EP 281- Siri Going Agentic, Apps Face the Agent Dilemma, and Big Tech Falls From Grace

Greg Sterling and Mike Blumenthal go it alone again this week (Myles Anderson of BrightLocal moves to next week) for a two-act episode: first Apple's quiet build-out of a permission-based agentic Siri and what it means for the publishers and directories that feed AI platforms, then Google's ultimatum to schools and the midterm results that turned data centers and tech billionaires into campaign liabilities.

Agentic AI is a contest over who owns the customer relationship — and with voters now punishing coercive, opaque tech, the platforms that ask permission of both the consumer and the merchant (Apple's slow, gatekept Siri) are better placed than the ones that impersonate users or bundle AI into products people can't refuse.

Takeaways

• Apple Maps ads now come with a human rep. Mike tried to create an Apple Maps ad when the program launched and has now been assigned a "client strategist." Greg's point: no one has built a small business ad business on self-serve alone. Google relies on agencies and a large human support layer, and he told Anthropic's head of SMB the same thing at its small business roadshow. (02:09–04:40)

• Siri AI opens to every app in iOS 27.3. The 27.3 developer beta connects Siri's natural-language layer to the App Intents developers have exposed for years, and adds "app skills" that bundle several actions into one request. Mike reads it as Apple underpromising and overdelivering: none of it was announced in June. He also says it follows the Agent Skills spec Anthropic developed, so any AI on the iPhone could use the same hooks. (04:40–06:38, 06:38–08:37) [1]

• The phone is becoming a full-funnel federated search interface. Mike traces Apple's arc from Maps and Apple Pay to its own web index, Spotlight and now Siri AI, which ties messages, email and calendar to commerce. His example request: find the sneaker my daughter recommended, size nine wide, and check whether Walmart has it in stock. He has already moved his ad hoc TV-watching questions from Google to Siri. (04:40–05:53, 10:20–10:40) [2]

• Permission on both sides separates Apple from Muse. Square now syncs merchant hours, menus, ordering and booking links straight into Apple Maps and Siri. That is the merchant opting in, and the user opts in too — the opposite of Meta's Muse clicking through Amazon as if it were the customer, which is why Amazon blocked it. (06:38–08:45, 15:28–15:56) [3]

• Apple's slow pace is a gatekeeper's luxury that fits real adoption. Greg: only a platform everyone has to genuflect to could move this leisurely. Mike: it also matches how people adopt. His wife is on an iPhone 13 and uses DuckDuckGo; it will be years before she has a Siri AI-capable phone and longer before she searches with it, so Apple is planning on roughly a five-year window. Meanwhile on-device models and Google's TPUs push inference costs down, and usage limits feed a paid tier. (09:06–12:35)

• Every consumer-facing site faces the same choice: feed the aggregator or protect the visit. Google, ChatGPT and Claude all fold partner data into their own brand (Reserve with Google, home-services booking in AI Mode, apps inside ChatGPT). For a B2B player like Square it doesn't matter. For Expedia, TripAdvisor or OpenTable it means becoming a data vendor, and if there's no visit, there's no ad impression. Yelp sits in the middle: it can push its quote system to Siri to start the lead and still pull users back into its own environment. (12:42–16:04)

• The Yellow Pages already ran this experiment. When print directories were a $15 billion industry, Google pitched publishers on reselling AdWords and handing over their listings for extra distribution. The publishers who said yes ceded their traffic and their brands faded. Today the industry is about 90% gone, with Thryv selling its remaining print directories to private equity, and Yelp is the only horizontal local search survivor out of a dozen-plus. (16:06–19:48)

• Weak consumer brands end up as agencies. Brands without consumer pull, such as Thumbtack and Angi, are pushed into selling distribution to contractors, which is what the Yellow Pages publishers became before they disappeared. Thryv has little left but its small business SaaS. Strong brands like Amazon can afford to block agents outright. (19:48–21:40)

• Google wants the checkout; Apple routes you to the app. Google's Universal Commerce Protocol lets the purchase happen on Google, turning every retailer into a commodity data provider. Siri sending Mike to the right product in the Amazon app still lands him on Amazon's ads and Amazon's ranking, which gives retailers a reason to participate. (21:40–23:28)

• Google told schools: take AI or lose Search. Schools that turn off Google's AI tools lose Google Search for students, because Google says the two are integrated. Mike calls it hubris that will come back to haunt Google. Greg calls it a pattern: let us use your reviews or lose indexing entirely (Yelp, TripAdvisor), take no payment for news or be de-indexed (Europe), stay in the index only by staying in AI Overviews. (23:28–26:31)

• Voters turned against data centers and the companies building them. Ballot measures on data centers went 38 of 41 against development, and Greg links Greg Abbott's loss in Texas partly to his support for them. Greg's three self-inflicted wounds: years of job-loss and doom rhetoric with no societal benefit to show for it, opaque deals for water and power, and an all-in embrace of Trump that now invites congressional investigations. (27:59–32:54) [4]

• A constrained buildout favors incumbents. Mike's counterpoint: if data center growth is capped by local opposition and regulation, Google and Amazon, with the lobbyists and the experience to navigate it, are better placed than OpenAI and Anthropic. Markets may already be pricing that in. (32:54–34:21) [5]

Related Reading

• iOS 27.3 Beta Finally Gives Siri AI Third-Party App Support (Beebom) — what the developer beta opens up, including app skills, and when users will see it.
• Agent Skills — the open format for packaging instructions an AI agent can load, which Anthropic introduced.
• Square Integrates Apple Business to Help Local Businesses Stand Out on Apple Maps — Square's announcement of direct listing sync to Apple Maps and Siri.
• Amazon blocks Meta's Muse agent from shopping on users' behalf (SiliconANGLE) — why Amazon locked out the agent the hosts keep returning to.
• Google will let schools disable its AI tools, but not without sacrificing Search (TechCrunch) — the school policy discussed in the second half.
• Midterm elections 2026: AI data centers lost big (Fast Company) — the ballot measure tally and the Texas governor's race.
• Midterm elections could zap the data center buildout, experts say (CBS News) — what the Texas result could mean for data center capacity.
• Apple Maps Ads launch in the US and Canada (TechRepublic) — background on the ad program behind Mike's new rep.
• EP 271 — The Path of Enshittification: Apple Ads, Google's Toll Booth, and Meta's $18B Escape Hatch — the earlier hosts-only episode on Apple Maps ads and Google's leverage.
• EP 273 — Fragmented Discovery, Google's €460M DMA Fine and Apple's Distribution Bet — Apple's distribution advantage, from an earlier episode.

Practitioner Notes

• Apple's local layer is now worth managing directly. With Siri AI drawing on Apple Maps data and Square pushing merchant data into it, Apple Business listings (hours, menus, booking links) belong in the same maintenance routine as GBP for local clients. Square merchants can sync from their dashboard; everyone else needs to keep Apple Business current by hand.
• Inventory feeds are the common thread in agentic local search. Mike's sneaker test, Let Google Call backfilling from Merchant Center last week, and the Siri-to-Walmart example all depend on structured local inventory. Retail clients without a product feed are invisible to every one of these agents.
• Google's school policy is the same playbook publishers face with AI Overviews. Greg's list — Yelp and TripAdvisor reviews, European news, AI Overviews opt-outs — is a ready-made history of Google using index access as leverage. It is useful context for any client weighing whether to block Google's AI features or restrict their content.

Concepts

• Full-funnel federated interface — Mike's term for the iPhone as a single entry point that connects Maps, Pay, Apple's web index, Spotlight and Siri AI to personal data and to every app, covering discovery through purchase.
• Permission-based agency — agentic actions that both the consumer and the merchant have opted into, as with Square's Apple Maps sync, as opposed to an agent posing as the user on a site that hasn't agreed.
• Disintermediation dilemma — the choice every consumer-facing site faces: feed data to an aggregator or agent for distribution, or protect the direct visit and its ad revenue.
• Commodity data provider — what a publisher or retailer becomes once an aggregator owns the customer and the checkout.
• Gatekeeper's luxury of time — Greg's phrase for Apple's ability to move slowly because every developer and data source has to deal with it anyway.
• Coercive bundling — Greg's pattern for Google tying something users need (indexing, Search) to something it wants them to accept (review use, AI Overviews, AI tools in schools).
• Broligarchs — Greg's label for the AI and tech leaders whose rhetoric and political choices turned the industry into a campaign issue.
• Surveillance pricing — setting individual prices from personal data, which Rep. Robert Garcia said Democrats would move to shut down.

Pick your starting point:

00:00 Intro
02:09 Apple Maps ads get a human rep
04:40 iOS 27.3: Siri AI meets App Intents
06:38 Agent skills and the full-funnel iPhone
08:41 Permission vs. impersonation: Apple and Muse
09:06 Apple's gatekeeper pace
12:42 Feed the platform or protect the visit?
16:06 The Yellow Pages precedent
19:48 Weak brands become agencies
21:40 Apple's middle path vs. Google's checkout
23:28 Google's ultimatum to schools
26:32 From "don't be evil" to villains
27:59 Data centers lose at the ballot box
32:54 Constraints favor the incumbents
34:43 Public distrust and privacy exhaustion
38:30 Gilded Age to Progressive Era? + wrap

Full Transcript -->on-line