Duna

When a supervisor comes knocking, the final AI decision isn’t enough. You need to show how you got there: → what the AI was asked to do → what evidence it considered → what it returned → how that output contributed to the decision But that doesn’t mean handing over every event your system has ever recorded. Instead, give them a clear view of the final decision and supporting evidence, with the full history available if a supervisor wants to dig deeper. So what should that record look like? Read the article to find out: https://lnkd.in/e4j7p656

Duna is on the Sifted AI 100 for the second year in a row. The ranking lists Europe’s 100 most promising AI-native companies in 2026. Proud to be alongside inspiring companies like Attio , Cradle , and Swap . Read our press release: https://lnkd.in/e6z7_jRh Thank you to Sifted and N47 for the recognition! #SiftedAI100

As a long-time bol fanatic, I had a small fanboy moment when they chose Duna to onboard their ~43,000 sales partners under their payments licence. Numbers so far: • Partners now take just 12.2 minutes to onboard - down from 14 days • 81% go straight through (STP) • Partners score the onboarding 83/100 on ease (high for business customers) bol’s team raised the compliance bar for a payments licence and made onboarding faster at the same time. That’s a rare feat. Built together with Josje Fiolet, Evelien Mols, Coen van Meulen, Alexander van Blommestein, Lizzy Glaser Mike van Zon, Damien Folkerts, Dylan de Beer, Ferdinand Admiraal. Thank you!

McKinsey & Company 's State of Organizations 2026 found that 86% of leaders believe their organizations aren’t ready to adopt AI in daily operations. In our work with bank compliance teams, we consistently see the same two issues: 1. Policies aren’t ready for automation 2. Tooling is outdated and fragmented These are the two dimensions we focus on when assessing AI readiness for customers. We evaluate the strength of each across lifecycle stages, from initial customer experience through policy changes. This blueprint helps organizations understand what needs to change so they can confidently embed AI and automation into their everyday processes. Find out how Duna can help with your own AI assessment. Drop us a note or read more about it here: https://lnkd.in/eVJmHAyU

Compliance policies written for humans are unusable by AI. They rely on analysts to interpret what the policy means and how it applies to a particular case. But unlike humans, a machine can’t be expected to resolve uncertainty on its own and come up with the same result every time. To automate policies or even run AI agents on top of them, you must first provide machine-readable rules that explain exactly what needs to be done. Start by reviewing each policy and listing questions a machine would need answered before it can act. Then use those answers to convert the policy into agent-ready instructions. Learn more about how Duna and our policy engine help you translate policy into code. Read the full piece: https://lnkd.in/edfEuHzm

New partners on bol used to wait two weeks for an activation code to arrive in the post. Today onboarding takes 12.2 minutes. 81% of cases now close first-time right, with no follow-ups. Analysts spend their time on the cases that need professional judgment. Read the full story: https://lnkd.in/eQHdz-br

We see the company as the second product we are building, and we design it with the same care as the first. Talent density is the core of our company product. We look for builders who want real ownership of their work, and we keep layers few and information open so they have what they need to decide. Read the letter from our founders and see all open roles on our careers page: duna.com/careers

A KYB file arrives as documents, registry data, screening hits, and a company website. Most of a review is reading all of that and lining it up against a policy. That reading is what we automate. Duna's AI agents multiply your output. The judgment stays with your analysts. Five jobs the agents do: 1. Document validation: Every document checked against your accepted types, your validity windows, the fields that have to match. 2. Screening optimisation: PEP, sanctions, and adverse media hits matched against what the file already establishes about the subject, so the queue holds names that warrant a look. 3. Website screening: What a company sells on its own site is evidence. The agent reads it at onboarding, keeps reading it afterward, and raises what falls outside policy. 4. Decision recommendations: A suggested action with the reasoning visible: which rule, which document, what is missing. 5. Smart case summaries. Every source document named and dated, so the audit trail is written while the case is worked. Duna's AI agents assemble and read the file so the analyst can apply their judgment.