Order of operations
Rights, then tools
Name the owner, the evidence, and the failure path before you compare models.
Research
Decision Rights Before Model Choice
Takeaway. Buy model access only after someone owns the decisions that model is allowed to make. Rights and evidence come before tool comparison.
Most AI programmes still open with a vendor comparison. That meeting is too late. If nobody can say who may let an AI output commit the organisation, you are not buying intelligence. You are buying a second review layer you did not budget.
Planning weeks
46
A working year used for planning estimates, not a measured client result.
Default rework share
25%
A planning starting point: a quarter of unofficial AI hours redone because no one owned the decision.
The false first meeting
A bake-off feels rigorous. It is usually a delay tactic. Teams compare latency, context windows, and demo charm while the operating question stays empty: who is allowed to act on the output?
The Decision Maker's Handbook to Data Science is written for executives who have to own that question. You do not need to be the modeller. You do need to own the strategy, the hire, and the standard of evidence.
Four decision rights
Write these on one page before procurement. If a row is blank, you do not have a programme. You have a pilot looking for a sponsor.
- Commit right: who may let an AI output bind a customer, a regulator, or a member of staff.
- Evidence right: what must sit next to the output (source, date, human review).
- Failure right: what happens when the model is wrong, and who is paged.
- Rework right: which budget pays for the second pass.
Uncertainty does not retire when the prose is fluent
Predicting the Unknown is about living with incomplete information, not pretending a fluent interface removes it. If your process assumes the model is correct, you have not designed a control. You have purchased a story.
Fluent text is a delivery format. It is not a proof. Leaders who treat tone as evidence create shadow AI: unofficial use, unofficial edits, and an unofficial second shift of cleanup.
Seats are not a business model
As lead author of Business Models in Emerging Technologies, the test I use is simple: can the organisation price, govern, and repeat the capability? A pile of licences fails that test. A named owner, a measured workflow, and a review path can pass it.
That is why the offer ladder on this page is practice first, then a shared standard, then advisory. Access without judgment is how AI budgets die in the committee.
Tesseract AI Coach
Run the rework math in the Coach
Name the owner, then size unofficial AI hours in the AI Coach inside the Free AI Fluency Course. It is the same product as the AI Coach hub, not a calculator on this page.
Enter your volume and labour rates there. Paid certification saves the result to your Coach as an Opportunity Map.
Open the AI Coach in the free course
FAQ
What is the core claim?
The expensive mistake in 2026 is buying access before anyone owns the decisions the model may make.
What should come before model choice?
Named decision rights, required evidence, and a failure path. Then compare tools.
Is this a software buying guide?
No. It is an operating brief on ownership and governance before procurement.
What is a sensible next step?
Write the decision owner and kill criteria for one workflow, then use free courses or certification to practise the standard.
Sources cited
- The Decision Maker's Handbook to Data Science. Apress / Springer Nature, 2020 and 2024 editions, sole author. How leaders own data and AI strategy.
- Predicting the Unknown. Apress (Springer Nature), 2023, sole author. ISBN 9781484295045. Uncertainty stays after the interface becomes fluent.
- Business Models in Emerging Technologies: Data Science, AI, and Blockchain. Business Expert Press, 2022, lead author with Mourouzis, Cardoso, and Zinopoulou. ISBN 9781637423134. Value is captured when a capability can be governed and repeated.
Practice, then a shared standard
Free courses first. AI Fluency Certification when you want a four-week, project-based credential. Advisory when the organisation needs named decision rights, not another licence pile.
Tesseract Academy · Decision Rights Before Model Choice · Planning estimates only · More research