What the hell does Duvo do?
Apparently "AI for enterprise operations" is not a satisfying answer. Here is the proper one.
I founded Rohlik and grew it from zero to over $1.6 billion in revenue across five European markets.
My co-founder, Marek Paris, came from Rossum, the AI document automation pioneer. He led AI Automation at Rohlik. Duvo’s first deployments went into production at Rohlik through full enterprise procurement. Both sides of the table: buyer and builder. We are not learning the enterprise.
At Rohlik, the hard part was never knowing that an action had to happen. The hard part was knowing which action mattered. When. Why. With what context. Under which control.
Here is what I saw.
Most enterprises do not just have work being done manually.
They have high-leverage work that is not done at all.
The supplier follow-up nobody has time for. The fraud alert that ages out before someone reviews it. The denied healthcare claim that never gets appealed. The long-tail invoice dispute that is not worth a human’s time. The process improvement everyone knows should happen but never gets staffed.
That is where the future is.
Duvo maps how operations actually run, identifies the work that should run, redesigns the process, and then runs the improved version with AI agents. Under the control of the people who own the business outcome.
Most AI companies are building agents. Agents execute tasks.
Duvo runs the loop those agents live in.
At the beginning, Duvo captures the real process: screen walkthroughs, interviews, SAP data, documents, handoffs, exceptions, workarounds, costs, and controls. The documented process is usually fiction. The real process lives in inboxes, Excel, old portals, Slack threads, and the operator who says: “We don’t do it that way anymore.”
Then Duvo turns that reality into an executable operating model.
Detect. Decide. Act. Document. Learn.
A supplier misses confirmation. Duvo sees it, gets the context, follows the policy, asks for approval when judgment matters, acts when the decision is low-risk, writes back to the system, and leaves an audit trail.
The operator does not disappear. The operator becomes the controller of the system.
They set boundaries. They review exceptions. They approve changes. They stop carrying the process in their head and start operating the loop.
This is the part I am most excited about. This is what we are building towards.
Agents should not only run processes. They should understand whether the process is working.
If an agent runs the same workflow every day, sees the same exception every week, and knows the business outcome it is meant to improve, then it should not just execute. It should suggest a better process.
This supplier always replies late. Move the chase two days earlier.
These deductions below €500 are almost never worth fighting. Auto-close them.
This price movement creates margin leakage before the weekly review. Trigger a negotiation the same day.
This handoff creates 40% of the delay. Remove it.
The agent proposes. The company’s process owner decides. Duvo turns that decision into the next run. The process improves the next day.
That is a different way to run an enterprise.
Today, transformation happens in projects. Big diagnostic, steering committee, deck, implementation plan, partial adoption, drift.
In the future, transformation becomes continuous. Every run teaches the system. Every exception improves the map. Every operator decision sharpens the agent. Every process can get better tomorrow, not in next year’s transformation program.
That is also why Duvo has to be a different kind of company.
The default AI playbook now is forward-deployed engineers. Palantir invented it. Anthropic just stood up a $1.5B FDE-led venture with Blackstone and Goldman Sachs. OpenAI built a $4B Deployment Company on the same model. Send engineers into the customer, wire agents into messy systems, learn, repeat. It is powerful. But operational transformation is not only an engineering problem.
The bottleneck is knowing which work should exist.
So we forward-deploy consultants, not just engineers. Ex-MBB or Big-4. They give us credibility with enterprise buyers, accelerate go-to-market, and help customers think through what to automate and how to transform. The runtime is their leverage. Together with our FDEs, one transformation manager does what a partner-led engagement at Big-4 does, at a fraction of the cost.
At Duvo, the transformation memo becomes the agent spec. The change plan becomes the run plan. The steering committee becomes a control loop. The output is not a deck. It is work running every day.
Customer support already has its winner. So does engineering. So does legal. Operations do not yet. The biggest of these markets is still up for grabs. We have about two years.
This is what makes Duvo hard to copy. Not the phrase “system of action.” Everyone will use that phrase. Not agents. Everyone will have agents.
The moat is the operating loop.
Operators inside the customer. Transformation managers inside Duvo. Engineers building the runtime. Clarity discovering the real process. Agents running the work. Business outcomes telling the system what better means. Human owners deciding what changes.
We collapse consulting, integration, and runtime into one accountable operating model. The handoffs that kill traditional transformation are no longer the customer’s problem.
All of this is real today. We started selling in January and we’re already past $1M in revenue.
Five playbooks already run in production: supplier onboarding, out-of-stock exception triage, chargeback resolution, commodity renegotiation, promo setup quality. Each one was built through a real deployment and captured in Clarity.
The consulting channel is moving with us. Deloitte signed as a sales and delivery partner. MBB engagements in motion. We are ISO 42001 certified, one of around 50 companies globally. Backed by Index Ventures and Northzone.
Operations work is universal. Every industry has it. The runtime travels.
This is not enterprise software as a place to record work.
It is enterprise software designed to improve work. All-the-time.
$240 billion in operations consulting and operational-process BPO globally. Trillions more in the operations labor enterprises run in-house. That is what we capture when transformation stops being a project and starts being a system.
So what the hell does Duvo do?
Duvo turns operations into a living system: real work discovered, better work designed, agents running it, humans controlling it, and the process improving tomorrow because it ran today.
Are we there yet? No. But that’s what building a company is about.