Engineering notes · 2026-06-12 · 4 min read
Where AI belongs in a business system
Most AI projects fail not because of the technology, but because they start from the wrong place. Companies first say “we want AI”, then go looking for a problem for it. The order should be reversed.
Our approach is simple: look at the process first. Where are people doing repetitive work that can be expressed as rules? Where does data exist, but decisions stay slow? Those points are where AI actually belongs.
The second principle is measurement. Before deploying anything, we take a baseline: processing time, error rate, response time. After deployment we measure the same indicator again. If there is no difference, we don't keep the system.
The third principle is human oversight. Automation doesn't erase responsibility for decisions. At critical points, a person still approves — the system just makes their work faster. That's how AI becomes a tool instead of a promise.
Back to insights