Groupon's 8-K Just Certified 400 AI-Native Operators. The Window Is 90 Days.
Groupon's Project Foundry cut ~400 people who shipped production AI agents on a public marketplace. Here's how to source the bench before Q3 2026 ends.
Every JD written in 2026 asks for "AI-native" engineers, product managers, and operators. Almost no candidate can prove the claim. On May 21, Groupon's board gave roughly 400 people the receipt: an 8-K, a named program (Project Foundry), and three shipped production systems on a marketplace serving 16.2M active customers. The bench hits the market between now and the end of Q3 2026. That is your window.
What the filing actually says
The 8-K approved cutting up to 400 positions globally, roughly 23-24% of a ~1,734-person workforce reported in the December 2025 10-K. Pretax charges land at $7-13M (mostly cash severance), with $20-25M annualized run-rate savings and $10-12M gross savings in 2026. Up to half gets reinvested in "marketing, AI infrastructure, and talent density." FY26 Adjusted EBITDA guidance was raised from $70-75M to $75-80M on the same news. The stock rose about 5%.
COO Jiri Ponrt resigned the same day, effective July 10, 2026. The filing states his departure is unrelated to any disagreement. He receives no severance. That last detail matters, and we'll come back to it.
The cuts span engineering, sales, customer service, and HR. Groupon already trimmed ~5% of headcount in Q1 2026 before the board-approved Foundry round, so what you're seeing now is the second, larger wave. Most reductions complete by end of Q3 2026, which means the exits are already in motion and the LinkedIn "Open to Work" surge will lag by four to eight weeks.
Why this cohort is different from every other 2026 layoff
Groupon has cut before. There were 500 layoffs in January 2023. What makes the Foundry cohort valuable is not that they were let go by Groupon. It is that they spent the six months before being let go inside a publicly documented, AI-native production transformation on a consumer marketplace with real traffic (8.1M units in Q1 2026 alone).
Read the Q1 earnings call from May 8, 2026. CEO Dusan Senkypl described product teams "starting from AI-built demos rather than written specifications" and, in some workflows, shipping consumer-facing functionality "without traditional engineering involvement." He framed the reorganized teams as "speedboats" shipping in "days or weeks." Every leader was expected to use AI agents by the end of Q2 2026. Senkypl himself built an "AI chief of staff" on Claude Code and took Groupon leadership to visit ElevenLabs and other frontier SF companies before writing the thesis.
Three named production systems came out of this work:
- Groupon IQ, the AI deal-creation platform, live in production.
- AI-generated review summaries, live across the marketplace.
- AI voice agents for outbound SMB merchant outreach, in pilot with a goal of majority of new merchant meetings AI-booked by end of 2026.
Anyone who touched those systems has a resume claim that a bootcamp graduate, a Copilot power user, or a "prompt engineer" contractor cannot match. They shipped AI into a public consumer product under an SEC-named program. That is the credential.
The 8-K is the resume
Coach candidates you place to lead with the filing. "I shipped on Groupon IQ during Project Foundry (Groupon 8-K, May 21, 2026)" is a claim any hiring manager can verify in thirty seconds. Compare that to the median AI-native applicant, whose portfolio is a Next.js clone of ChatGPT and a Medium post about RAG.
If you're a founder writing an AI-native JD right now, the honest version of your ask is: "someone who has already lost a job to the transformation you're building." That's a small population. The Foundry cohort is the largest publicly identified batch of them shipped into the market this quarter.
This is also where sourcing-by-keyword breaks. Boolean strings like ("AI agent" OR "LLM") AND (Groupon) will catch maybe a fifth of the bench. The rest have job titles like "Senior Product Manager, Marketplace" or "Staff Engineer, Merchant Platform" and never wrote "AI-native" into their headline because they were too busy doing the work. This is exactly the problem Refolk solves: describe the person in plain English (someone who shipped an AI voice agent or LLM-generated summaries at Groupon in the last 18 months) and get a ranked shortlist across GitHub, LinkedIn, and the open web, not a keyword-matched pile.
The geography almost everyone will get wrong
The headline in Hoodline read "Groupon's Chicago HQ Rocked." Chicago is the corporate address. It is not where the engineering density lives.
Pale Fire Capital, Groupon's largest shareholder, is Czech. Senkypl is Czech. CTO Ales Drabek joined in May 2025 from Lampenwelt and Conrad Electronic, both Central European operators. The engineering center of gravity has drifted eastward for years, and Foundry accelerated it. The public locations page lists Prague, Brno-adjacent CEE, Warsaw, Katowice, Berlin, Valencia, Bengaluru, and Chennai as active engineering hubs.
In a single index query of currently-tagged Groupon engineers and PMs (~99 profiles), the top five cities were Bengaluru, Seattle, Prague, Chennai, and Warsaw. Chicago did not crack the top five.
If you are running a US-only sourcing motion out of the Chicago HQ narrative, you will miss the actual builders. If you can hire in Prague, Warsaw, or Bengaluru (or you have an EOR that can), you have less competition and a stronger cohort. Reach out to the Prague and Warsaw pods first. They are furthest from the American recruiter feeding frenzy and closest to the CTO who architected the reorg.
Source the product managers before the engineers
Every recruiter reading a layoff announcement reaches for the engineers first. In this cohort, that instinct is wrong.
Senkypl said the quiet part loud on the earnings call: product teams are "starting from AI-built demos rather than written specifications" and shipping "without traditional engineering involvement." Translate that. The PMs and analysts who learned to prototype in Claude Code, wire up an ElevenLabs voice agent, and ship a working consumer surface without an engineering ticket are the rarest profile on the market right now. They are also the least likely to describe themselves as "technical" on LinkedIn, because they came up as PMs.
Backend engineers displaced by AI transformations are a growing supply. AI-native PMs who shipped to 16M customers without waiting for a sprint are not. Source them first. Pay them like engineers.
Every JD asks for AI-native. Groupon's 8-K is the only place 400 people can prove it. </pull>
Wait. Correcting that syntax below where it belongs.
The stock went up. Read the room accordingly.
When a layoff moves a stock down, the emotional posture of the cohort is grief plus solidarity. When a layoff moves the stock up 5%, the posture is different. The market publicly rewarded the company for removing them. That is a specific kind of raw.
Outreach that leads with sympathy ("sorry to hear about the layoff") will underperform outreach that leads with respect for the work. "I saw your name attached to Groupon IQ. What's the piece of that you'd rebuild from scratch?" gets a response. "Sorry about Foundry" gets ignored or resented.
This is a place where hiring engineers displaced by AI requires a different playbook than hiring engineers displaced by cost-cutting. The Foundry cohort was cut precisely because their work succeeded. They automated their own function. That is a story they will tell for the rest of their careers, and it is a story worth hearing before you pitch a role.
The adjacent bench nobody is counting
Ponrt's exit is the tell. A COO leaving voluntarily on the same day as the RIF, with no severance, is not a routine transition. That signals cultural upheaval one layer above the WARN-eligible cuts.
Expect quiet departures over the next two quarters from senior operators who were not officially impacted but who saw the shape of things. Their LinkedIn will not say "impacted by Project Foundry." It will say a new role at some AI startup in October or November 2026. The tenure end date is the signal, not the bio.
Widen the search. Anyone with "Groupon" tenure ending between June and October 2026, especially at director and above, is worth a conversation. The formal Foundry list is a floor, not a ceiling.
This is another place general-purpose Boolean fails. You cannot Boolean "left Groupon quietly after Foundry because the culture flipped." You can, however, describe that person in plain English to a tool that reads tenure timelines and cross-references company events. That is what Refolk was built for. Sourcing AI-native talent from the adjacent layer of a public restructuring is exactly the query shape where natural-language search beats a filter stack.
A 60-day operational plan
Here is the compressed version if you have a Q3 headcount to close.
Week 1-2. Build the target list. Prague, Warsaw, Bengaluru, Chennai, Berlin, Chicago, Seattle, in that order of density. Filter for tenure that touches 2025-2026, not just current impacted status. Prioritize PMs and product-adjacent engineers who shipped Groupon IQ, review summaries, or the voice agent pilot.
Week 3-4. First-touch outreach. Lead with the specific product. Reference Senkypl's speedboats framing if you want a signal that you actually read the earnings call. Skip the sympathy.
Week 5-8. Second-touch and interview loops. The Foundry cohort will be talking to five to fifteen companies each. Your differentiator is not comp, it is whether your role lets them keep building the thing Foundry started. If your engineering culture still routes AI prototypes through a two-week spec review, do not pitch this bench. They will bounce in 90 days.
Week 9-12. Backfill from the adjacent layer. Senior operators who resigned voluntarily post-Foundry. Contractors from the ElevenLabs and Anthropic tool stacks. Ex-Groupon PMs from the January 2023 cut who have since built AI-native muscle elsewhere and might rejoin the diaspora.
By early Q4 2026, the cohort will have dispersed into stealth startups, series-A infra companies, and the AI teams at Amazon, Instacart, and DoorDash. The 8-K certification will still be on their resume. The response rate to cold outreach will not.
FAQ
How do I verify a candidate actually shipped on Project Foundry versus just working at Groupon during the period?
Ask them to name the system, the traffic surface, and the collaborator on the other side of the interface. Groupon IQ, AI review summaries, and the AI voice agent pilot are the three named production systems. A real Foundry contributor can tell you which one they touched, who the PM was, what shipped in Q1 versus Q2 2026, and what the rollback plan looked like. A pretender will speak in generalities about "AI initiatives."
Is the Chicago HQ still worth sourcing from, or should I focus entirely on Europe and India?
Chicago is worth sourcing but not worth leading with. The commercial, marketing, and some senior product roles sit there. The engineering density is in Prague, Warsaw, Bengaluru, and Chennai, in that order. If your role is US-based and requires colocation, Chicago and Seattle are your realistic pool. If you can hire in the EU or India, start there. You will find stronger candidates with less recruiter competition.
What compensation should I expect the Foundry cohort to command?
The 8-K certification is a real premium but it does not reset the market. Prague and Warsaw base ranges hold. Bengaluru and Chennai hold. What shifts is the leverage on scope: this cohort will negotiate hard for autonomy, direct access to the CEO or CTO, and the right to keep using their preferred AI toolchain (Claude Code, ElevenLabs, and whatever internal stack they built on top). Give ground on scope and you will save on cash.
Where does Refolk fit in a sourcing motion this specific?
For a named cohort like Foundry, the bottleneck is not finding candidates who list "Groupon" on LinkedIn. It is finding the PMs and engineers who shipped specific products, in specific cities, during a specific window, without the right keywords in their headline. Refolk lets you describe that person in plain English, including the product they touched and the timeline, and returns a ranked shortlist across GitHub, LinkedIn, and the open web. For a 60-90 day window against a public 8-K cohort, that is the difference between hiring the bench and reading about where they landed.