ML product management (16 November): product and ML teams cannot have a clean hand-off, because each owns both the "what" and the "how" with different weight (Notes on ML product management).
Other things we've been reading
- -The search for the AI-native firm. David Peterson of Angular Ventures argues that AI-powered enterprise software is still "SaaS 2.0", sold per seat, and compares SaaS 2.0, AI roll-ups and AI-native services firms. He suggests three questions for designing a new organisation: where the bottleneck is, where economies of scale lie and what compounds (Angular Ventures, 17 November).
- -Owning the intelligence layer. Justin Borgman, co-founder and CEO of Starburst, argues that agents need access, meaning and proof across many systems, and that centralising all data is the wrong goal. This is a vendor post, and the claims are his (Starburst, 18 November).
- -The death of personalization. Jeriad Zoghby, who led personalization work at Accenture, argues that segments and past behaviour give way to a high-definition view of each consumer built with LLMs, and that data has to go from traffic report to Waze. Part 1 of 3 (Machine Speed, 10 November).

Work with us
If one of these is useful to you, send us a message or visit bettingproduct.ai/products. We start with a fixed-scope AI Discovery and Pilot: one problem, one pilot, one decision.
Ian and Vlad | BettingProduct.ai