Duvo, an enterprise artificial intelligence company founded by Tomáš Čupr, the serial entrepreneur behind Czech online grocer Rohlik, has raised $32 million in a Series A round led by Balderton Capital, as businesses search for ways to turn AI experiments into results.
Existing investors Index Ventures, Northzone and Credo Ventures also participated. The round comes about 10 months after Duvo raised a $15 million seed round and brings its total funding to $48 million, according to Tech Funding News. The company did not disclose its valuation, revenue or annual recurring revenue.
Duvo will use the money to enter the United States, starting with retailers and consumer goods producers.
A Founder With an Operator's Perspective
Čupr is one of Central Europe's best-known technology founders. He started Rohlik in 2014; the online grocer became a unicorn in 2021 after a €100 million Series C at a €1 billion valuation and raised a further $170 million in 2024. He previously founded Slevomat and food delivery service Dáme jídlo, which was acquired by Delivery Hero in 2014. He has kept his role at Rohlik while building Duvo.
His co-founders, Marek Paris and Martin Pecha, are former Rohlik colleagues.
That background shapes how Duvo approaches AI. Čupr says he once measured AI by how many jobs it could replace, a yardstick he now considers wrong. Instead, he argues that AI should create value from work that people simply never had time to do.
He gives an example from Rohlik. Planners there used to check about eight at-risk products per market each day. With AI agents, every signal is now checked around the clock. The goal, in Čupr's framing, is to raise each employee's output rather than cut headcount.
The Pilot Problem
Duvo is targeting one of the most persistent frustrations in enterprise AI: projects that perform well in pilots but never make it into everyday operations.
Gartner has predicted that more than 40 per cent of agentic AI projects will be cancelled by the end of 2027, mainly because of rising costs. Many companies have found that building an AI agent for a single process is relatively easy, but connecting it reliably to legacy systems, ensuring it follows approval rules and scaling it across multiple processes is far harder.
Rob Moffat, a partner at Balderton, said that getting AI agents to work reliably across the enterprise "remains incredibly difficult". He credited Duvo's ability to map how work actually happens and where value is lost, tying that capability to Čupr's operating experience.
How Duvo Works
Duvo describes its product as an "Enterprise Execution OS". Rather than treating each automation as a separate project, new processes run on shared connections, approval rules and result tracking, which the company says makes it faster to add new workflows over time.
A tool called Duvo Clarity maps workflows through AI-led interviews, screen recordings and document analysis, helping companies understand how work is really done before automating it. AI agents then carry out approved steps across enterprise resource planning systems, supplier portals, spreadsheets and email inboxes, using application programming interfaces where available and falling back to user interfaces where they are not. Decisions and exceptions are routed to the relevant human owners.
Duvo says initial processes can go live within weeks without heavy engineering resources, a claim aimed squarely at companies that have been frustrated by long and expensive implementation projects.
