THE AI LEADERSHIP PROGRAMME
AI is not another IT project. It is your operating model.
20-Session Programme · 25 Minutes Per Day · Free · For sitting executives
WEEK 4 OF 4 · THE NINETY DAYS THAT PROVE IT · SESSIONS 16–20 · WORK AT YOUR OWN PACE
Execution · Days 1–30SESSION 16
Days 1–30: The Diagnostic That Does Not Flatter Anyone
Why This Matters
The first thirty days cannot be presented as progress. Nothing is demonstrated at the end of them — only an honest account of where your business loses money and what data you cannot trust. I have read diagnostics written to protect three senior people, and they cost their company a year. This session gives you the instrument instead: small teams, one question each, a fixed end date, and one page of findings you can compare across the business. Commission it properly and the next sixty days almost design themselves.
25-Minute Module
0:00 – 5:00
The five questions Days 1–30 must answer
- Days 1–30 produce one deliverable: an AI Readiness Diagnostic answering five questions about your own business. Where is value being lost? What data exists? What data cannot be trusted? Which use cases pay fastest? What must be fixed before anything is scaled?
- No vendor, no architecture, no tool. Those questions are not unimportant; they are premature, because the answers change once these five are answered honestly.
- THE DOCTRINE
A diagnostic that protects egos is worthless. The one that names the real gaps is what makes the next sixty days productive.
If the diagnostic tells you nothing you did not already believe, it was a confirmation exercise.
5:00 – 11:00
Small teams, one question, a fixed end date
- You do not run this yourself. You deploy a rapid-assessment team: a sponsor, a team lead, an operations person with the ground truth, a data lead, one domain specialist. Seven to fourteen working days, one problem.
- Its charter is written before it deploys: the mandate as a single question, the scope in and out, the timebox, the access needed, and the date you will decide.
- Then the rules of engagement. Time-boxed — the sprint ends on the report date, finished or not. Evidence over opinion. Quantify the value. One problem at a time. Standard format. End in Go, Hold or Stop.
Write the decision date into the charter before the team starts. A finding with no scheduled decision becomes a document.
11:00 – 18:00
The one-page findings report
- Every team reports on one page, in the same format. That is what lets you compare nine opportunities in an afternoon.
Line on the page What it must contain Readiness score The scored current state, with its evidence Top three findings Each backed by a record, a log or a reconciliation Value at stake Quantified per month in your own units Recommended pilot One bounded, measurable pilot, in one sentence Risks and blockers What would stop it, named Decision ask What the team needs from you to proceed Recommendation Go, Hold or Stop - THE DOCTRINE
If a team cannot fill in the value-at-stake line, the work is not finished. Value is the currency of every decision that follows.
- ENERGY · THE WORKED CASE
- A mature onshore asset: most of last quarter's losses sat under one deferment code, operational. Nobody could attribute the loss, so nobody owned it.
- Value at stake became barrels deferred per month against the allocation record, a figure the production manager could defend to finance.
OUTSIDE ENERGY · THE PARALLEL- A contractor logged site rework by trade and never priced any of it.
- Value at stake became rework hours per project against the approved programme, a figure the project director could defend to the client. Same instrument, different unit.
Ask for the page. If a team sends nine slides instead, send them back for the page.
18:00 – 22:00
Your job, and the three words you will use
- THE DOCTRINE
You own the what and the whether. The team owns the how and the how-much. Neither does the other's job.
- Three things only you can do: commission it with real access, protect it from politics, read it honestly. Soften one finding on its way to your desk and you no longer own an instrument.
Recommendation What it means What you do next Go Value quantified, data adequate Fund the pilot at the next gate Hold Value real, a precondition missing Fund the precondition, not the pilot Stop Value too small, or the data cannot support it Say so in writing and release the team
Hold is the most useful of the three words and the least used. Stalled programmes are full of pilots that should have been holds.
22:00 – 25:00
Before you close the laptop
- Write down four things: the area you will diagnose first, who leads it, the report date, and the date you decide.
- This session produces Section 16 — your diagnostic findings and the value at stake. Every later section is priced against it.
- The assignment is a compressed diagnostic on one area of your own organisation, in the standard format, ending in one word.
The area you least want to look at is usually the one carrying the number.
Key Terms
AI Readiness Diagnostic
The deliverable of Days 1–30: a written answer to five questions — where value is lost, what data exists, what cannot be trusted, what pays fastest, what must be fixed first.
Rapid-assessment team
A small, cross-functional, time-boxed team on exactly one problem, existing to assess, quantify and recommend rather than to build.
Value at stake
The loss a finding carries, quantified per period in a unit your business already uses, with the record it came from. An adjective is not a value at stake.
Timebox
A fixed start and report date agreed before work begins. The sprint ends on the report date, and any remaining gap is reported as a finding.
Self-Check — Before You Submit
1. A team returns a scored current state, three well-evidenced findings and a clear recommended pilot. The value-at-stake line reads “significant”. What do you do?
- A. Approve the pilot, since the findings are well evidenced
- B. Score it into the portfolio using a provisional value you set
- C. Send it back: the work is not finished until value is quantified in a real unit
- D. Reject the recommendation and stop the initiative
C. Send it back: the work is not finished until value is quantified in a real unit — An unpriced finding cannot be ranked against anything else. Approving on the strength of the evidence is the tempting error: a loss you have not priced still leaves you funding on instinct.
2. What is the deliverable of Days 1–30?
- A. An AI Readiness Diagnostic answering five questions about the current state
- B. A prioritised portfolio with the first three initiatives named
- C. A signed pilot charter with a baseline and two named owners
- D. A vendor shortlist and a target architecture
A. An AI Readiness Diagnostic answering five questions about the current state — The portfolio belongs to Days 31–60 and the charter to Days 61–90, each depending on the diagnostic before it. Naming the vendor first commits money before anyone knows where value leaks.
3. A report is due Friday. On Wednesday the function being assessed asks for three more weeks so the picture can be complete. What do you do?
- A. Grant it, because an incomplete diagnostic is worse than a late one
- B. Hold the report date and publish what the team has, with the gaps named as findings
- C. Move the sprint to another function and return later
- D. Extend by one week as a compromise
B. Hold the report date and publish what the team has, with the gaps named as findings — The sprint ends on the report date, finished or not, and a gap is itself a finding. Extending to complete the picture turns a diagnostic into a consultation, and the request comes from the most exposed function.
Today’s Assignment
Run a compressed diagnostic on one area of your own organisation and write it up as the one-page findings report.
- Choose one area where you believe value is leaking: one asset, one process, one function, one customer segment.
- Name the team you would deploy by role, and by name where you can, with the timebox and the decision date you would set.
- Score the current state and write the evidence beside each score. A score with nothing beside it earns nothing.
- Quantify the value at stake in your own unit per month and name the record it came from. Then name the data gaps, including data you would not trust, and who would own each.
- Close with one bounded pilot in a single sentence, a recommendation of Go, Hold or Stop, and the honest paragraph: the finding your organisation would find uncomfortable, and who would try to soften it. Four hundred to nine hundred words.
YOUR CHARTER
Section 16 — Our diagnostic findings and the value at stake.
How this is marked — the rubric
| Quantifies value at stake in a real unit per period and names the record it came from | 30 |
| Scores the current state with specific evidence — a record, a log or a reconciliation, not an opinion | 25 |
| Names the data gaps and a prospective owner by role for each | 20 |
| Ends with one bounded pilot in a single sentence and an unambiguous Go, Hold or Stop | 15 |
| Includes the honest paragraph naming the uncomfortable finding | 10 |
| Total · pass mark 70 | 100 |
Submitting session 16
Write your answer in Word or Google Docs. The first line must read
Write your answer in Word or Google Docs. The first line must read
Session: 16, or the submission is refused. Upload the Word file (.docx), or switch to the Google Doc tab and paste your document’s share link.Marked against the rubric above and returned as a written letter, saved to your progress page. Do not submit until you have passed the self-check above.Capabilities Covered
Diagnostic commissioningEvidence over opinionValue quantificationTimebox disciplineGo, Hold or Stop
Execution · Days 31–60SESSION 17
Days 31–60: Designing the Operating Model and Choosing What to Fund
Why This Matters
This is the session where AI stops being a conversation and becomes a budget line. By the end of Days 31–60 you must be able to say in public which three things your organisation is doing first, and why those three rather than the nine others people are enthusiastic about. Most executives make that choice on advocacy: whoever presented last, or loudest. I would rather you made it on arithmetic you can show, using weights you fixed before you saw the candidates. Unlike advocacy, it survives a board challenge.
25-Minute Module
0:00 – 4:00
What the second thirty days produce
- Days 31–60 are for structure and choice. The structure — governance committee, data-ownership model, integrations, cyber controls, training plan, partner strategy, budget, metrics, risk controls, timeline — you have largely drafted already.
- The choice is the part nobody can make for you: a prioritised portfolio and the priority pilots. Structure without choice is a committee. Choice without structure is a favour.
- THE DOCTRINE
Opportunities will always outnumber the money. The executive act is not generating candidates; it is ranking them against a rule you set before you saw them.
Fix the weights before you look at the list. Weights chosen afterwards are a preference with decimal places.
4:00 – 9:00
Five criteria, and what each is weighted
- Every candidate is scored 1 to 5 on five criteria, and those scores collapse into one priority score.
Criterion What it measures Weight Value Impact on output, cost, safety, emissions or revenue 35% Readiness Can we do it now? The diagnostic score, not the ambition 20% Speed How fast to a measurable result 20% Strategic fit Alignment with stated company priorities 15% Risk, inverted Lower delivery, safety and reputational risk scores higher 10% - The shape of the weights is the argument. Value dominates but cannot win alone, because readiness and speed together outweigh it, 0.40 against 0.35: a brilliant thing you cannot yet do is worth less than a good thing you can do this quarter.
- Risk is the only inverted criterion: scored 5 it contributes (6 – 5) x 0.10, scored 1 it contributes (6 – 1) x 0.10. Risk is managed by the gates, not by refusing to score.
Readiness is the criterion people inflate. Take it from the diagnostic, not from the pitch.
9:00 – 16:00
The arithmetic, worked
- The priority score is one line. Value x 0.35, plus Readiness x 0.20, plus Speed x 0.20, plus Fit x 0.15, plus (6 – Risk) x 0.10.
- Take one candidate slowly. Failure prediction on critical rotating equipment: value 5, readiness 3, speed 4, fit 5, risk 2. Value gives 1.75, readiness 0.60, speed 0.80, fit 0.75, and risk of 2 inverts to 4, giving 0.40. The score is 4.30.
Candidate — illustrative Val Rdy Spd Fit Rsk Score Deferment and loss attribution by cause 5 4 4 5 2 4.50 Failure prediction on rotating equipment 5 3 4 5 2 4.30 Tender compliance and bid assembly 4 4 5 3 1 4.15 Field-report automation 3 5 5 3 1 4.00 Autonomous control of plant setpoints 5 1 2 2 5 2.75 Enterprise data lake, built first 3 1 1 3 3 2.20 - Those figures are illustrative. Read the bottom two rows: autonomous setpoint control carries the joint highest value and ranks fifth, because value cannot rescue low readiness and high risk. The data lake built first is the sequence error in numbers.
Do the sums in front of the team. A score computed in private is an opinion with a number attached.
16:00 – 21:00
Where you start, and where you never start
- Now plot value against readiness. Quick wins — high value, high readiness: fund now. Foundations — modest value, but they unlock what sits above: fund to unlock. Big bets — high value, low readiness: stage-gate, never fund in one tranche. Park — do not start here, however good the demonstration.
- THE DOCTRINE
Never more than three initiatives in active pilot at once. Finish them, bank the value, prove the credibility, then take the next three. Spreading thin is the most common way a transformation stalls.
- ENERGY · THE WORKED CASE
- Illustrative: an operator ranked loss attribution, failure prediction and tender compliance first. The digital twin scored high on value and low on readiness, so it became a foundation: reconcile the as-built record and it is fundable next quarter.
OUTSIDE ENERGY · THE PARALLEL- Illustrative: a bank's most exciting candidate, an autonomous pricing engine, had the highest value and the lowest readiness. It was stage-gated as a big bet, and a data contract on the pricing inputs was funded instead.
When the fourth good idea arrives, and it will, the answer is not no. It is next.
21:00 – 25:00
The four gates that govern the money
- Four gates, and money moves only through them. Gate 0 — Deploy: charter a team on a priority problem. Gate 1 — Findings: review the report and score it into the portfolio. Gate 2 — Pilot Go: fund the ninety-day pilot for the top-ranked opportunities. Gate 3 — Scale Go: on a proven, measured pilot, approve scale-up and standardisation.
- THE DOCTRINE
The steering committee owns the gates, and nothing passes Gate 3 without a measured result at Gate 2. Nothing scales on a promise.
- Write down your weights, your six candidates, your provisional first three, and the date of your next Gate 1 review. This session produces Section 17 — your ranked portfolio.
The gates are where an executive's authority lives. Give them away and you no longer control the programme.
Key Terms
Priority score
One number per candidate: Value x 0.35, plus Readiness x 0.20, plus Speed x 0.20, plus Fit x 0.15, plus (6 – Risk) x 0.10, from scores of 1 to 5. Built for comparability, not precision.
Readiness
Whether the organisation can do this now — the data, the owner, the workflow, the skills — taken from the diagnostic rather than from the proposal seeking funding.
The Rule of Three
Never more than three initiatives in active pilot at once. Finish them, bank the value, prove the credibility, then take the next three.
Stage gate
A scheduled point at which money is released or withheld. There are four — Deploy, Findings, Pilot Go, Scale Go — and Scale Go requires a measured result.
Self-Check — Before You Submit
1. A candidate scores value 4, readiness 3, speed 3, strategic fit 4 and risk 2. What is its priority score?
- A. 3.20
- B. 3.60
- C. 3.40
- D. 4.00
B. 3.60 — 1.40 plus 0.60 plus 0.60 plus 0.60 plus 0.40 is 3.60. The trap is the risk term: inverted, a score of 2 contributes (6 – 2) x 0.10, which is 0.40 and not 0.20. Forgetting the inversion flatters the dangerous candidates.
2. Your highest-value candidate scores 1 for readiness: its data has no owner and no agreed definition. A senior sponsor wants it in the first three. What do you do?
- A. Put it in the first three, since value carries the heaviest weight
- B. Park it, because low readiness means it should not be attempted
- C. Fund the foundation that raises its readiness, and stage-gate the candidate as a big bet
- D. Rescore readiness upwards to reflect the sponsor's commitment
C. Fund the foundation that raises its readiness, and stage-gate the candidate as a big bet — High value with low readiness is a big bet: fund the foundation that unlocks it and govern the bet through the gates. Putting it straight into the first three loses a year, and rescoring to please a sponsor destroys the engine's purpose.
3. What does Gate 2 release?
- A. Funding for the ninety-day pilot of a top-ranked opportunity
- B. Approval to scale and standardise a proven pilot
- C. The mandate to deploy a rapid-assessment team
- D. Sign-off of the annual technology budget
A. Funding for the ninety-day pilot of a top-ranked opportunity — Gate 2 is Pilot Go, the money to run the pilot. Scale comes at Gate 3 and only on a measured result. Confusing the two is how an unproven pilot becomes an enterprise rollout.
Today’s Assignment
Score at least six candidates from your own organisation through the prioritisation engine, show the arithmetic, and name the three that will run first.
- List at least six real candidates from your own business. Real means someone has proposed it, or you know the loss it addresses.
- Score each 1 to 5 on value, readiness, speed, strategic fit and risk, with one line justifying the value score and one the readiness score.
- Compute the priority score with the weights exactly as taught, and show the arithmetic in full for at least two candidates.
- Rank the list, then label every candidate a quick win, a foundation, a big bet or park.
- Name the three that will run first under the Rule of Three, then the candidate your organisation is most enthusiastic about that did not make the three, and who you must tell. Close with the date of your next Gate 1 review. Four hundred to nine hundred words.
YOUR CHARTER
Section 17 — Our ranked portfolio and the first three we will run.
How this is marked — the rubric
| Six or more real candidates from the student's own organisation scored on all five criteria, with a justification for each value and readiness score | 30 |
| Priority score computed with the exact weights, including the inverted risk term, and the arithmetic shown for at least two candidates | 30 |
| Ranked portfolio with every candidate labelled quick win, foundation, big bet or park | 20 |
| Names the first three under the Rule of Three, plus the excluded enthusiasm and the person who must be told | 20 |
| Total · pass mark 70 | 100 |
Submitting session 17
Write your answer in Word or Google Docs. The first line must read
Write your answer in Word or Google Docs. The first line must read
Session: 17, or the submission is refused. Upload the Word file (.docx), or switch to the Google Doc tab and paste your document’s share link.Marked against the rubric above and returned as a written letter, saved to your progress page. Do not submit until you have passed the self-check above.Capabilities Covered
Portfolio prioritisationWeighted scoringThe Rule of ThreeStage-gate controlFunding discipline
Execution · Days 61–90SESSION 18
Days 61–90: Pilots That Carry a Number
Why This Matters
Ninety days ends in one of two places. Either you can stand in front of your board and say what moved, by how much, against what it was before — or you can show them a screen. The screen is the more common outcome, and it is not a failure of technology. It is a failure to write the number down before the work started. This session is about the document that prevents it: a pilot charter carrying a baseline, two named owners and a decision date. One page, one afternoon, and the whole difference between a pilot and a demonstration.
25-Minute Module
0:00 – 4:00
Restraint first: three to five, not twenty
- The final thirty days launch pilots, and the discipline is restraint. Three to five, chosen for clear value and adequate data, and never more than three in active pilot at once.
- Adequate data is the filter most often skipped. A candidate with excellent value and no trustworthy data is not a pilot; it is a data contract with a pilot on the end.
- THE DOCTRINE
A pilot without a measured business outcome is not a pilot. It is a demonstration, and demonstrations do not compound.
If you cannot name the number the pilot will move, you are not ready to start it. You are ready to scope it.
4:00 – 12:00
The charter, field by field
- Every pilot carries a charter, agreed before work begins and signed by the sponsor. Nine fields, each closing a specific way that pilots fail.
Field What it must contain How it fails The problem The decision it will improve, in one sentence Written as a technology Business owner Accountable for the outcome, by role and name A committee Data owner Accountable for the data and its definition Assumed to be IT Metric and baseline One metric, its current value, source, period A target with no starting point Timeline Start, 30-day checkpoint, Day-90 decision Open-ended Budget The cost, including internal time Only the licence fee Risk and control Consequence class, approval point, undo Deferred until it works Executive sponsor You, by name, able to clear a blocker A signature with no time behind it - The ninth field is what the other eight serve: the measured outcome at Day 90, against the baseline. Where a proposal cannot fill a field in, that empty field is the finding.
Read the charter aloud in the steering meeting. Any field that cannot be read without hedging is not agreed.
12:00 – 17:00
The baseline is the whole argument
- One field decides whether Day 90 means anything. The baseline is the current value of the metric, the record it came from, and the period it covers, established before anything is built.
- Establishing it is bookkeeping, not analysis. Where the number does not exist, the pilot's first thirty days measure it: legitimate work with an owner and a date, rather than a figure reconstructed after the result.
- THE DOCTRINE
No number without a baseline. Without a defined starting point no improvement can be credibly claimed, and every claim becomes a matter of opinion.
- ENERGY · THE WORKED CASE
- A predictive-reliability pilot on critical rotating equipment. Baseline: unplanned downtime hours per month on the named equipment across four quarters, from the work-order and deferment records.
- The Day-90 metric is downtime avoided against that record, and the owner is the maintenance manager, not the analytics team.
OUTSIDE ENERGY · THE PARALLEL- A discharge-delay pilot in a hospital group. Baseline: bed-days lost to delayed discharge per month, from the admissions system, across the same four quarters.
- One metric, one source, one period, and the owner is the clinical operations lead.
If two functions hand you two different baselines for one metric, you have found the pilot's first deliverable.
17:00 – 21:00
The checkpoint, and what Day 90 means
- At thirty days you are not asking whether it works. You ask whether the data is flowing, the workflow is being used, and the baseline still holds. Most pilots that fail at Day 90 were failing one of those at Day 30.
- THE DOCTRINE
Prove has four parts at Day 90: it ran live, the impact was measured against the baseline, the decisions were owned and actioned, and the value was documented.
Day-90 decision What it means What you do Continue The number moved and the pattern held Run it on and prepare the Gate 3 case Expand The pattern is replicable Standardise it and take it to the next asset Adjust The workflow is right, the instrument is not One bounded revision with a new decision date Stop The number did not move Stop cleanly, write what you learned, release the people - Stopping well is rarer than starting. A pilot killed in public at Day 90, with the reason stated, protects every pilot after it. And before Day 90, write down what would have to be true to run this on the second asset.
Put the Day-90 decision date in your own diary. Pilots survive on the sponsor's calendar.
21:00 – 25:00
Before you close the laptop
- Write down five things: the metric, its current value, where that value came from, the two owners by name, and the Day-90 decision date.
- This session produces Section 18 — your first pilot charter, signed. It is the only section that commits you to a number on a date.
- The assignment is that charter, in full, for your own top-ranked initiative. Pop Quiz 4 opens after this session.
Sign it. An unsigned charter is a proposal, and a proposal is what people negotiate with when the number does not move.
Key Terms
Pilot charter
The one-page document agreed before a pilot begins: problem, business owner, data owner, metric with baseline, timeline, budget, risk review, governance control, sponsor.
Baseline
The current value of the metric a pilot will move, with its source and period, recorded before work starts. Without it, improvement cannot be claimed, only asserted.
30-day checkpoint
The review one month in, testing whether data is flowing, the workflow is used and the baseline still holds — not yet whether the number has moved.
Demonstration
Work that shows a capability without measuring a business outcome. It may be impressive, and it does not compound, because nothing in it can be banked or replicated.
Self-Check — Before You Submit
1. At Day 90 the pilot has run live and adoption is high, but the metric has not moved against the baseline. The team asks for three more months. What is the disciplined response?
- A. Grant the extension, because adoption is a leading indicator
- B. Scale it anyway, since the workflow is demonstrably used
- C. Decide: one bounded adjustment with a new decision date, or stop cleanly and write down what you learned
- D. Replace the metric with the adoption figure and report that
C. Decide: one bounded adjustment with a new decision date, or stop cleanly and write down what you learned — Day 90 exists to force a decision: continue, expand, adjust or stop. An open extension turns a pilot into a project nobody chose, and high adoption of a workflow that moves nothing is an activity metric standing in for an outcome.
2. Which of these makes a piece of work a demonstration rather than a pilot?
- A. It has no measured business outcome against a baseline
- B. It runs on a single asset rather than across the business
- C. It keeps a human in the loop on every consequential decision
- D. It is delivered alongside an external implementation partner
A. It has no measured business outcome against a baseline — The rule is about measurement, not scale, autonomy or sourcing. A one-asset pilot with a human in every loop is a good pilot if it carries a number, and a company-wide rollout with no baseline is still a demonstration.
3. The metric your first pilot will move has never been measured in your organisation. What belongs in the charter?
- A. A target, with the baseline reconstructed at Day 90
- B. The published industry benchmark, used in place of a baseline
- C. A qualitative success statement agreed with the business owner
- D. The first thirty days spent establishing the baseline, named as the pilot's first deliverable with an owner and a date
D. The first thirty days spent establishing the baseline, named as the pilot's first deliverable with an owner and a date — A baseline built after the fact is shaped by the result you hoped for, and a benchmark measures someone else's operation. Measuring the starting point is legitimate pilot work with an owner and a date.
Today’s Assignment
Write the complete pilot charter for the top-ranked initiative in your own organisation, and sign it as the sponsor.
- Take the first-ranked candidate from the portfolio you built for your own business last session. If it has changed, say in one line why.
- Complete all nine fields: the problem as a single decision, both owners, the metric with its baseline, the timeline with checkpoint and Day-90 date, the budget including internal time, the risk review and governance control, the sponsor, and the measured outcome.
- For the baseline, give the figure, the record it comes from and the period it covers. If it does not exist, write in the thirty days that will establish it, with the owner and the reporting date.
- State the guardrails: the consequence class, the human approval point, and the documented undo.
- Write the Day-90 decision in advance — what makes you continue, expand or stop — then sign and date it as sponsor and name the field you found hardest to fill in. Four hundred to nine hundred words.
YOUR CHARTER
Section 18 — Our first pilot charter, signed.
How this is marked — the rubric
| All nine charter fields completed for an initiative in the student's own organisation, with no field left as an intention | 30 |
| Baseline stated as a figure with its named source and period, or a written thirty-day plan to establish it with an owner and a date | 30 |
| Both owners named by role, plus a governance control with a human approval point and a documented undo | 20 |
| Continue, expand and stop conditions for Day 90 written in advance in measurable terms | 20 |
| Total · pass mark 70 | 100 |
Submitting session 18
Write your answer in Word or Google Docs. The first line must read
Write your answer in Word or Google Docs. The first line must read
Session: 18, or the submission is refused. Upload the Word file (.docx), or switch to the Google Doc tab and paste your document’s share link.Marked against the rubric above and returned as a written letter, saved to your progress page. Do not submit until you have passed the self-check above.ASSESSMENT NOW OPEN
Pop Quiz 4 — The Ninety Days opens now. Sit it before you start the next session — it is how you find out whether the doctrine landed, and the practice papers may be resat as often as you like.
Sit Pop Quiz 4 →Closed book, no notes, and Coach Ritchie switches himself off while you sit it.Capabilities Covered
Pilot charteringBaseline disciplineOwner accountabilityCheckpoint governanceStopping cleanly
Execution · CapabilitySESSION 19
Capability, Partners and the Five Failure Modes
Why This Matters
Two capabilities decide whether any of this survives contact with your organisation, and they are the two most often bought from outside: data engineering and AI governance. Rent them both and your operating brain sits in someone else's building, on someone else's renewal terms. This session is about building them, mapping the rest of the capability to roles rather than training everyone identically, and choosing partners on a single test. Then the five ways programmes like yours actually die, so you can name which one is most likely to kill yours before you start.
25-Minute Module
0:00 – 5:00
Capability is an appreciating asset
- Training sits in most budgets as a cost line, which is why it is cut first and measured last. Capability is an asset that appreciates, and every task outsourced that your own people could do is margin lost permanently.
- There is a curve underneath this. Awareness — leaders know what AI can and cannot do. Literacy — staff use the tools safely. Application — AI is in real decisions. Ownership — your own people build and govern the models.
- The scarcest skills here are data engineering and AI governance, and they are the ones most often outsourced. They are unglamorous, hard to interview for, and a vendor will always offer to carry them.
- THE DOCTRINE
Build the data-engineering and the governance track first. Their absence stalls everything else, and no amount of general literacy compensates for data nobody can engineer.
Ask your HR director how many data engineers the company employs, then how many it rents.
5:00 – 11:00
Capability by role, not by headcount
- The point is not to make everyone a data scientist. It is to give every role the right depth: awareness, working, proficient or expert.
Role Depth needed What the person must be able to do Engineers and specialists Deepest process analytics Interrogate a model and trust their judgement against it Front-line supervisors Practical tool fluency Read an alert, check it against what they see, act Maintenance and operations Interpretation and feedback Act on a criticality score and code the work so the model learns Commercial and business development Literacy and prompt fluency Stress-test an analytical answer before a client sees it Compliance and assurance Governance depth Explain how an AI-derived number was produced and defended Leadership and the board Deepest governance literacy Set the risk appetite, fund it, tie it to an outcome - A capability map turns upskill everyone into something costable: for each role, which skills, to what depth, by when. The trap is the certificate, so design the application loop before the curriculum.
Measure training-to-performance conversion, not completion. Completion only tells you that people attended.
11:00 – 16:00
Own the intelligence layer, and test the partner
- In Nigerian energy this has a name, Local Content 2.0, and a general form for any company that does not want its operating brain outside the business: own the intelligence layer rather than renting it.
- The aim is not zero partnership but a phased shift from imported intelligence towards owned capability. Ask of every model: who built it, can we maintain it, and who is accountable when it is wrong.
- THE DOCTRINE
The decisive question for any partner or senior hire is whether they can take you from a slide to a number. Choose for implementation discipline, not demo polish.
- ENERGY · THE WORKED CASE
- Illustrative: an operator scored two bids for a reliability programme. One brought a polished demonstration on the operator's own data.
- The other brought a reference client's baseline, the movement against it, and the manager who owned that number. It was the shorter document and the stronger bid.
OUTSIDE ENERGY · THE PARALLEL- Illustrative: a telecoms operator applied the same test to network-operations analytics.
- What separated the bidders was which of their previous clients could still run it without them. One could name two.
Ask a vendor for a reference client's baseline and the movement against it. The answer, or the silence, is the evaluation.
16:00 – 21:00
The five ways this dies
- Five failure modes kill programmes like the one you have designed. Not one is technical, and not one arrives announced.
Failure mode How it shows up The control that prevents it Starting with technology instead of value A platform chosen before the loss is quantified No candidate enters the portfolio without a value-at-stake figure Waiting for perfect data A data lake built first, ahead of any decision Build the foundation around the priority decisions Insight with no owner Alerts reaching no named person Every insight carries a named owner and a deadline Automating past your ability to govern Autonomy granted faster than controls and undo Autonomy earned rung by rung, matched to consequence Measuring activity instead of outcomes Models deployed, dashboards built, people trained Report outcomes against a baseline - THE DOCTRINE
Name the failure mode most likely to kill your programme before you start, and the one control that prevents it. That paragraph is a pre-mortem, and it is worth more than a risk register.
- In my experience the most common in a capable, well-intentioned business is the third, insight with no owner, because it is the only one that looks like progress all the way to the end.
Write the pre-mortem yourself, in one paragraph, and read it again at Day 60.
21:00 – 25:00
Before you close the laptop
- Write down three things: the two tracks you will build first and their sponsors, the role you will map first, and your pre-mortem in one sentence.
- This session produces Section 19 — your capability plan, partner test and pre-mortem.
- Practice Exam 4 covers the full syllabus and opens after this session. The Final Exam is sat in Session 20.
Capability is the only part of this that cannot be bought at the speed you need it.
Key Terms
Capability map
A statement, role by role, of which skills are needed and to what depth — awareness, working, proficient or expert — with a date and an assessment method.
Data engineering
Building and maintaining the pipelines, definitions, lineage and quality controls models depend on. The most reliable cause of a stalled AI programme when it is absent.
Intelligence layer
The models, pipelines, definitions and analytical systems your decisions depend on. The question is whether your organisation owns and can maintain it, or rents it.
Pre-mortem
A short written judgement, made before work begins, naming the most likely cause of the programme's failure and the single control that prevents it.
Self-Check — Before You Submit
1. Two partners bid for your first pilot. One demonstrates a polished product on your own data within a week. The other is slower, brings a client's baseline and the measured movement against it, and asks who will own the outcome internally. Which is the stronger bid?
- A. The first, because speed of demonstration proves delivery capability
- B. The second, because a measured movement against a baseline is the only evidence they convert slides into numbers
- C. Neither: run the first pilot entirely in-house
- D. Both, split across the pilot so they can be compared
B. The second, because a measured movement against a baseline is the only evidence they convert slides into numbers — The decisive question is whether a partner can take you from a slide to a number, and a reference baseline answers it. A polished demonstration tells you about their sales engineering, and splitting a first pilot means neither partner owns the outcome.
2. Which two capability tracks should be built first, and why?
- A. Prompt fluency and field analytics, because they reach the most people
- B. Executive literacy and change management, because adoption decides everything
- C. Data engineering and AI governance, because their absence stalls everything else
- D. Data science and model development, because models are the product
C. Data engineering and AI governance, because their absence stalls everything else — Those two are the scarcest skills and the most often outsourced: there is little point training analysts on data nobody can engineer or models nobody can govern. Data science is easier to buy in once governance exists.
3. A programme reports twelve models deployed, four dashboards live and three hundred staff trained, and cannot state one outcome moved against a baseline. Which failure mode is this?
- A. Measuring activity instead of outcomes
- B. Waiting for perfect data
- C. Automating past your ability to govern
- D. Starting with technology instead of value
A. Measuring activity instead of outcomes — Deployed, live and trained are activity counts, not business outcomes. It may also have started with technology, but the reportable symptom is the measurement failure, which lets a programme look healthy until the budget round.
Today’s Assignment
Write your own capability plan by role, your partner selection test, and a one-paragraph pre-mortem.
- Map capability for at least five roles in your own organisation: the role, the depth it needs, what the person must be able to do, and how you would assess it.
- Name the two tracks you will build first and the sponsor of each. If they are not data engineering and governance, defend the substitution.
- State your organisation's honest position on the maturity curve — awareness, literacy, application or ownership — with the evidence you would show a sceptic.
- Write your partner test as three questions you will ask every bidder, one of which must take them from a slide to a number, and say what an inadequate answer looks like.
- Write the pre-mortem: which of the five failure modes is most likely to kill your programme, why that one in your organisation, and the control that prevents it. Four hundred to nine hundred words in total.
YOUR CHARTER
Section 19 — Our capability plan, partner test and pre-mortem.
How this is marked — the rubric
| Five or more roles from the student's own organisation mapped with a stated depth and an assessable description of what the person must do | 30 |
| Two first tracks named with a sponsor for each, and the maturity-curve position stated with evidence rather than aspiration | 25 |
| Partner test written as three answerable questions, one requiring a reference baseline and a measured movement | 20 |
| Pre-mortem names one of the five failure modes, argues it from the student's own organisation, and names one preventive control | 25 |
| Total · pass mark 70 | 100 |
Submitting session 19
Write your answer in Word or Google Docs. The first line must read
Write your answer in Word or Google Docs. The first line must read
Session: 19, or the submission is refused. Upload the Word file (.docx), or switch to the Google Doc tab and paste your document’s share link.Marked against the rubric above and returned as a written letter, saved to your progress page. Do not submit until you have passed the self-check above.ASSESSMENT NOW OPEN
Practice Exam 4 — Full Syllabus opens now. Sit it before you start the next session — it is how you find out whether the doctrine landed, and the practice papers may be resat as often as you like.
Sit Practice Exam 4 →Closed book, no notes, and Coach Ritchie switches himself off while you sit it.Capabilities Covered
Capability mappingOwned versus rented capabilityPartner selectionPre-mortem disciplineFailure-mode literacy
The Leadership Mandate · CapstoneSESSION 20
The Leadership Mandate and Your 90-Day Charter
Why This Matters
Nineteen sessions ago I asked how your organisation classifies AI. Today you sign what you have built. This session has three parts: an honest look at the leader this work requires, which is a self-assessment rather than a portrait; the whole doctrine in one pass; and the capstone, which is the charter itself — nineteen sections, one document, one named first pilot, one baseline, one owner, one start date. I have tried to be direct with you rather than encouraging, and I will keep that up for one more session.
25-Minute Module
0:00 – 5:00
What kind of leader this actually requires
- Everything here resolves into a leadership problem. AI in a consequential, regulated business cannot be led by a pure technologist or a pure operator. It needs someone who holds operations, data, risk, technology, finance, people and governance in the same head.
- THE DOCTRINE
The decisive figure is not the most technical person in the room. It is the one who connects field reality, data, governance, commercial return and execution speed into one operating model, and owns the result.
- That profile is scarce because it sits where worlds rarely meet: the technologist has never carried a production number, the operator has never owned a data definition. If you are a functional head rather than a chief executive, the demand is not that you become all of it, but that you know your weakest part and who covers it.
The fastest way to fail at this is to hire the gap before you have named it. Name it first.
5:00 – 11:00
The eight capabilities, as a self-assessment
- Eight capabilities define the leader this work requires. Score yourself 0 to 5 on each and write the evidence. Not your organisation. You.
Capability The question to answer honestly about yourself Strategic clarity Do I know where AI creates value, and where I refuse to chase it? Operational discipline Is every model I fund linked to a decision and an action? Data accountability Can I name the owner of every data domain my decisions need? Governance maturity Could I show the board the model inventory, approvals and undo? People development Am I building the two scarce tracks, or renting them? Commercial focus Is every investment anchored to a baseline? Have I killed one? Execution speed Do I scale by replication, or by demonstration? Owned local capability Do we own the intelligence layer, or rent it? - Two matter most for the executives I meet: data accountability and commercial focus. Neither can be delegated. Your lowest two scores are your development plan.
Score this alone, in ten minutes, and show nobody. The value of this one is in the honesty.
11:00 – 16:00
The whole doctrine in one pass
Week What you built The reflex it leaves you with 1 — The Inflection Six frameworks, from the operating-intelligence loop to the operations-excellence lens Ask what the company must become, not which tool to buy 2 — The Operating Model Visibility, ownership, decision rights, twelve pillars Governance and data first, autonomy last 3 — Trust and the Board Governance standard, traced number, safe use, scoreboard Trust the accountability around the model 4 — The Ninety Days Diagnostic, portfolio, pilot charter, capability plan Connect the capability to the result, or do not build it - THE DOCTRINE
AI will not save your industry. A leader who embraces it will.
- The thesis has not moved since Session 1: AI is not another IT project, it is the operating model of the company. You now hold the document that says so, with your own numbers in it.
If one of those four rows reads thin, that is the week to revisit before the Final Exam.
16:00 – 21:00
Move now
- The technology will keep improving whether or not you act. Regulation will keep becoming measurement-based. The only variable you control is whether your company is built to use any of it.
- THE DOCTRINE
Pick the one loss that costs you most, name the owner, set the baseline, and start the ninety days on Monday. Not a committee. Not a strategy offsite. A start.
- ENERGY · THE WORKED CASE
- For an operator that usually means the deferment that recurs every month under a code nobody owns. Name the loss, name the reliability manager, take the baseline from the allocation record, and start.
OUTSIDE ENERGY · THE PARALLEL- For a contractor it is usually rework, or the schedule slip found at the monthly report. Same moves: name the loss, name the project-controls manager, take the baseline from the last four projects, and start.
- A committee formed to decide where to begin is a decision to begin later. Put your first Gate 0 charter in the diary for next week.
Twelve weeks from now you will hold either a measured number or a longer plan.
21:00 – 25:00
The capstone, and what happens when you finish
- The capstone is the charter itself: nineteen sections assembled into one signed document, with a named first pilot, a baseline, an owner, a start date and a one-page executive summary.
- Two things close the programme, in order. First the capstone, submitted as Session 20. Then the Final Exam, sat on this page: fifteen questions, three of them rebuilt from your own earlier mistakes, thirty minutes, closed book. The certificate needs all twenty assignments, the capstone, and the final at 70 or above.
- That summary is the only part some board members read. Seven lines: the decision requested, the value at stake, the first pilot, its baseline, the owner, the start date, the review date.
- When you submit the capstone, your completion gift is released: The Intelligent Barrel — Digital Edition plus its three executive bonuses, $39.99 — The Reader's Toolkit, The 90-Day AI Pilot Charter and Playbook, and The Value–Risk Map and Autonomy Playbook. It is given free, as thanks for the time you put in, and arrives by email as a code at the address you signed in with. Your certificate and reference letter follow.
- This programme was free, and it was written for people who intend to run the ninety days rather than read about them. The work is hard. It is also worth it. Build it well, and build it yours.
Send me the charter even if two sections embarrass you. I would rather mark an honest document than a tidy one.
Key Terms
90-Day AI Operating Model Charter
The signed document assembled from all nineteen sections, from the honest classification and the broken loop through governance and the scoreboard to the portfolio, the first pilot and the capability plan.
Executive summary
The one page a board decides from: the decision requested, the value at stake, the first pilot and its baseline, the owner, the start date and the review date.
Strategic clarity
The first of the eight leadership capabilities: knowing where AI creates value in your business and refusing to chase it where it does not.
Execution speed
Moving from pilot to scale by disciplined replication rather than a parade of demonstrations. Measured in value banked, not initiatives launched.
Self-Check — Before You Submit
1. Your charter is complete except for one thing: the baseline for the first pilot has never been measured, and your board meets in a fortnight. What goes in the document?
- A. The target, with the baseline reconstructed once results arrive
- B. A comparable figure from a peer company, clearly attributed
- C. The first thirty days named as baseline establishment, with the owner, the source and the reporting date
- D. A qualitative success statement, so the pilot is not delayed
C. The first thirty days named as baseline establishment, with the owner, the source and the reporting date — Measuring the starting point is legitimate, chartered work, and naming it as a deliverable keeps the board honest. A baseline reconstructed afterwards is shaped by the result, and a peer's figure measures someone else's operation.
2. What does the doctrine mean when it says the decisive figure is not the most technical person in the room?
- A. Technical depth is unnecessary at executive level
- B. AI leadership means connecting operations, data, risk, technology, finance, people and governance into one model and owning the result
- C. Technical work should be outsourced to specialists
- D. The board should govern AI without executive involvement
B. AI leadership means connecting operations, data, risk, technology, finance, people and governance into one model and owning the result — The claim is about integration and accountability, not about technical depth being useless: data accountability and governance maturity demand real understanding. What is rejected is the idea that the deepest technologist automatically leads an operating-model change.
3. Which of these belongs on the one-page executive summary of your charter?
- A. The model architecture and the vendor shortlist
- B. The twelve-pillar scores for every function
- C. Every candidate in the portfolio, with the arithmetic
- D. The decision requested, the value at stake, the first pilot with its baseline and owner, the start date and the review date
D. The decision requested, the value at stake, the first pilot with its baseline and owner, the start date and the review date — A board approves a decision, a number and an accountability, and needs the review date to hold you to all three. Architecture and scorecards belong in the body, where they cannot bury the decision.
Today’s Assignment
Assemble the nineteen charter sections you have written into one signed 90-Day AI Operating Model Charter for your own organisation, with a one-page executive summary.
- Assemble all nineteen sections in order and edit them for consistency. Where a later session changed your mind, correct the earlier section and say so in a line.
- Name the first pilot. It should be the charter you signed in Session 18 unless something has genuinely changed, in which case argue the change.
- State the baseline as a figure with its source and period; both owners by role and name; the sponsor; and the start date as an actual date, not a quarter.
- Write the one-page executive summary: the decision requested, the value at stake, the first pilot and its baseline, the owner, the start date, the governance control and the review date.
- Read the document as your most sceptical board member and fix whatever would not survive the question “how do you know?”. Then sign and date it, and name the one commitment you are least confident your organisation will keep.
YOUR CHARTER
The whole charter, assembled and signed.
How this is marked — the rubric
| All nineteen sections present, in order, internally consistent, with no placeholders and no section unresolved | 30 |
| A real baseline for the first pilot — a figure with its named source and period — rather than a target, a benchmark or an intention | 25 |
| A named owner by role and name for every commitment, plus the sponsor's signature and an actual start date | 25 |
| A defensible first pilot with its value at stake quantified, and a one-page executive summary carrying the decision requested, the owner, the start date and the review date | 20 |
| Total · pass mark 70 | 100 |
Submitting session 20
Write your answer in Word or Google Docs. The first line must read
Write your answer in Word or Google Docs. The first line must read
Session: 20, or the submission is refused. Upload the Word file (.docx), or switch to the Google Doc tab and paste your document’s share link.Marked against the rubric above and returned as a written letter, saved to your progress page. Do not submit until you have passed the self-check above.FINAL EXAM
Your Final Exam is sat in this session. 15 questions, of which 3 are rebuilt from questions you got wrong on the practice papers. Sit it once your capstone is submitted.
Sit the Final Exam →Closed book, no notes, and Coach Ritchie switches himself off while you sit it.Capabilities Covered
Integrative AI leadershipHonest self-assessmentCharter assemblyBoard-ready summaryExecution start discipline