Duvo’s new funding puts a Czech-founded company’s approach to enterprise automation in the spotlight: understand how work happens, connect the systems involved and keep people responsible for decisions.
A supplier bonus needs checking. The agreement sits in one system, transaction data in another and the explanation for a discrepancy in an email. Someone has to bring those pieces together before the business can settle the amount.
That kind of work helps explain the proposition behind Duvo, co-founded by Rohlik founder Tomáš Čupr. On 14 October 2026, the company announced a US$32 million Series A, led by Balderton Capital, with participation from existing investors Index Ventures, Northzone and Credo Ventures. Duvo — funding announcement
The investment brings reported total funding to US$48 million and supports international expansion, starting with retail and consumer-goods companies. Balderton confirms the round in its own announcement. The two figures describe the new round and cumulative funding respectively; neither is a company valuation. Balderton — investment announcement
A founder bringing retail operations into software
Duvo was founded in 2025 by the team behind Rohlik Group. Its connection to retail is useful context: purchasing, stock availability, supplier agreements and financial reconciliation create recurring work across several systems.
The new funding also supports a commercial push into the United States. Duvo’s announcement names Spencer Insolia as Managing Director, U.S., and Michael Green as Chief Commercial Officer. The company says its initial expansion will focus on retailers and consumer-goods businesses. Duvo — company background and expansion plans
For international readers following Czech technology, this is a concrete example of operating experience becoming the basis for a software business. Experience can help a team identify a valuable problem. The product still has to work reliably for customers with different systems and working practices.
What Duvo proposes to automate
An AI agent is software that can carry out defined steps using connected tools and information. In a business process, its usefulness depends on the task, its access and the decisions it is allowed to make.
Duvo describes a platform that first maps existing work through interviews, recordings and documents. Its agents then perform approved steps across enterprise resource planning systems, supplier portals, spreadsheets and inboxes. The company says it uses application programming interfaces where available and screen-based interaction where needed. Decisions and exceptions are directed to responsible people. Balderton — how the platform works
The appeal is easy to understand. Many companies have already bought the systems they need, yet employees still spend time moving information between them. Bringing that work into a controlled process could reduce repetitive handling and make outstanding decisions easier to see.
The practical challenge is keeping the process dependable when a document is incomplete, a supplier changes its portal or a transaction falls outside the usual pattern. Those are useful situations to include in a demonstration.
Start with one process and a measurable result
Duvo’s release reports a specific example at Notino: quarterly supplier-bonus reconciliation shortened from 70 days to 15, with specialist effort falling from around 120 hours per cycle to under 10. These are supplier-published customer results, rather than independently verified performance figures or a forecast for another business. Duvo — reported Notino example
The example gives a prospective customer something useful to investigate: the time between starting a task and completing it, the staff effort required and the quality of the result.
A first pilot should establish those measures before automation begins. A sensible review might cover:
| What to measure | Why it matters |
|---|---|
| Time from task intake to completion | Shows whether work moves through the whole process faster |
| Staff time, including exception handling | Captures the review work that remains |
| Correctly completed cases and subsequent corrections | Tests output quality alongside speed |
| Decisions waiting for approval | Reveals where human responsibility affects throughput |
| Integration, usage and support costs | Helps assess the operating economics of the deployment |
Suggested evaluation measures for an automation pilot; these are not Duvo’s published contractual commitments.
For instance, a business could begin with one supplier-reconciliation workflow, a defined set of documents and a named reviewer. It can then test routine cases, missing records and disputed amounts before expanding the scope.
The human decision still needs a clear owner
Approvals become useful when everyone understands what they mean. A finance team may allow software to collect records and propose a match while keeping settlement approval with an authorised employee.
The following diagram illustrates that division of work. It is an example for planning a pilot, rather than a reproduction of Duvo’s interface or technical architecture.
Loading diagram…
Read diagram as text
flowchart TD
A["Agreement and transaction records"] --> B["Agent compares records and prepares a result"]
B --> C{"Exception or approval required?"}
C -->|Yes| D["Responsible employee reviews the case"]
C -->|No| E["Complete the permitted step"]
D -->|Approved| E
D -->|Rejected or unresolved| F
E --> F["Record the outcome and measure the process"]
classDef source fill:#ffffff,stroke:#ded5ca,color:#282623;
classDef process fill:#e8ded1,stroke:#c7b8ac,color:#282623;
classDef human fill:#ece2e5,stroke:#c2adb3,color:#57373f;
classDef result fill:#282623,stroke:#282623,color:#f8f4ed;
class A source;
class B,C,E process;
class D human;
class F result;Before deploying such a workflow, the customer should agree which records the software can read, which changes it can make and when it must stop for review. The team also needs a way to correct an error and continue the work if an integration becomes unavailable.
What this means for companies operating in Czechia
International companies with Czech operations can use the same approach when assessing an automation project. Choose a recurring task involving several systems, identify its owner and define a result that the business can measure.
Local delivery can include Czech and English process documentation, a review of source data, integration work and support for the people using the system. These functions deserve attention alongside the choice of AI product.
For a prospective partner, the first discussion should establish the customer segment and the role to be delivered: introducing a buyer, mapping a process, implementing connections or providing ongoing support. A funding announcement creates a reason to investigate the company; a commercial relationship depends on a specific fit.
The main conclusion
Duvo’s Series A supports an international expansion built around a recognisable business problem: work that falls between existing systems. Its progress is worth following through customer deployments, completed processes and the effort required to maintain them.
For a potential customer, the next step is a defined pilot with agreed permissions, human approvals and measures of accuracy, time and cost. That gives the business evidence for deciding whether to expand.
How Kodo can help
Kodo helps international software, retail-technology and automation companies prepare the commercial and operational side of entering Czechia. We support early market validation, partner mapping, local positioning, English and Czech communication, web content and coordination of the first implementation steps.
For legal, tax, regulatory, certification and investment-incentive matters, companies should work with qualified specialists and the responsible authorities.
Related reading
- Flae Robotics: AI for the hotel front desk
- Mews: from Prague hotel to global platform
- Czech startups: sectors, funding and opportunities
Sources
- Duvo: $32 million Series A, company background, expansion plans and reported customer example — 14 October 2026
- Balderton Capital: investment announcement and platform description — 14 October 2026
This article provides general market and technology information, not legal, tax, regulatory or investment advice. Companies should verify product capabilities, commercial terms and requirements for their own deployment.