The AI Transformation Brief—November 4, 2026
Company AI spending rose more than fivefold in a year, faster than buyers can check what they get or who answers for it.
The AI Transformation Brief
// Today’s Signal
AI spending is outrunning the buyer's ability to check it. Company AI spending reached $176.8 billion, 5.4 times last year, and three suppliers took 69% of it. Menlo Ventures Snowflake is building AI workers that run unattended and call a person only for sign-off. Unite.AI Anthropic showed a prototype that proves which AI produced an answer. Anthropic Hg and Anthropic are funding a firm to put AI engineers inside European companies. The AIC The buyer who can check the work will hold the power.
// Top Stories
Menlo Ventures estimates that company spending on using AI reached $176.8 billion by September, up from $33 billion in 2025. Anthropic, OpenAI and Google took $122.7 billion, or 69%, up from 50% a year earlier. New budgets funded 55% of the added spending, and 22% came from existing software budgets. The report draws on a survey of nearly 400 US decision-makers responsible for more than $4 billion in AI spending. Menlo Ventures
Most of this money is new, so few companies have the cost controls they use for older software. Concentration is the bigger risk: when three suppliers hold 69% of spending, they set prices, usage rules and contract terms. Anthropic alone took 55% of what companies spend on direct access to AI, up from 40%. Menlo Ventures Give every AI budget a named owner, track results against it each quarter, and keep a second supplier tested.
Snowflake announced always-on CoCo and CoWork Bots on Tuesday. A user sets a goal, such as preparing briefings for every customer meeting, and the bot plans and does the work using that user's access rights. It surfaces only when a critical decision or final approval is needed. The bots are coming soon for testing with select customers. CoWork is used by more than 5,800 accounts and CoCo by more than 9,100. Unite.AI
Software that works while no one watches changes who answers for the result. The person whose access the bot borrows is accountable for work they never saw. Snowflake is adding a safety check that stops AI from acting beyond its access and flags staff who use it to skip approvals. Snowflake Decide which decisions need a person before switching on any AI that runs unattended, and review a sample of its finished work every week.
The employee whose access the bot uses answers for its work, and the person who approves the final step checks it. If no one reviews the steps in between, mistakes reach customers under that employee's name.
Anthropic released FlashProve, an open-source prototype that attaches a proof to each AI reply showing it came from the exact AI the provider promised. Anyone can check the proof on an ordinary computer without seeing the provider's secrets. Anthropic says users cannot currently confirm an answer did not come from a cheaper AI "secretly substituted in by a routing service for cost reasons." Anthropic There are no plans to put it into service at Anthropic or any other lab. TechCrunch
The prototype is expensive: proving an answer takes roughly 70 to 95 times the computing time of producing it. Anthropic The lesson for buyers arrives now. Few AI contracts say which AI must answer each request or how a buyer would know. Add that clause at renewal: name the AI your supplier must use, require notice before any switch, and ask how the supplier will show it kept that promise.
Hg, a London investor in software businesses, agreed to invest with Anthropic and angel investors in a new company that will help European businesses use AI and build AI-based products. It will launch before year-end, with engineers building on Claude. Terms were not disclosed. The AIC Hg says the firm will work across Europe's many languages and regulatory regimes and aims for "measurable operational impact." Hg Anthropic backed a similar services company with Blackstone, Hellman & Friedman and Goldman Sachs in May. FourWeekMBA
AI suppliers are moving into the work of putting their AI to use. That puts them in competition with the consulting firms companies hire for neutral advice. An adviser funded by one AI supplier will tend to recommend that supplier. When you hire help, ask who funds the adviser, insist on a comparison of at least two AI suppliers, and own every rule and workflow they build for you.
// Shelly Palmer Pulse
Palmer looks at OpenAI's promise to cut its data centers' electricity use when local grids are strained, including a 25-year deal with Georgia Power. He asks how each commitment will be checked and enforced, and who pays if the campus arrives late or uses less power than expected. Shelly Palmer Every AI promise in this edition deserves the same test.
// What It Means For Your Business
AI spending is growing faster than your ability to check it.
Name one executive this quarter to own every AI budget, every unattended AI worker and every supplier promise, reporting results to you monthly.
With 69% of spending at three labs, and those labs funding the advisers who help companies choose, buyers lose bargaining power unless they keep a tested alternative and independent advice.
Before any AI works unattended, list the decisions that need a person, set who approves each one, and review a weekly sample of finished work against those rules.
Ask management which AI works without a person watching, whose access it uses, and whether each AI contract names which AI must answer and how the company can verify it.
// The Take
The shift this edition is that AI spending, and AI working alone, are growing faster than anyone's ability to check either. The assumption it breaks is that a supplier's word settles which AI did the work and whether it followed the rules. The decision it forces is to build checks into contracts and processes before the next renewal.
If your AI supplier quietly swapped in a cheaper AI next month, how would you know?
Make AI accountable. Measure the results. Redesign the business.
Read On
The AI Transformation Brief—November 10, 2026
AI is landing inside cloud bills, manager pay plans and national law, and each one changes who answers for the results.
Read the brief →The AI Transformation Brief—November 9, 2026
Your AI supplier's security gaps, unfinished work and power deals are now your business risks, and this week showed each one.
Read the brief →The AI Transformation Brief—November 8, 2026
Businesses are using more AI for the same money, while exposed hacking tools and voter pushback show who pays AI's hidden costs.
Read the brief →