AI Leadership Development
Enough understanding to judge a proposal, not enough to build one.
What is AI leadership?
AI leadership development gives senior people the judgement to evaluate AI proposals, question vendors usefully and decide what to fund, without requiring them to understand how the technology works internally.
Senior people are being asked to approve AI spending they cannot assess. The usual response is either approving whatever the enthusiastic person proposes, or blocking everything until the risk is understood, and both are expensive.
What actually helps is a small set of questions that separate a serious proposal from a weak one. What problem does this solve, what does it cost per year rather than to build, how will we know it worked, and what happens when it is wrong.
That is judgement rather than technical knowledge. A leader who asks those four questions consistently will make better decisions than one who has learned how a model works.
Why we teach leaders to question us
Duara works with boards and executive teams from Nairobi, usually in short sessions rather than programmes, because senior calendars do not support anything else.
We teach leaders to interrogate proposals, including ours. A leadership session that leaves the room easier to sell to has failed at the thing it was supposed to do, however well it was received.
Who this is for
- You approve AI spending you cannot assess
- Vendors are difficult to evaluate
- Your team disagrees and you must decide
- The board is asking what your AI strategy is
Industries we serve
- Banking and financial services
- Insurance
- Manufacturing groups
- Healthcare organisations
- Public sector
- Large professional services
Problems this solves
These are decision problems under uncertainty, which is ordinary senior work applied to an unfamiliar subject.
- Approving what cannot be assessed. Proposals are approved on the confidence of whoever presented them rather than on their merits.
- Vendors difficult to compare. Every supplier sounds credible and there is no framework for separating substance from presentation.
- Blocking everything instead. The alternative failure is refusing all AI spending until it is understood, which costs opportunity quietly.
- Board pressure for a strategy. A strategy is requested urgently, and producing one that sounds impressive is easier than producing one that is true.
What we deliver
A short set of questions that separate serious proposals from weak ones, a working understanding of what AI does and does not do, and a framework for deciding what to fund and what to refuse.
A question set for evaluating any AI proposal
A working understanding of capability and limits
A framework for funding decisions
Vendor evaluation guidance
Realistic views on cost and timeline
Language to use with your board honestly
Features
- Judgement rather than technical depth. Four good questions asked consistently beat understanding how a model works internally.
- Designed to make you harder to sell to. Including by us. A session that leaves you easier to sell to has failed at its purpose.
- Short by design. Senior calendars do not support programmes, and the useful content fits in half a day.
- Cost framed annually. Build cost is the smaller number and running cost is the decision, which most proposals present the other way round.
- Refusal treated as a skill. Knowing which proposals to decline is more valuable than knowing which to approve, and rarer.
Benefits
- Better funding decisions. Weak proposals get identified before money is committed rather than during the second year.
- Vendors get questioned properly. A supplier who cannot answer what happens when it is wrong has told you something important.
- You stop blocking by default. Understanding enough to judge removes the need to refuse everything as a risk control.
- Board conversations get easier. You can describe your position honestly, including its uncertainty, rather than performing confidence.
Technologies we use
- Capability and limits. What AI reliably does today, what it does badly, and the difference between the two in practice.
- Proposal evaluation. The questions that separate a serious business case from a technology enthusiasm.
- Cost structures. Why running cost matters more than build cost, and how AI spend scales with use.
- Risk framing. Which risks are real, which are overstated, and which are being quietly omitted.
- Vendor questioning. How to test a supplier claim without technical knowledge.
Leaders do not need to know how models work. They need to know what the technology reliably does, what it does badly, and which claims should prompt a follow up question.
How the process works, step by step
Leadership sessions are usually half a day or a series of short ones, because anything longer does not survive contact with a senior diary.
-
Step 1
Free scoping conversation
We establish what decisions you are actually facing rather than delivering a standard session.
-
Step 2
Session built on your decisions
Content is framed around proposals in front of you, since abstract material does not transfer.
-
Step 3
The question set
Leaders leave with a short set of questions they can apply to any proposal, including ours.
-
Step 4
Applied to a live case
A real proposal is worked through in the room, which is where the questions prove themselves.
-
Step 5
Follow up on request
Short sessions later, usually when a specific decision arrives rather than on a schedule.
Quality assurance and how we work
The commitments attached to advising the people who decide.
- We teach you to question us too. A leadership session that leaves the room easier to sell to has failed, however well it was received.
- We do not sell from the session. Training that is a pitch in disguise is worth less than either a good session or an honest pitch.
- We state what AI does badly. A session covering only capability produces confident leaders making poor decisions.
- Running cost is presented as the decision. Build cost is the smaller and more attractive number, and presenting it alone misleads by omission.
- We keep it short. A programme that does not fit a senior diary does not get finished, and an unfinished programme teaches nothing.
Frequently asked questions about AI leadership
Do we need to understand the technology?
No. You need to know what it reliably does, what it does badly, and which claims should prompt a follow up question.
What questions should we ask any proposal?
What problem does this solve, what does it cost per year, how will we know it worked, and what happens when it is wrong.
Why is running cost the important number?
Because build cost is the smaller and more attractive figure. A proposal presenting only that is misleading by omission.
Will this make us harder to sell to?
That is the intention, including for us. A session leaving you easier to sell to has failed at its purpose.
Do you sell during the session?
No. Training that is a pitch in disguise is worth less than either a good session or an honest pitch.
How much of our time will this take?
Usually half a day, sometimes a short series. A programme that does not fit a senior diary does not get finished.
Can you cover a live proposal?
Yes, and it is the most useful format, since abstract material does not transfer to a real decision.
How do we evaluate vendors?
By asking what happens when the system is wrong. A supplier who cannot answer that has told you something important.
What if our team disagrees internally?
Common, and usually the disagreement is about risk appetite rather than technology. Naming that resolves more than technical detail does.
Should we have an AI strategy?
Probably a short one. Producing something impressive is easier than producing something true, and boards eventually notice the difference.
What is a reasonable timeline to expect?
Most useful projects run weeks rather than months, and anything promising transformation in a quarter deserves questioning.
How much should we budget?
Less than most proposals suggest for a first project. Starting small is both cheaper and more informative.
Is it appropriate for a board rather than executives?
Yes, with a different emphasis. Boards need governance and risk framing more than operational judgement.
Do we need a Chief AI Officer?
Rarely at first. The role usually makes sense later than people expect and we would say when.
What are the risks people overstate?
Existential ones. The practical risks are wrong outputs acted on, cost overruns and projects nobody uses.
What risks are understated?
Running cost, staff resistance and the maintenance burden after the exciting part finishes.
Can our managers attend too?
A separate session usually works better, since the decisions and the vocabulary differ by level.
Will this help us say no?
That is much of the value. Knowing which proposals to decline is rarer and more useful than knowing which to approve.
Is what we discuss confidential?
Yes. What is said about your proposals and internal disagreements stays in the room.
What does a leadership session cost?
Format and group size determine it, quoted after the scoping conversation.
Glossary of terms
Terms leaders encounter in AI proposals.
- Proof of concept
- A small test of whether something works, frequently presented as though it proved it would work at scale.
- Running cost
- What the system costs per year in use, as distinct from what it costs to build.
- Hallucination
- Confident output that is wrong, which no supplier can eliminate and honest ones will say so.
- Human in the loop
- A person reviewing before action, which is a design decision with a real cost and a real benefit.
- Model
- The trained system doing the work, which leaders do not need to understand internally.
- Scope creep
- A project growing beyond what was approved, which AI projects are unusually prone to.
Related services in AI Training
Talk to us about AI leadership
Take the last AI proposal you saw and ask what it costs per year and how you would know it worked. The answers usually settle it.
Start your project