"Agents aren't a new tool layer. They're the team a merchant could never afford to hire." That's the premise behind a new piece by Masha Sharma, VP of Merchant Experience AI at Groupon, written ahead of her panel at the AI Agent Conference in NYC this week. The reframe: stop thinking about AI as a set of features. Start thinking about it as the headcount your business has been short on the entire time. The pricing analyst you couldn't afford. The front desk you couldn't keep staffed. The merchandiser who never had time. Three roles every merchant has always wanted to hire, and now can. What makes the piece worth reading isn't the agentic enthusiasm. It's the operator's lens: how the merchant's job actually changes (you become a manager of policies, not a doer of tasks), what to hire next, and why platforms that charge on execution, not exposure, are about to define agentic commerce. Read the full piece on our Leadership Insights page. Link in comments. #TeamGroupon #AgenticCommerce #AIAgentConference
About us
We believe the best things in life happen offline. That belief shapes everything we build and every partnership we form. For consumers, we make it easy to say yes to their city. For merchants, we deliver real customers through real doors. Not impressions. Not clicks. People who show up. We move fast, we own outcomes, and we're hiring people who want to do the same. Explore our open roles and join a tech company that actually believes in scrolling less, living more. #TeamGroupon
- Website
-
https://www.groupon.com
External link for Groupon
- Industry
- Software Development
- Company size
- 1,001-5,000 employees
- Headquarters
- Chicago, IL
- Type
- Public Company
- Founded
- 2008
- Specialties
- technology, local commerce, social media, marketing, community, customers, customer worship, and e-commerce
Locations
Employees at Groupon
Updates
-
“Precision beats volume. Every time.” That’s the premise behind a new piece by Ernesto Martín Rodés, VP of Revenue and CRM Marketing at Groupon, on what it took to rebuild CRM from the ground up. His team cut managed-channel message volume in half, overnight, no soft landing. They thend use the savings to fund a new Customer Data Platform, rebuild lifecycle programs across 13 markets..and they keep cutting. The result: email volume down 66% year-on-year. The channel growing, not flat. Unsubscribe rates down. Customer tenure up. What makes the piece worth reading isn’t the volume cut. It’s the operating logic behind it: that short-term volume is borrowed from long-term retention, that you don’t argue for new infrastructure on a slide deck, and that respecting customer attention is not a soft metric: it’s a growth strategy. Concrete, grounded, no hype. Read the full piece on our Leadership Insights page. Link in comments. #GrouponAI #TeamGroupon #TechAtGroupon
-
-
This May, we're celebrating the people behind every Groupon milestone. To everyone marking a Groupon anniversary this month, thank you. Your energy, your craft, and your commitment are what make this company what it is. A special thought for Stefan Lück, Justin Blanton, and Nadine Leidecker, all marking 16 years with Groupon this month, whose long-standing impact continues to shape our teams every day. Here's to you, and to everything still to come. #TeamGroupon #ScrollLessLiveMore #GrouponAnniversary
-
-
Our CTO Ales Drabek just shared what he took back from Google Cloud NEXT, and the through-line is one of our operating principles: Speed Over Comfort. It's the way we build, and it's the way we hire. The shift from AI pilots to AI agents in production is not a slogan at Groupon. It's the daily work for our engineering teams: shipping fast, simplifying ruthlessly, and standardizing what is repeatable so the next agent ships faster than the last. If that sounds like the bar you want to be held to, our engineering org is hiring. Roles open across AI, platform, and product engineering. Read Ales's full take below. 👇 #Groupon #Engineering #AI #Hiring #LifeAtGroupon https://lnkd.in/gmSw9Mma
From "AI as sidekick" to AI as infrastructure — my Google Cloud Next ‘26 takeaway Spent a few days at #GoogleCloudNext. Most of the news was incremental. One announcement was not: the Universal Commerce Protocol (UCP) — an open standard that makes catalogs and checkout flows directly usable by AI agents. This matters more than the agent demos. At Groupon we already operate agentic-first internally. Over 60% of our code is written with AI. Agents handle first workflows that used to take analysts hours: deal approval reviews, supply health narratives, weekly business reports and marketing in different channels. We are not "experimenting" with AI — it is how we work. The gap is on the outside. Today a customer wanting a spa day in New York, a haircut in Chicago or a weekend with full of experiences in Madrid comes to a search bar. Tomorrow they will ask an agent. If our local-experience inventory is not agent-readable, none of our internal speed matters. That is why UCP is the announcement to watch (till now focus on this setup is more on goods, but I believe soon we will also see expansion in other categories). Not the flashiest one. The one that decides whether marketplaces become infrastructure for AI ecosystems — or get disintermediated by them. We have a real opportunity here. Groupon has unique inventory (300k+ local merchants, experiences you cannot ship in a box) and a team that is already building agent-native. The next move is making that inventory speak to the agent. The second interesting learning was about expansion from website agents mostly only replacing customer service on all steps in customer journey on retail websites. In my opinion we are already done with the first level of agents in post purchase area (FAQs, solving customers issues) and we see now expansion from agents on product detail page, in pre-purchase or by platforms on integration between merchants and customers. And this is exactly the Groupon direction. We already have customer service AI Agents in 14 countries with tons of daily interactions and now we are starting with first pilots on our Deal Page or in communication with merchants. More to come in next weeks … What was your take on UCP? Did anyone else see it as the most important thing in Vegas?
-
-
"And we failed." That's our CTO Ales Drabek talking about one of our first attempts to put AI agents on a complex sales workflow at Groupon. The agents hallucinated. Not because the technology wasn't ready. Because the context wasn't. The knowledge needed to run that workflow was spread across people's heads, Google Sheets, scattered docs. We tried to automate before we'd done the homework. In the new episode of AI Monstera by ACTUM Digital, Aleš shares the lesson most companies skip: Don't start with the agent. Start with the data, the playbook, the knowledge structure. Then automate, step by step. We've built hundreds of agents across engineering, customer service, sales and compliance since. But the failures taught us more than the wins. Worth 30 minutes if you're past the hype phase. #TeamGroupon #GrouponAI
-
"Speed comes from decomposition, not better prompts." That's the premise behind a new piece by Viktor Bezdek, VP Engineering at Groupon, on the orchestrator-subagent pattern. Take a complex business question. Break it into independent research threads. Run them in parallel through specialized AI subagents. Synthesize the findings into a weighted hypothesis with confidence levels. A day of digging across tools, answered in 20 minutes. What makes the piece worth reading isn't the speed claim. It's the engineering discipline behind it: independence between threads, graceful degradation when subagents fail, synthesis logic that doesn't just average confidence scores. Concrete, grounded, no hype. Read the full piece on our Leadership Insights page. Link in comments. #GrouponAI #TeamGroupon ##TechAtGroupon
-
-
Most companies are bolting AI onto how they already work. We are doing the opposite. At #CzechCrunchFuture last week, our CEO Dušan Šenkypl shared how Groupon is using AI not to optimise the old operating model, but to rebuild it: faster decisions, less internal friction, more focus on the work that actually matters. The real difference today is not which AI tools a company uses. It is how willing it is to change the way it works because of them. Full coverage from CzechCrunch in the comments. #GrouponAI #TeamGroupon #ScrollLessLiveMore
-
-
A few days of working sessions across our Madrid and Valencia offices. Country leaders, local teams, and shared priorities in the same room. Performance reviews, new projects taking shape, and a clear focus on AI in execution. Strong teams, clear direction, and the kind of alignment that only happens face to face. #TeamGroupon #GrouponAI #ScrollLessLiveMore
-
-
Not a discipline problem. A friction problem. Our CTO Ales Drabek on why we stopped forcing teams into JIRA and built an AI Assistant that lives where they already work. Stop building better gates. Start building better paths. Read Aleš's full post below ⤵️ #TeamGroupon #GrouponAI
The Problem Wasn’t Discipline. It Was Friction. Before joining Groupon, I worked with several German companies where "follow the process" was the cultural baseline. Last May when I joined Groupon I quickly recognised that this company has a different way of working. For example every bigger business area has their own chat in which there is a communication with Engineering about specific Applications. So we have Salesforce Chat, Merchant Center Chat, Deal Creation Chat … In this chat people are sharing bugs, asking for technical help or changes in the system. I spent months trying to get teams to use JIRA Service Desk for this type of issue. The result? Everyone kept posting in Google Chat anyway. I used to think this was a discipline problem. I was wrong. It was a friction problem. People live in Google Chat. That is where the context is and where the conversation happens. Asking them to context-switch to a ticketing tool to report a problem they are already describing in a message is unnecessary overhead. They weren't being lazy; the process was just poorly designed for how humans actually work. If you can't move the people to the tool, move the tool to the people. We stopped fighting the behavior and built an AI Assistant in Chats (starting with Salesforce Chat) that lives directly in Google Chat. Now, the moment someone posts a Salesforce issue in our support spaces (like MC or DCT), the bot starts investigating automatically. There is no human triage and no "cold start" for escalations. How it works under the hood: Real-Time Investigation: The bot uses the #Google ADK and #Claude Sonnet 4.6 to chain complex API calls. Deep Context: It runs SOQL queries, checks metadata state via the Tooling API, and pulls real-time instance health via the Status API. Dual-Layer Response: It posts a plain-English answer for the user and a structured technical summary—root cause, impact, and suggested fix—for the engineer. Closing the Loop: It even handles on-call tagging via JSM and can create Salesforce cases on demand. The Stack We built this to be scale-out safe and secure, running on Google Cloud Run with Firestore for persistent session state and MCP Protocol to bridge our LLM to Salesforce tools. The Real Lesson The value of AI isn't in forcing new, "modern" workflows on your team. It’s in making the workflows they already use dramatically smarter. Stop building better "gates" and start building better "paths."
-
-
The engineer is dead. The builder is what comes next. As our CEO, Dušan Šenkypl puts it: the bottleneck is no longer code - it’s clarity. AI isn’t just making people faster. It’s removing the boundary between thinking and building. Most companies are scaling AI usage. More tools. More prompts. More output. But they’re not building systems that learn. So knowledge stays fragmented. Mistakes repeat.Nothing compounds. One of our teams flipped this. 8 engineers. 25 repos. 4 domains. Led by Minas Arustamyan, they built a shared AI system where every correction, pattern, and decision is captured and reused. “Knowledge lives in the system, not in people’s heads.” That’s the shift. From speed → to learning. Full article in the first comment. #TeamGroupon #GrouponAI
-