HubSpot's board approved a restructuring on October 1 that eliminates roles across approximately 7% of the company's workforce(opens in a new tab), with $65 million to $75 million in charges, most of them in the fourth quarter. The CEO memo filed with it puts the number at nearly 660 people and says HubSpot will reduce management layers and move decisions closer to the people doing the work. It ties the change to a strategy shift around AI, then says plainly that the cut is not driven by AI-related efficiencies(opens in a new tab).
That sentence is the interesting part. CEO Yamini Rangan describes a shift from building software that helps customers grow to delivering outcomes for them with AI, and in the same letter declines to claim AI savings. Both statements can be true. But a reader of the filing gets a headcount number and a charge, and no measure of what the AI strategy is expected to return. When a company does not publish its own AI scorecard, outsiders build one from headcount and share price.
Most large companies are in the same position. In its State of Markets update, a16z says nearly 30% of S&P 500 companies report some quantifiable impact from AI, but only about 2% report a metric they track over time(opens in a new tab). A one-time claim of impact is a press release. A metric tracked quarter to quarter is evidence.
Airbnb shows what the second kind looks like. Its CTO, Ahmad Al-Dahle, said roughly half of support tickets are now resolved purely by AI(opens in a new tab), in line with the nearly 45% in its second quarter results, which was up from 40% in the first quarter. That is one number, reported quarter after quarter, so a reader can see it move. It also comes with a stated limit: Airbnb deliberately keeps some tickets, such as safety issues, with humans.
For an enterprise running AI pilots, the lesson from this week is about sequencing. If you restructure first and name the metric later, the board, the analysts and the press will choose the metric for you, and they will choose headcount. Pick one or two operating measures per AI program now, record the baseline before any org change, and report them on the same schedule as the financials. Keep restructuring charges in a separate line from any AI return claim, so neither one gets used to explain away the other.
Action items
HubSpot reorganized around AI this week and declined to claim AI savings. Only about 2% of the S&P 500 disclose an AI metric they track over time.
For the CFO
Keep restructuring charges and AI return claims in separate lines, so one is never used to explain the other.
For the CAIO
Name one or two operating metrics per AI program and record the baseline before any org change.
For the board
Ask for the same AI numbers every quarter, measured the same way. A single claim of impact is not a trend.
Researched and drafted by an automated workflow, then reviewed and edited by a human editor before publication. Every source is linked. See how we use AI here.
Also worth knowing
- U.S. Securities and Exchange CommissionHubSpot Form 8-K: workforce restructuring plan(opens in a new tab)
Roughly 7% of roles and $65 million to $75 million in charges, mostly in Q4. The filing is the template for how a restructuring tied to an AI strategy gets disclosed, and what it leaves out.
- U.S. Securities and Exchange CommissionHubSpot CEO memo: reducing management layers, not driven by AI efficiencies(opens in a new tab)
Nearly 660 roles go and management layers shrink, yet the memo says the cut is not driven by AI-related efficiencies. Expect your own board to ask which story your AI reorganization tells.
- Andreessen HorowitzState of Markets(opens in a new tab)
Nearly 30% of S&P 500 companies report some quantifiable AI impact, but only about 2% report a metric tracked over time. A repeatable number is now a way to stand out, not a formality.
- Latent.SpaceAirbnb's CTO on resolving half of support tickets with AI(opens in a new tab)
About half of Airbnb's support tickets are resolved purely by AI, in line with nearly 45% in Q2 (up from 40% in Q1), after synthetic testing before launch and with safety cases kept with humans. A model for one metric, reported consistently.
- Constellation ResearchMeta, Sierra pitch Personal Agent Protocol to govern how personal AI agents deal with business(opens in a new tab)
Sierra and Meta launched a draft standard with Shopify, Stripe and Walmart for how customers' AI agents deal with businesses. The spec is early, but customer service and commerce leaders should track it.