Groupon's Project Foundry Just Trained 400 Voice-Agent Hires. Bland Has Until July 10.
Groupon's May 21 AI-native restructuring frees the merchant-outreach pool Bland, Vapi, Retell, and Sierra have been trying to script. Here's the window.
On May 21, 2026, Groupon's board approved Project Foundry: up to 400 positions cut, roughly 24% of headcount, with the explicit goal of replacing human merchant-outreach reps with AI voice agents. Same day, COO Jiri Ponrt resigned effective July 10. If you run sourcing at Sierra, Bland, Retell, or Vapi, the seven weeks between those dates are the entire game.
This is not a generic AI-pivot layoff. Groupon just produced the most domain-trained merchant-outreach automation pool in the market: people who spent years cold-calling SMBs and negotiating deal economics, now reading their own obituary in an 8-K filing. They are the training data the voice-agent vendors have been buying synthetically. With W-2s.
What Project Foundry actually did
The SEC 8-K is unusually direct. Groupon framed Foundry as "a company-wide initiative to transform our operating model by embedding AI agents into the core of every function," intended to let the company "operate with the speed required to succeed in an AI-native world." Pre-tax charges run $7 to $13 million, mostly cash severance, unlocking $20 to $25 million in annualized savings. The company simultaneously raised EBITDA guidance from $70 to $75 million up to $75 to $80 million. The stock popped about 5% on the news.
The math behind the 400 number matters. Groupon's April 2026 proxy listed roughly 1,734 employees. A 400-person reduction, even spread across employees and contractors, lands somewhere between 23% and 24% of the workforce. Most cuts complete by the end of Q3 2026.
Q1 2026 financials explain the panic. Groupon posted a $0.32 loss per share against a $0.19 profit a year earlier. Revenue was flat YoY at $117.2M. Net loss: $12.8M. Foundry is not a strategic flex. It is a board trying to print EBITDA before the next earnings call, and the cheapest variable cost to cut is the merchant-outreach floor.
Why the 8-K, not a press release, is a signal
Consumer brands almost never break a layoff this size inside an 8-K with no companion CEO memo. Groupon did. Affected employees largely learned through financial media before internal channels caught up. That delivery mode breeds the fastest-moving candidates: people who already feel disrespected by their employer's communication, and who started updating profiles the night the filing hit EDGAR.
The pool nobody else has indexed yet
Here is the part recruiters keep missing. The merchant-outreach population at Groupon is not 400 generic "salespeople." It is a specific, mappable cohort.
Our index of professional profiles surfaces approximately 89 current or recent Groupon merchant-facing sales people. Of those, 24 still list Groupon as their employer right now. The dominant titles are Account Executive (17) and Sales Representative (6). Geography clusters hard in the Greater Chicago Area (10), with secondary nodes in Atlanta, Phoenix, LA, NYC, and two German metros (Kassel and Hannover-Braunschweig, the legacy CityDeal footprint).
That is not a vague abstraction. That is a finite, named list you can work in an afternoon. The trap is that LinkedIn keyword search will surface a much larger pool of ex-Groupon people who left in earlier rounds (2020, 2022, 2023), and most of them have moved into adjacent industries and lost the muscle memory. What you want is the still-employed-on-May-21 cohort, the people whose calendars are about to clear.
This is exactly the kind of search where a Boolean string fights you. "Groupon" AND "Account Executive" returns ten years of alumni. "Currently at Groupon" filters too aggressively because Sales Navigator's freshness is stale. The cleaner approach is to describe the person you want in plain English, which is why we built Refolk: ask for current Groupon merchant-sales ICs in Chicago with three-plus years of outbound cold-calling, get a ranked shortlist that respects the May 21 employment cutoff.
Why these people, specifically, are the voice-agent prize
The voice-AI market crossed $22 billion this year. Gartner forecasts conversational AI will cut contact-center labor costs by $80 billion in 2026 alone. Vapi just closed a $50M Series B at a $500M valuation after Amazon Ring evaluated 40-plus voice vendors and chose Vapi to handle inbound. Vapi processes 62 million calls a month. Parloa is at $3B. PolyAI is past $500M. Every one of these companies is hiring conversation designers, prompt engineers, and outbound-call ops leads faster than they can interview.
The product these vendors ship is, fundamentally, a script. A really sophisticated script with branching, latency budgets, and tool calls, but a script. Scripts need authors who know the conversation. Groupon merchant reps have spent years inside the exact workflow these voice agents now imitate: cold-calling an SMB owner who is on the floor of a restaurant, negotiating a 50/50 deal split, handling the "I tried Groupon in 2014 and got burned" objection, surviving 70% rejection rates without burning out by lunch.
The people getting replaced by the voice agent are the only ones who actually know what the voice agent should say.
This is the asymmetry. The voice-agent vendor's engineering team can ship lower latency, better TTS, cheaper inference. What they cannot ship from scratch is merchant-domain conversational fluency. They have to either spend twelve months gathering it from customer data or hire it. Foundry just made hiring it possible.
The vendor-by-vendor fit
Bland AI is the natural home. Their entire pitch is high-volume outbound: up to 1 million concurrent calls, optimized for throughput and conversion rather than conversation quality. That is merchant acquisition dialing, exactly. A Groupon AE who hit quota in Phoenix already lives inside Bland's target use case.
Vapi needs go-to-market depth more than conversation design right now. With ~100 employees and a $500M valuation, the next 20 hires shape the company. The merchant-sales managers, not the ICs, are the prize here.
Retell AI plays in the regulated, multilingual lane (about 600ms latency, HIPAA, 31+ languages). The Kassel and Hannover-Braunschweig cohort matters specifically for Retell's European expansion. German-language outbound sellers with US-managed-team experience are rare.
Sierra is the stretch hire. Bret Taylor and Clay Bavor have built an enterprise/brand-governance product on a proprietary Agent OS, and their voice capability is still maturing after a late-2024 launch. They are the platform most lacking gritty SMB-merchant context, which means a Groupon hire reframes their roadmap rather than slotting into it. That is a harder sell, but a higher-ceiling one.
The July 10 corridor
The clock is the COO's calendar. Jiri Ponrt notified the company on May 21 and exits July 10. His resignation was voluntary, meaning no severance, meaning he was already out the door before Foundry was announced. He is the operational sponsor of merchant outreach. When he leaves, the mid-level managers who report into that org lose their internal advocate and their last reason to stay.
Practically, that means the seven weeks between May 21 and July 10 are when those managers update LinkedIn, take coffees, and decide. The ICs follow the managers. After July 10, the best of them are either in process at Bland or Vapi already, or they have signed with whoever the new merchant-focused COO turns out to be. Analyst notes flag that incoming COO appointment as a bull signal precisely because Groupon is trying to retain the merchant-outreach IP it just torched.
Groupon's own retention pitch will lean on the "talent density" language buried in the 8-K: management plans to reinvest up to half of the $10 to $12 million gross savings into marketing, AI infrastructure, and talent density. Translation, "we are cutting you but we promise to spend on the survivors." That works on senior engineers. It does not work on an AE in Atlanta who just learned about the layoff from CNBC.
How to actually run the search
Three concrete moves.
One: index, do not Boolean. The cohort is small enough (24 currently-employed targets, plus the 65 who have left in the last 18 months and may be open to a second move) that you should treat it as a named list, not a search. Build the list once, work it in waves.
Two: lead with geography, not title. Chicago is the prize because Groupon HQ is at 600 W. Chicago Ave, and the merchant-outreach floor was physically there. SF and NYC-default voice-agent companies will lose this fight unless they pitch remote-first or open a Chicago hiring pilot. A Sierra recruiter pitching relocation to the Presidio will lose to a Bland recruiter offering remote with quarterly Chicago meetups.
Three: separate the IC from the manager track. The AE-titled reps fit Bland and Retell. The Sales Representative-titled reps fit conversation-design and QA roles. The managers (not surfaced in the 24-person count but adjacent to it) are the Vapi and Sierra GTM hires. Mixing them in one outreach sequence wastes both.
The reason a plain-English query beats a saved search here is that "merchant outreach automation talent" is not a LinkedIn skill, it is a description of lived experience. Tools like Refolk that read intent and triangulate across GitHub, LinkedIn, and the open web are what surface the German-language seller in Kassel who happens to have managed a US team. Boolean will never find that person.
What happens if you wait
The voice-agent market is consolidating fast. Vapi winning the Amazon Ring evaluation against 40-plus rivals tells you the enterprise-buying side is making decisions now, which means the GTM build-out is now, which means the hiring window is now. Wix cut 1,000 on May 26. Cisco cut 4,000 on May 14. Foundry is one of three AI-native restructurings in a two-week window, and recruiters at every voice-agent vendor are looking at the same calendars.
The team that moves on Groupon's Chicago cohort before July 10 will hire from a 24-person pool with no competition. The team that moves on July 20 will be the fourth call those candidates have taken that week.
FAQ
How many Groupon people are actually in scope for voice-agent vendors?
The top-of-funnel pool is roughly 89 current and recent Groupon merchant-facing sales people identifiable in professional-network data, with 24 still listing Groupon as their current employer as of the May 21 announcement. The most relevant subset is the 10 Chicago-based Account Executives, plus a smaller cohort in Atlanta, Phoenix, LA, and NYC. The European nodes in Kassel and Hannover-Braunschweig matter specifically for vendors with German-language roadmaps.
Why is Bland a better fit than Sierra for these candidates?
Bland's product is purpose-built for high-volume outbound campaigns, up to one million concurrent calls, optimized for throughput and conversion. That maps directly to merchant-acquisition dialing, which is exactly what Groupon AEs did. Sierra is enterprise/brand-governance focused with a maturing voice layer, so a Groupon hire there is a roadmap-shaping bet rather than a slot-fill. Both can win, but Bland will close faster.
What is the realistic window before this pool dries up?
The COO exits July 10, 2026, and most Project Foundry cuts complete by the end of Q3. The practical recruiting corridor is the seven weeks between the May 21 announcement and July 10, when affected managers are still on the floor and ICs are watching them. By August, the top quartile is in process or signed. By October, the remaining candidates are the ones who waited, which is itself a signal.
Will Groupon's retention pitch hold any of them?
Some, but not the best ones. Groupon's 8-K explicitly commits to reinvesting up to half of the $10 to $12 million in gross savings into marketing, AI infrastructure, and talent density. That pitch works on senior engineers and product leaders the company explicitly wants to keep. It does not work on a merchant-outreach AE who learned about the restructuring from CNBC because the announcement was an 8-K, not an all-hands.