May 7, 2026
If You Want ROI From AI Agents, Organise Them Like People
AI agent ROI comes from an operating model: scoped roles, accountable owners, clean inputs, risk-based controls, and metrics that prove work moved rather than noise increased.
I’m Frank van den Brink, an interim CTO and engineering leader for European scale-ups and tech departments of 20–250 people. I step in when delivery has slowed down, the tech is fighting the business, or nobody can make a decision stick.

I work alongside the people doing the work. I look at the code, the delivery system, the organisation and the decisions being avoided. Then we remove the constraint and make the changes stick.
No imposed framework. No deck handed over at the end. No process theatre.
An 80-person SaaS team got a backlog moving
Planning overhead dropped by 80%. A feature that had taken six weeks shipped in two.
A 60-person SaaS team found the real bottleneck
Out of 47 microservices, one dependency was blocking delivery. Coordination for affected features fell from 14 weeks to five.
A 150-person post-acquisition fintech stopped the decline
Clashing cultures and duplicate systems had cut delivery speed by 50%, and key talent was leaving. The teams were unified around shared practices, stopping the decline and restoring momentum.
Organisations I have worked with
May 7, 2026
AI agent ROI comes from an operating model: scoped roles, accountable owners, clean inputs, risk-based controls, and metrics that prove work moved rather than noise increased.
Mar 8, 2026
Most engineering organisations skip three foundations before chasing AI productivity. Boundary clarity, technical ownership, and contribution norms matter more than tooling.
Feb 12, 2026
A rewrite proposal is not a technical plan. It is the final signal in a feedback loop that has been failing for months. How leadership responds determines whether the engineer stays.