Enterprise AI is often judged by whether it can complete familiar tasks more quickly. A potentially larger prize lies elsewhere: work that organisations know matters but rarely manage to carry out consistently.
Duvo, a Czech AI-agent company co-founded in 2025 by Rohlik Group founder Tomáš Čupr, describes this neglected activity as “abandoned work”. The problem, as outlined in the article in The Patent Lawyer, is not a lack of awareness. Teams may know what needs doing, ...
Continue Reading This Article
Enjoy this article as well as all of our content, including reports, news, tips and more.
By registering or signing into your SRM Today account, you agree to SRM Today's Terms of Use and consent to the processing of your personal information as described in our Privacy Policy.
For intellectual property departments, this offers a different way to assess AI: not simply asking which existing tasks can be automated, but identifying valuable processes that are skipped, delayed or performed only occasionally.
Rohlik, an online grocer operating in the Czech Republic, Germany, Austria, Hungary and Romania, provides one example. The company’s commercial teams tracked commodity prices and sought corresponding changes to supplier terms, but the work was carried out quarterly at best. Duvo’s agents began monitoring prices and initiating supplier contact, extending the process beyond periodic manual reviews.
The Patent Lawyer reports that Rohlik recorded more than €2.8 million in annualised savings within three months and lifted inbound delivery confirmations from 52% to 90%. Duvo’s own customer-stories page also cites the improvement in confirmations. These are company-reported outcomes, not independently audited findings; the available accounts do not set out a complete methodology or baseline.
A separate Duvo customer story describes more than €1.45 million in annualised supplier savings after the company renegotiated terms on more than 120 stock-keeping units with over 15 suppliers in a week. That is a distinct reported result, with its own scope and period, and should not be conflated with the €2.8 million figure. Together, the accounts illustrate the potential to expand the reach and frequency of commercial work, but they do not establish what other organisations would achieve.
The distinction matters when measuring AI. Time saved can capture improvements to work already being done, but it misses tasks that have no reliable baseline because they happen irregularly or not at all. The article argues that organisations should also track how much work is covered, value recovered and interventions made earlier. Software licences and prompt volumes may indicate adoption; on their own, they do not show whether neglected work is being addressed.
Duvo’s founding reflects the operational challenge it is targeting. WIRED Czech reported that the company was established by Čupr alongside former senior Rohlik executives Marek Paris and Martin Pecha. Its agents are designed to operate across internal systems and supplier portals, as well as handle spreadsheets and business correspondence. The report also said Duvo raised a $15 million seed investment led by Index Ventures. That background explains the company’s focus, but does not independently substantiate its reported customer results.
For IP and research teams, one opportunity is regular portfolio reconciliation. Records in an intellectual property management system can be checked against patent registers, assignment information and deadlines, helping identify inconsistencies before they become costly problems. The same approach could prompt follow-up on invention disclosures set aside for later review, when a product or market change may have made them more relevant.
Competitor monitoring is another candidate. Instead of commissioning landscape reviews only before a product launch, negotiation or dispute, teams could track filings and other public developments on an ongoing basis. That may help turn intelligence gathering from an occasional exercise into a routine source of information.
Patent maintenance decisions also merit repeated scrutiny as commercial circumstances change. The United States Patent and Trademark Office sets maintenance-fee deadlines at 3.5, 7.5 and 11.5 years. In Europe, costs depend on the form of protection: the European Patent Office explains that a Unitary Patent carries a single renewal fee, while a conventional European patent may require separate national fees in each country where it is maintained. Those arrangements make renewal reviews a question of where protection remains worth the expense, not simply whether a payment is due.
These processes can reinforce one another. Better-maintained records can support monitoring, licensing and decisions about where to continue protection. The central proposition is that AI’s value may be greatest not when it makes an established task marginally cheaper, but when it makes previously uneconomic work practical to perform.
Source: Noah Wire Services



