Bjorn Claes · Writing
Decision-grade blueprints
Four articles on one early discipline in AI products: making a picture of the product accurate enough that a team, a client and a lawyer can make decisions on it, before anything is built.
The same candidate. Two products.
Demo
87%
AI match · Marek W.
An answer that looks final.
Blueprint
- Certificate incompletescan half legible
- Practical test proposedVictor decides
A decision that stays checkable.
PART 01 / 04
A picture that holds up
What are we actually building?
Every AI project starts with one sentence and four different products in four heads. An accurate picture turns them into one product, before anything is built.
7 min read
PART 02 / 04
One screen, ten decisions
Who decides, and on what basis?
A screen that looks like an answer is a stack of decisions. Here are ten of them, from the file of one welder with a Polish CV.
7 min read
PART 03 / 04
Showing uncertainty without a confidence score
What does the system really know?
A confidence score looks precise and says little. What works instead: showing what a finding rests on, and what should happen when that ground is weak.
9 min read
PART 04 / 04
Seen, out of scope
What happens to what you saw outside the brief?
Design precisely and you see more than you were asked to. Staying silent doesn't work, and neither does quietly delivering it. A list does.
6 min read
Written by
Bjorn Claes
I design the early phase of AI products: before the build, often before the quote.
If one of these articles touches a project of yours, I’m glad to hear about it.
- Emailbjorn@aikonaut.com
- LinkedInlinkedin.com/in/bjornclaes
