There is a specific genre of layoff press release that the market reads as bullish. The headline number is bad. The annualised savings number is good. The EBITDA guide moves up by less than the savings number. The company calls the cost out an “AI-native pivot.” The stock pops 5%.
Groupon ran the play on May 21 and the market priced it on May 27.
The Form 8-K Groupon filed with the SEC on May 21, 2026, announced the company would cut up to 400 positions — employees and contractors — by the end of Q3, take a pre-tax restructuring charge of $7 to $13 million (mostly cash severance), and unlock $20 to $25 million in annualised payroll savings. The company told investors it expects $10 to $12 million of gross savings to land in 2026 and intends to reinvest up to half of that in marketing, AI infrastructure, and “talent density.” Net savings in 2026: about $5 million. Adjusted EBITDA guide for the full year raised from $70–$75 million to $75–$80 million. Stock popped 5% premarket when the press cycle picked it up on May 27, closing at $21.48.
Groupon’s April 2026 10-Q reported roughly 1,734 employees including part-time, seasonal, and temporary workers. If the 400 cuts hit only the employee count rather than the contractor base, that is approximately 23% of total headcount. Fast Company and PYMNTS both rounded it to “nearly a quarter.”
Project Foundry, named at last
The 8-K introduces the project name the company has been workshopping internally: Project Foundry. The framing — Groupon is rebuilding itself as “AI-native” so it can operate with the speed required to succeed in an AI-native world — is now the standard 2026 layoff dialect. Cloudflare’s Matthew Prince used a slightly grander Drucker-quoting variant on May 20 (“builders, sellers, and measurers”). Wix’s Avishai Abrahami used a slightly more defensive variant on May 25 (the Base44 acquisition is eating the parent). Groupon’s variant is the most candid of the three: the company is openly piloting AI voice agents for merchant outreach, with a stated goal that AI sets the majority of new merchant meetings by year-end, and is explicitly redirecting the freed payroll into AI infrastructure spend.
The 8-K also says Phase 1 is the entrée. The filing explicitly notes the company is evaluating additional material cost-reduction and automation actions related to Project Foundry, subject to board approval, with any such actions expected to be completed by the end of 2027. That is the language of a multi-year program that has another two waves in the queue and is telling the market it has them in the queue. The bull case prices that in. The bear case wonders what the workforce floor looks like in 2027.
The COO leaves the same day
The other line in the 8-K that the press cycle has under-reported: on May 21, 2026, Chief Operating Officer Jiri Ponrt notified the company of his decision to resign, effective July 10, 2026. The filing notes it is not due to a disagreement. The same day the company announces a Project Foundry that will be COO-territory operational work in execution — automation actions, function-by-function AI agent rollouts, “talent density” hiring — the operator who would own that execution resigns. That is the kind of coincidence that prospectuses are written to defuse and that investor relations teams pre-brief sell-side analysts on. Both happened. Neither announcement explains the other. Make of it what you will.
CEO Dusan Senkypl, who has run the company as permanent CEO since 2024 after the Pale Fire Capital co-founder was installed by the Czech investor group that took an activist position on the board, owns the strategy. Q1 2026 billings were $383 million and falling. The growth story is in the AI-native rebuild, not the legacy daily-deal flow. Project Foundry is the bet that the company can shrink the bottom line fast enough to outrun the top line.
What the market actually bought
The 5% pop is buying three things at once: a real $20–25M annualised cost out, an EBITDA guide raised by about $5 million (i.e., roughly what the 2026 net savings number forecasts), and the option value of two more Project Foundry waves through end of 2027. The thing the market is not yet pricing is whether AI voice agents actually close more merchant meetings than the human team they are replacing. That is a Q3 disclosure, not a Q2 narrative. Box CEO Aaron Levie’s “AI psychosis” diagnosis from yesterday is the warning label: CEOs see AI at the demo layer and miss the last-mile delivery layer. The 400 people leaving Groupon by end of Q3 are the delivery layer.
What to watch
- Project Foundry Wave 2 announcement window. The 8-K reserves it for board approval and a 2027 completion date. The interesting tell is whether Senkypl pre-announces Wave 2 alongside Q2 earnings or holds it for Q3 once the first batch of AI voice-agent metrics are in.
- AI voice-agent merchant-meeting metric. The company has staked a year-end goal that AI sets the majority of new merchant meetings. That number, if disclosed, is the Project Foundry KPI.
- COO replacement. The 8-K does not name a successor. A general counsel or CFO acting-COO bridge is the cheapest path. A named outside hire is what the strategy actually needs.
- Q1 billings trajectory. Q1 billings were $383M and falling. If Q2 billings reaccelerate, the AI-native bull case writes itself. If they keep falling while the cost out lands, the next 8-K is the one investors will read sideways.
The press release ran the play. The market bought the play. The question is whether the operator who would run the play is the one who just resigned.