01Derk Disselhoff
Founder & Director
- The studio
AI expertise, human intuition.
A small, specialised team based in Amsterdam, bringing practitioner expertise to enterprise data and AI transformation.
Dissel AI was founded by Derk Disselhoff, bringing 10+ years of hands-on experience in complex commerce, product data management, and AI transformation.
We stay small by design so every engagement has a founder in the room - and we don't take on work we can't measure.
01
0%+
Data completeness needed before AI pays off
02
0 days
From fragmented data to AI-powered operations
03
0 mo
Typical payback period on data + AI investment
04
0+ yrs
Hands-on commerce, data and AI experience
- The team
The team
01
02Elko Lemiso
AI Engineer
Manifesto
Thestaffingmarketrunsonexceptions:deviatingcontracts,hoursandbillingforsecondment,freelancingandpayroll.WeapplyAIaroundyourcoresystemstofixdiscrepancies,bridgegapsandresolveedgecases—sothewholechainrunsmoreefficiently.
The problem isn't AI. It's a missing underlying foundation able to carry every exception in your business model.
Derk DisselhoffFounder, Dissel14Kfollowers
- Commercial exceptions become process exceptions — different rate, different contract, different hours flow
- The ATS covers part; the rest is built custom
- Year-one architecture choices block year-five sales
- Five systems each manage a piece; none knows the whole
- Growth is absorbed by adding more FTE
- Onboarding, contracts and hours stay manual
- Engineering delivers slowly — new propositions never reach the market
- More FTE per million in revenue: you grow, and your EBITDA shrinks
Why 
, 
, 
or 
alone are not enough
A licence on an AI model is not an AI strategy. The difference between users who type more cleverly and a company that structurally runs differently sits in integration: your data, your systems and your workflows connected, with agents built into your processes and kept under control.
An AI model alone is not a system
It is a component. The car still has to be built.
No foundation underneath
Without connected data and systems it stays a loose chat window.
No control on the output
Run an agent in your real processes and "sometimes wrong" is not acceptable.
No owner for continued development
Building is one-off; models, processes and integrations change continuously.