Today's thread is where AI money is going, and what it is not yet buying.
Before your 2027 plan is locked, make sure every AI dollar in it has a dated outcome attached.
Developments
Go-to-Market
OpenAI Says ChatGPT Now Handles About One-Fifth as Many Search-Style Queries as Google (November 3, 2026)
At an OpenAI dinner in New York, ChatGPT consumer product lead Adam Fry said an internal analysis puts ChatGPT's volume of search-like queries at roughly one-fifth of Google's, covering lookups of people, places, finance and health. Weekly users grew from 1 billion to 1.2 billion in about 60 days, and voice features reach 250 million weekly users. Fry also confirmed plans to merge ChatGPT's Chat and Work modes into one interface.
So what
If that figure is even close, a meaningful share of the discovery your buyers do now happens in an answer engine that cites a few sources and rarely sends a click. Ask your marketing lead for one report this month: how often ChatGPT names you, and which pages it cites, across your 25 most important buying questions. Fix the cited pages first, since those are what the model reads.
Hype check
The one-fifth figure is OpenAI's own internal analysis, and it has not disclosed how it counts queries or follow-up prompts.
Confidence: Informed perspective — Fry's remarks are on the record via Big Technology; the underlying metric is unaudited.
Go deeper Big Technology · The Deep View
Customer Behavior
Menlo Ventures Counts $176.8 Billion in Enterprise AI Deployment Spend, With Most of the Growth Funded by New Budgets (November 3, 2026)
Menlo Ventures' annual survey of nearly 400 U.S. enterprise decision-makers estimates $176.8 billion in AI deployment spending as of September, with OpenAI, Anthropic and Google capturing 69%. New budgets funded 55% of incremental AI spend, while only 22% was redirected from existing software budgets. Excluding the labs' own apps, startups took 61% of application spending. Anthropic led enterprise API spend at 55%, ahead of OpenAI at 31% and Google at 9%.
So what
For software sellers, AI money is mostly arriving as new budget lines rather than cuts to your renewal, which favors products that do work a buyer could not buy before. Frame your AI offer against a task budget, such as hours of analyst or support work, rather than as a seat upgrade. The report measures spending, not returns, so expect CFOs to ask for returns at next year's renewals and start building that evidence now.
Hype check
Menlo Ventures is an Anthropic investor; treat the vendor-share figures as one firm's survey, not audited market data.
Confidence: Established fact — published survey results, which measure spending rather than value delivered.
Go deeper Menlo Ventures · The Deep View
Product & Technology
CrowdStrike Adds an AI Kill Switch That Cuts Rogue Agents Off From Their Models (November 3, 2026)
CrowdStrike launched AI Kill Switch in its Falcon Guardian platform: an administrator can direct connected AI gateways to block an agent's access to the models that drive its decisions, and every use is logged. Extending containment to endpoints, cloud, identities and SaaS apps is still in development. CEO George Kurtz says the control sits outside the AI so an agent cannot argue with it or wait it out. The same day, CNBC reported that Apple's annual filing warns AI agents may act in unintended ways, including completing transactions without the right permissions.
So what
Agent governance is becoming a line item in security stacks, and enterprise buyers will start asking vendors how their agents can be stopped. If you deploy agents, route their model calls through one gateway you control so a single switch can halt them. If you sell agents, document your stop mechanism and audit log before a security questionnaire asks for it.
Confidence: Established fact — CrowdStrike's own announcement; broader enforcement is labeled coming soon.
Go deeper CrowdStrike · The Rundown AI
Competition
Google Will Switch Off Search, Not Just Gemini, for K-12 Students Until Schools Opt In (November 3, 2026)
Google said Gemini, Gemini Notebook, Gemini in Classroom and Google Search will all be off by default for K-12 student accounts until an administrator turns them on, with the controls arriving in the coming months. Education VP Chris Phillips told The New York Times that "Search and A.I. are deeply integrated," so students at schools that ban AI would lose Search. Google says each product can be enabled separately but has not shown whether Search can run without its AI features. Former FTC chair Lina Khan called the move coercive.
So what
This is a bundling decision presented as a safety control, and it shows how a dominant vendor can make AI non-optional by tying it to the product customers cannot drop. Expect the same pattern in your own stack as AI features arrive inside tools you already depend on. When renewing core platforms, ask in writing which AI features can be disabled without losing the underlying product, and if you sell into regulated buyers, offer that switch yourself.
Confidence: Established fact — Google's announcement and on-record comments; the final admin settings are not yet published.
Go deeper Google · The Neuron
Perspectives
Business
Stripe Benchmarks Show 29 of 48 SaaS Peer Groups Shrank in October as Churn Eased (November 3, 2026)
Mack Grenfell analyzed Stripe benchmark data covering about 181,000 subscription businesses and found 29 of the 48 peer groups above $100,000 ARR shrank in October, up from four in September. Customer churn fell slightly, from 7.45% to 7.24%, which leads him to suspect fewer new customers rather than more departures. Only companies above $5 million ARR still grew, at 0.58%. He cautions this is one month of data, and Menlo's survey finds AI only modestly slowing traditional SaaS growth.
So what
If the pattern holds, the squeeze shows up first in new logos as more products chase the same buyers. Split your pipeline reporting so new-customer win rates and sales cycles are visible on their own, separate from net revenue retention. If win rates are slipping, check whether prospects are comparing you with AI-built alternatives your team has not yet seen in deals.
Confidence: Informed perspective — one analyst's reading of a single month of benchmark data.
Go deeper Mack Grenfell · TLDR Founders
Competition
Euclid's Nic Poulos: Vertical AI Apps Win Where the Outcome Data Only Exists Inside Real Work (November 3, 2026)
Opening The Verticalist Summit, Euclid Ventures' Nic Poulos argued that vertical AI companies keep most of the application layer even as labs launch products aimed at their customers, citing OpenAI's legal offering arriving about a week after Harvey raised at $15.5 billion. He sorts the outcome data that trains models into three groups: synthetic, which labs can simulate; purchasable, which labs can buy; and job byproducts such as loan defaults or claim approvals, which exist only when someone does the work. Labs win the first two; vertical apps can win the third.
So what
This gives boards a usable test for the question of whether a lab will absorb your product. List every outcome your product touches and classify each one. If most are synthetic or purchasable, speed is your main defense. If most are job byproducts, make sure your product is present when the outcome happens and captures it as structured data from day one.
Confidence: Informed perspective — a venture investor's framework, and vertical AI investors have a stake in this conclusion.
Go deeper Euclid Ventures · TLDR Founders
Resources
Tools
Claude Team and Enterprise Admins Can Now Provision App Connectors Centrally, So Staff Get Them at First Login (November 2026)
Anthropic's enterprise-managed auth lets administrators authorize a connector once for the organization, choose which roles get it and at what permission level, and require managed-only connections so personal accounts stay out of work tools. Identity-provider-managed auth, currently through Okta, is generally available; a Claude-issued option that needs no identity provider is in beta. Supported connectors include Asana, Atlassian, Figma, Linear, Monday.com, Notion and Slack.
So what
The usual blocker to rolling out AI connectors is sign-in sprawl and the security review, not the tool itself. If you run Claude Team or Enterprise, pilot one connector with one team's role, confirm that deprovisioning removes access, then expand. Whatever AI platform you use, make managed-only connections a requirement before staff link work data.
Confidence: Established fact — Anthropic's help documentation.
Go deeper Claude Help Center · The Deep View
Operations & Finance
A CFO's Template for Presenting AI Spend in the 2027 Annual Plan (November 3, 2026)
OnlyCFO warns that AI can make an annual plan look leaner before any savings arrive. Its fix is to show AI spend beside the headcount plan and date when avoided hires or extra output reach the numbers, split AI costs between gross margin and operating expense, and break out revenue by pricing model, since consumption-based AI revenue can carry very different margins. It also recommends a separate line for the AI cost of free trials and proofs of concept, which can grow quickly.
So what
Boards will discount any plan where AI works as a plug that improves the forecast without changing how the business runs. Before your plan goes to the board, ask your CFO for one page that ties each AI dollar to a dated outcome and a named owner. Show department AI usage only next to what it saved or produced.
Confidence: Informed perspective — practitioner guidance from a finance newsletter.
Go deeper OnlyCFO · TLDR Founders
