From Rohlik to enterprise AI: Tomáš Čupr’s Duvo raises $32M Series A

Duvo aims to automate workflows across disconnected business systems without extensive engineering teams, as it prepares to expand in the US.

Duvo today announced a $32 million Series A led by Balderton Capital, with participation from existing investors Index Ventures, Northzone, and Credo Ventures.

The round brings Duvo’s total funding to $48 million and will support its expansion, initially targeting retailers and consumer packaged goods (CPG) companies.

Tomáš Čupr is a Czech serial entrepreneur, co-founder and CEO of Duvo, and founder and Group CEO of Rohlik Group, the European online grocery company valued at approximately $2 billion. He founded Rohlik in 2014 and expanded the business beyond the Czech Republic into several European markets.

I spoke with Čupr to learn more.

Duvo is an enterprise operating system that brings a company’s processes together on one platform. Businesses can automate multiple processes using the same software connections, approval rules, and monitoring tools, rather than treating each automation as a separate project.

How Rohlik’s operational challenges led to Duvo

Duvo emerged from the operational frustrations Čupr encountered while building online grocer Rohlik.

Despite recognising the potential of AI to improve productivity, he found that introducing automation into an established business was far more complicated than expected. One of the biggest challenges was understanding how work actually happened across the company. He found that employees often struggled to explain their processes in sufficient detail.

“I’d say, ‘Let’s build these agents. What can we automate?’ But when I asked people to explain a process, including all the exceptions, it became clear that much of that knowledge lived in people’s heads.”

Even when processes could be documented, the software involved often lacked the APIs needed to connect different systems. Employees also relied on a mixture of spreadsheets, internal applications, and email, frequently transferring information manually between them.

“We wanted to create an enterprise super app, an operating system that sits on top of existing systems and brings everything into one place.”

Rather than replacing existing enterprise software, Duvo aims to provide a single interface through which employees can manage tasks, make decisions, approve actions, and escalate issues.

Turning undocumented processes into AI automation

Duvo follows the same approach for each process. Its Duvo Clarity tool uses AI interviews with employees, screen recordings, and existing documents to understand how work is carried out in practice, including exceptions and decisions that may not be formally documented.

AI agents then map these processes, identify opportunities for automation, and determine where human intervention remains necessary. Other agents build the automations and generate interfaces tailored to individual employees.

The platform carries out approved steps across ERP systems, supplier portals, spreadsheets, and inboxes, using APIs where available and existing software interfaces where they are not.

This approach allows companies to introduce their first automated process within weeks, without large engineering teams or extensive systems integration projects. Čupr explained:

“Jane and John might have the same job title in merchandising, but because they work with different suppliers, their jobs are slightly different in practice.”

The platform can account for these differences while allowing managers to monitor tasks and approvals.

“We’re not claiming to replace the underlying systems. ERP and warehouse management systems still have their place. We’re abstracting away their messy interfaces and creating something more useful on top.”

When a process requires human input, Duvo routes the decision to the relevant employee. Dashboards show outstanding approvals and completed tasks, reducing the need to manually transfer information between disconnected systems.

Why Duvo became a separate company

Although Rohlik provided the inspiration and an early use case for Duvo, Čupr decided that developing the technology within the online grocery business would have been financially impractical.

“Rohlik could never have financed it.”

Building the platform required substantial research and development investment, which Čupr believed would be difficult to justify within a consumer business.

“But I thought that if we could solve Rohlik’s problems through a separate company, perhaps we could solve the same problems for everyone else.”

Rohlik subsequently became a Duvo customer, and Čupr credits the technology with contributing to its operational transformation.

AI as a productivity tool rather than a cost-cutting exercise

Čupr argues that the biggest opportunity for AI lies in increasing employee productivity rather than reducing headcount.

“For a while I measured AI in heads, counting how many people an agent could replace. It’s the wrong unit.

At Rohlik, the gains I’d point to came from work nobody had the hours for. Our planners used to check about eight at-risk products a day per market for shortages. Now every signal gets checked, around the clock.

We think of it as raising the GDP of every employee: more value from the people a company has, not the same value from fewer of them.”

He also points to software development as an example of how AI can significantly increase individual output. In sectors such as retail, where staffing costs represent only part of total expenditure, Čupr believes improving productivity across an organisation could deliver greater benefits than workforce reductions.

“You can save some money through automation, but you can create far more value by increasing productivity.”

He also rejects the idea that AI will necessarily take over entire jobs.

“I don’t believe there’s a world where AI simply does everyone’s job from beginning to end. But it can dramatically increase what people are capable of achieving.”

Building enterprise AI without armies of engineers

As competition in enterprise AI intensifies, Čupr sees Duvo’s ability to operate without extensive deployment teams as a key differentiator. Many AI companies rely on forward-deployed engineers who work directly inside customer organisations to understand their processes and implement technology. Čupr believes that approach will become increasingly difficult to sustain.

“Eventually, enterprises will question the sustainability of paying for dozens or hundreds of engineers to work inside their organisations, absorbing their knowledge.”

Duvo instead aims to make process discovery, automation development, and deployment part of the product itself.

“We want the product itself to do the work.”

Čupr argues that improvements in AI models are making this increasingly feasible. Tasks that previously required extensive engineering work can now be handled directly by software, potentially reducing implementation costs.

Fundraising and the sustainability of agentic AI

AI-powered enterprise software is a crowded market, and Čupr is sceptical that current customer growth necessarily indicates long-term viability.

“Some companies have significant traction today, but they may struggle when their contracts come up for renewal a year from now.”

He draws parallels with the online grocery investment boom of 2021, when companies such as Gorillas and Getir attracted valuations that exceeded Rohlik’s.

“I saw something similar in online grocery. In 2021, I looked at the valuations of Gorillas and Getir with envy. Of course I wished Rohlik had those valuations.”

Čupr expects a similar correction in the AI market as companies face pressure to demonstrate sustainable economics.

“When the market settles, the winners will be the companies that can deliver the most value at a sustainable cost.”

Why many agentic AI projects may fail

Čupr agrees with Gartner’s prediction that more than 40 per cent of agentic AI projects will be cancelled by the end of 2027 because of rising costs, unclear business value, and inadequate risk controls.

He believes security, implementation costs, and the ability to demonstrate measurable results will increasingly determine which AI providers survive.

Duvo is model-agnostic, allowing companies to switch AI models without rebuilding their workflows. Organisations with stricter security or data sovereignty requirements can run open-weight models on dedicated GPUs, connect models hosted in their own cloud or on-premises infrastructure, or use existing approved AI endpoints.

The platform also records approvals and changes made to connected enterprise systems, providing an audit trail. Duvo holds SOC 2 Type II, ISO 27001, and ISO 42001 certifications, and is available to US customers through Microsoft’s Azure Marketplace.

To control operating costs, Duvo selects different open-weight and proprietary models based on each task’s requirements. Čupr believes AI deployment economics will become increasingly important as the market matures.

“I believe we’ll see significant disruption among AI companies that have raised billions but ultimately prove too expensive to operate.”

Beyond retail: Banking, insurance, and energy

Duvo works with companies including Morrisons, Holland & Barrett, Boden, Notino, and Allegro. Duvo says its technology has generated more than $100 million in value for customers across finance, supply chains, customer service, marketing, and HR.

At Notino, the European online beauty retailer, Duvo says the reconciliation of supplier bonuses worth approximately $45 million annually has been reduced from 70 days to 15, while the specialist time required per cycle has fallen from around 120 hours to fewer than 10.

Although Duvo originated from the challenges of running an online grocery business, its technology is also being used in banking, insurance, and energy.

Čupr argues that the underlying problems are similar across industries: fragmented systems, undocumented workflows, and employees spending significant amounts of time on repetitive administrative tasks.

“A lot of agentic AI companies build agents and then ask businesses to change their operations to accommodate them. We take the opposite approach. We meet companies where they are.”

He gives the example of finance teams spending days reconciling information across multiple spreadsheets.

“Those employees are already overwhelmed, just trying to close the books and make sure everyone gets paid.”

Expansion into the US and beyond

Duvo is expanding into the US, where Čupr sees a different attitude towards adopting enterprise technology.

“The US is generally more willing to try something new, but customers can be less sticky.”

European companies, by comparison, may take longer to approve new software but are often more likely to remain with a provider once they have committed.

Duvo is also exploring opportunities in Asia, although Čupr says it is too early to draw conclusions about the region.

Duvo currently employs around 70 people and expects to direct most of its new investment towards sales, marketing, and partnerships. The company has appointed a senior vice-president to lead its commercial expansion and is developing relationships with consulting firms. Čupr sees an opportunity for firms such as Deloitte, BCG, and Bain to use Duvo Clarity to identify operational problems and tasks that could be automated within their clients’ businesses.

“We’ve been relatively invisible until now, but I think that’s going to change.”

Two executives are leading the expansion. Michael Green joins from Databricks as Senior Vice President, Go-to-Market, having previously held a senior vice-president role at Salesforce. Spencer Insolia has joined as Managing Director, US, following roles as President of Kmart and Managing Director at Accenture.

Looking beyond the AI hype

Čupr expects AI to become an ordinary part of enterprise technology rather than a standalone selling point.

“At some point, we won’t talk about AI as something special. It will simply be like electricity or the internet.”

He believes the industry will increasingly be judged by the same economic considerations as other businesses, including profitability, customer retention, and long-term value.

“I’m building a legacy company, not one designed to raise money quickly and sell.

As someone who has spent years building businesses, I feel a responsibility to talk about increasing the productivity of employees and the collective intelligence of companies.”

Rob Moffat, Partner at Balderton Capital, also points to the difficulty of deploying AI across large organisations.

“Getting AI agents to work reliably across the enterprise remains incredibly difficult,” he said.

“Duvo’s unique ability is in how it uses AI to capture how work actually happens in the enterprise and what the causes of loss are. Duvo then automates these processes in the same integrated system. This is only possible with the deep operational experience of the founder, Tomáš.”