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 2 OF 4 · THE OPERATING MODEL · SESSIONS 6–10 · WORK AT YOUR OWN PACE
The Doctrine · Framework 5SESSION 6
The Sequencing Discipline: The Order That Survives
Why This Matters
You hold four frameworks. Today you get the one that decides whether any of them survive a budget round. I have watched capable companies, with real money and genuine intent, produce nothing after eighteen months — not because their models were poor, but because they did the right things in the wrong order. Sequence is the most reliable predictor I know of whether an AI programme is still alive in two years. It is also the cheapest thing to fix, because it costs nothing but the discipline to say not yet to something your organisation is already excited about.
25-Minute Module
0:00 – 6:00
The order that survives
- Framework 5 is the least glamorous of the six and the most predictive. The question is not which capability you build. It is which one you build first.
- The sequence that works: build the data foundation around the decisions that matter most; prove value in a bounded pilot against a measured problem; keep a human in the loop where stakes are high; scale only what has paid.
- THE DOCTRINE
Sequence is not bureaucracy. It is what separates a programme that compounds from one that is quietly defunded.
Move The test before you move on Foundation Can the decision-maker defend these numbers to a sceptic Bounded pilot Did a named owner act, and did the baseline move Human in the loop Is it clear who approves, and what the undo is Scale Has this version already paid, in your own units
You move on when the test is passed, not when the calendar moves.
6:00 – 13:00
The order that fails, and the three sentences that reveal it
- The failing sequence is the exact inverse, and it is the default: buy a platform, hire data scientists, build a proof of concept, and find eighteen months later that nothing in the operation has changed.
- Capable people choose it because each step is easier to approve than the one it replaces. A platform is a purchase order; a data foundation is an argument about who owns a number.
- Three sentences tell me a programme is out of order. One: “we will start with the platform and find the use cases later” — technology before value. Two: “we are building the team first” — capability before a named decision. Three: “the proof of concept will show what is possible” — nothing provable, because there is no baseline and no owner.
Listen for what a proposal avoids. You need not be technical to hear it.
13:00 – 19:00
Data and governance first, autonomy last
- Data first, because of Framework 2. A capable model on an ungoverned foundation does not buy intelligence. It industrialises whatever is broken, confidently.
- Governance next, because governance is what makes scaling defensible. Automation on an ungoverned foundation does not accelerate the company. It accelerates its risk.
- THE DOCTRINE
Governance and data first, autonomy last. Skipping that order is the most expensive failure available to a serious company.
- Sequencing is not waiting. Build a foundation trustworthy enough around the decisions that matter most, then expand — never the data lake first, in hope of a use case.
If someone says the sequence will slow you down, ask how fast the last out-of-order programme delivered.
19:00 – 25:00
Re-sequencing, and your wrap
- Same capability, opposite order. The only difference is what the sponsor insisted on before the build.
- ENERGY · THE WORKED CASE
- An operator buys a predictive-maintenance platform for its compressors. Work orders close as “general fault”, so there is no failure history to learn from, and alerts reach no named owner.
- Re-ordered: fix failure coding on the ten most critical machines, baseline last year's unplanned downtime, then one bounded pilot routing predictions to a named planner.
OUTSIDE ENERGY · THE PARALLEL- A food manufacturer funds a demand-forecasting model before anyone owns the product master data. Three functions count an active product line differently, so the forecast ranks lines on a definition the planners do not use.
- Re-ordered: one data contract for the product domain, a named owner, one definition, a measured baseline for forecast error at one plant, then the model.
- Notice what the sponsor does in both. Not cancel. Re-sequence. The foundation work gets funded because it now sits on the critical path to something the business wants.
- THE DOCTRINE
Decide what paid means before the pilot starts, never after it finishes.
- Write three lines before you close: the initiative furthest out of order, the move that should have come first, and the item you will put in the not yet column. That is Section 6.
Anyone can add to a roadmap. A leader can defend an omission.
Key Terms
Sequencing discipline
Building AI capability in a fixed order — foundation, bounded pilot, governed autonomy, scale — entering each stage only when the last passed its test.
Bounded pilot
A narrow test against one real, measured problem, with a baseline, an owner and an end date. It exists to produce a number.
Proof of concept
A demonstration that something can technically work. It proves feasibility, never value, and cannot carry a business case.
Platform-first procurement
Buying technology before naming the decision it must improve: the opening move of the failing sequence.
Self-Check — Before You Submit
1. Which order is the sequence that survives?
- A. Hire the data team, buy the platform, run a proof of concept, then find the use cases
- B. Data foundation around priority decisions, bounded pilot with a baseline, human in the loop where stakes are high, scale only what has paid
- C. Build the data lake, select tools, automate, then measure
- D. Automate the low-risk work first, then govern whatever breaks
B. Data foundation around priority decisions, bounded pilot with a baseline, human in the loop where stakes are high, scale only what has paid — Foundation, proof, control, scale. C sounds foundational but spends years without touching a decision; D inverts governance and autonomy.
2. A functional head brings you a credible vendor proposal. The business case says the proof of concept will demonstrate the value. There is no named data owner and no baseline. You are the sponsor. What do you do?
- A. Approve it — the cost is small and the learning useful
- B. Reject it until the organisation is more mature
- C. Approve in principle but re-order it: name the decision, appoint the data owner, set the baseline, then fund a bounded pilot that must move it
- D. Send it to IT for a feasibility review
C. Approve in principle but re-order it: name the decision, appoint the data owner, set the baseline, then fund a bounded pilot that must move it — Re-sequencing fixes the fatal defect while keeping both initiative and relationship. Approving as written buys a demonstration you cannot defend; rejecting outright stops good people bringing you ideas.
3. Why does governed autonomy come last rather than first?
- A. Autonomous systems cost more than analytical ones
- B. Regulators prohibit autonomous decisions in most sectors
- C. Vendors cannot yet deliver it reliably
- D. Automation on an ungoverned foundation accelerates risk rather than performance
D. Automation on an ungoverned foundation accelerates risk rather than performance — The reason is structural: autonomy multiplies whatever sits beneath it. Blaming cost, regulation or vendors excuses the foundation work you control.
Today’s Assignment
Audit one AI initiative already under way or proposed in your own organisation against the four-move sequence, and write the re-ordering you would impose as its sponsor.
- Name the initiative in one sentence of plain business language and say who is driving it. Choose the one carrying the most money or political weight.
- For each of the four moves, state done, partly done or skipped, with evidence: the data owner by name and role, the baseline figure and its source, the approval point, the scale decision.
- Quote, as closely as you honestly can, a sentence you have actually heard about this initiative that resembles one of the three tells.
- Write the re-ordering: first, second, and the not yet item, with a named owner and a date against the first two.
- The honest part: name what your organisation is most excited about that your own sequence says is not yet, and the sentence you would use to tell its champion.
- Aim for 400 to 900 words. This becomes Section 6 of your charter.
YOUR CHARTER
Section 6 — Our sequence: what we do first, second and not yet.
How this is marked — the rubric
| Audits a real named initiative from the student's own organisation against all four moves, with evidence stated or its absence declared | 35 |
| Names a data owner by role and a baseline figure with its source, or states that neither exists today | 25 |
| Re-ordering gives a first step, a second step and a not-yet item, with a named owner and date against the first two | 25 |
| States the not-yet decision and the sentence that would deliver it to its champion | 15 |
| Total · pass mark 70 | 100 |
Submitting session 6
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: 6, 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
Sequencing disciplineProposal interrogationBaseline disciplineDeliberate omissionSponsor authority
The Doctrine · Framework 6SESSION 7
The Operations-Excellence Lens: Connect It to the Result or Do Not Build It
Why This Matters
This session closes the toolkit, and it is the one your finance director will thank you for. Every AI conversation reaches a moment where sophistication is offered as evidence — the architecture, the accuracy, the number of models in production. None of that is value. The test of any model, platform or pilot is its effect on something your business already measures and already cares about. Most executives accept a technical proposal without ever forcing that translation, and cannot answer the board a year later. Today you learn the translation, and the baseline discipline that makes it honest.
25-Minute Module
0:00 – 6:00
The only test that matters
- Framework 6 is a lens, not a method. Hold every proposal against what the business already measures: output protected, downtime avoided, decisions accelerated, compliance strengthened, cost removed, revenue assured. If it lands on none of those, it is not an investment.
- In the book the instruction is written for the industry it came from — connect the capability to the barrel, or do not build it. Read it in your own unit; it is the same sentence.
- THE DOCTRINE
Connect the capability to the result, or do not build it.
- This is not anti-technology. A valuable model presented without a result is indistinguishable, to a board, from a worthless one.
Resist evaluating the technology. Ask which existing number it moves, then wait.
6:00 – 12:00
Activity or outcome
- Activity metrics count what the programme did. Outcome metrics record what the business gained, against a baseline. Telling them apart is most of the executive work.
- Activity metrics are not dishonest. They are the wrong evidence, and the standard disguise of a failing programme — always available, always rising, never falsifiable.
What the pack reports What you ask for instead Eleven models in production Which decision each changed, by how much against baseline Fourteen dashboards built Decision latency reduced on the decisions they serve Four hundred people trained How many acted on it in a real decision last month Licences at three sites Cost removed or revenue assured there, against last year - I have sat in a review where every activity metric was moving and no outcome metric had been defined. Nobody was lying. The programme was defunded within two quarters.
Deployed and trained is activity. Reduced and protected is outcome.
12:00 – 18:00
The baseline discipline
- A baseline is the defined starting point a claim is measured from: the figure, the period, the source, and the person who stands behind it. Without one, every claim is an opinion.
- THE DOCTRINE
No number without a baseline. Without a defined starting point, no improvement can be credibly claimed.
- Usually the baseline does not exist. That is not permission to proceed. It makes establishing the baseline the first deliverable — a baseline built after the result is a justification, not a measurement.
Ask for the baseline in the first meeting. If it cannot be produced in a fortnight, that is the finding.
18:00 – 25:00
Your unit of value, and six frameworks as one reflex
- Every business counts in its own unit — the barrel, the case, the claim, the shipment, the customer, the bed-day, the naira. The drill: which existing measure moves, by how much, from what starting point, and whose name is on it.
- ENERGY · THE WORKED CASE
- The proposal: an anomaly-detection platform for well performance, presented with model accuracy and a screen of charts.
- The translation: barrels of deferment avoided per month on one field, against the last four quarters of the deferment ledger, each intervention owned by the production superintendent.
OUTSIDE ENERGY · THE PARALLEL- The proposal: route optimisation for a logistics operator, presented as a percentage gain in a simulation.
- The translation: on-time delivery and cost per shipment on two named lanes, against those lanes last year, with the regional operations manager accountable.
Framework The question it forces Operating-intelligence loop Where does our loop break — sense, decide, act, learn Data-foundation precondition Would this be confidently wrong on the data we have Autonomy ladder What does this touch, and what rung has it earned Value–risk map What does being wrong cost, and is the value worth it Sequencing discipline Is this in the order that survives, or the inverse Operations-excellence lens Which number we already report does this move - Write your unit of value in one word, the number you would stake this programme on, and whether a defensible baseline exists today. That is Section 7.
Six questions, one page. Ask all six and you will be wrong far less often.
Key Terms
Activity metric
A count of what a programme did — models deployed, dashboards built, people trained. Useful for managing work, never evidence of value.
Outcome metric
A measure the business already reports, moved against a baseline: downtime avoided, cost removed, revenue assured, decisions accelerated, compliance strengthened.
Baseline
The starting point a claim is measured from: figure, period, source system, and the person accountable. Set before the build, or it is a justification.
Unit of value
The unit your organisation already counts performance in — the barrel, the case, the claim, the shipment, the customer, the naira.
Self-Check — Before You Submit
1. Your finance director asks what the AI programme delivered last year. The answer: eleven models in production, fourteen dashboards live, four hundred staff trained. What is the correct response?
- A. Accept it — adoption and coverage are the leading indicators in year one
- B. Ask which existing reported number each model moved, from what baseline, and who owns that figure
- C. Commission a technical review of the eleven models
- D. Ask for the same metrics next quarter to establish a trend
B. Ask which existing reported number each model moved, from what baseline, and who owns that figure — Every figure quoted is activity, and rising activity can coexist with zero value for years.
2. A proposal will improve a process, but no baseline for it exists anywhere in the company. What is the disciplined way forward?
- A. Proceed and reconstruct the baseline afterwards from whatever records exist
- B. Use an industry benchmark instead of your own starting point
- C. Make establishing the baseline the first deliverable, before anything is built
- D. Defer until the finance system reports the measure automatically
C. Make establishing the baseline the first deliverable, before anything is built — A baseline built after the result is a justification, and a borrowed benchmark says nothing about your own starting point.
3. Which of these is an outcome metric rather than an activity metric?
- A. Reduction in unplanned downtime at one site, against last year's recorded hours
- B. Use cases approved by the governance committee
- C. Staff who completed the AI awareness training
- D. Data domains with a documented owner
A. Reduction in unplanned downtime at one site, against last year's recorded hours — Only the first states a business result against a defined starting point. The others are programme management — two of them necessary foundations, but not value.
Today’s Assignment
Convert one real technical AI proposal from your own organisation into your own unit of value, with the baseline you would measure it against.
- Summarise the proposal in no more than five sentences, as it was given to you, including the technical claims it leads with and what it would cost.
- State your unit of value in one word, then translate: which existing measure moves, by how much you would accept as success, over what period, and whose name is on the result.
- Define the baseline — figure, period, source system, definition, owner. If it does not exist, give the plan to establish it before any build: who, from what source, by what date.
- List every activity metric the proposal offers as evidence, and replace each with the outcome metric you would accept instead.
- The honest part: name one thing your organisation funded in the last two years whose value it cannot now demonstrate, and what was missing.
- Aim for 400 to 900 words. This becomes Section 7 of your charter.
YOUR CHARTER
Section 7 — Our unit of value and the baseline we will measure against.
How this is marked — the rubric
| States one unit of value and translates the proposal into a specific existing measure, with a success threshold and an owner named by role | 30 |
| Baseline fully specified — figure, period, source, definition, owner — or a dated plan to establish it before build | 30 |
| Every activity metric in the proposal is identified and replaced with an outcome metric | 20 |
| Names one real past investment whose value cannot be demonstrated, and what was missing | 20 |
| Total · pass mark 70 | 100 |
Submitting session 7
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: 7, 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
Value translationBaseline disciplineMetric discriminationFramework fluencyBoard-ready evidence
The Operating Model · VisibilitySESSION 8
From Reporting to Visibility: Running the Business on Live Signals
Why This Matters
Ask most executives whether they have visibility of their operation and they will point to a reporting pack. I have done it myself, years ago, and I was wrong. A report tells you what happened; visibility tells you what is happening, what is changing and what needs you now. Conflating the two is the quiet reason interventions arrive late — by the time some reports land, the window to act has closed and the loss is already in the numbers. This session is the shift from a monthly story to live signals surfaced by exception.
25-Minute Module
0:00 – 6:00
History, or now
- Reporting and visibility are different instruments answering different questions. The difference decides whether your intervention lands inside the window where it still changes the outcome.
The question A monthly report answers An operating picture answers What is our position What it was at month end What it is now, with each signal's vintage What is wrong What went wrong, after the fact What is outside its threshold, and who was told Who is acting Who will be asked at the next review Who owns the open exception, and by when - THE DOCTRINE
Reporting tells you what happened. Visibility shows you what is happening, what is changing, and what needs attention now.
- Two shifts are needed and only one costs money. The technical shift is from batch reporting to near-real-time exception management. The management shift is to run the meeting from live signals.
You can tell which instrument a company runs on by what leaders bring to a Monday meeting.
6:00 – 12:00
What real visibility covers
- Real visibility is broader than the operational dashboard most companies stop at. In the doctrine it spans projects, output, equipment health, maintenance backlog, procurement delays, safety and quality observations, compliance deadlines, the sales pipeline, cash exposure and risk escalation.
- In your business that reads as the order book, the plant, the fleet, the arrears, the claims queue, the ward, the branch network, the regulatory calendar.
- Six to ten signals is the right size, each with a stated vintage. Nobody reads the thirty-signal version, and data three weeks late supports archaeology, not operations.
Choose signals by the decisions you make, not by the data you happen to have.
12:00 – 19:00
The anatomy of an exception
- An exception is not an alert. An alert is a notification; an exception is owned work. Four things make one: a threshold, a named owner, a deadline, and an escalation when it passes.
- THE DOCTRINE
An alert that reaches no named owner is not intelligence. It is noise with a timestamp.
- ENERGY · THE WORKED CASE
- A flow station's output drifts below its expected envelope on a Tuesday. Under monthly reporting it surfaces in the deferment line at month end, three weeks after the barrels were lost.
- Under exception management it breaches a threshold within the shift, routes to the named production supervisor with a deadline, and escalates to the asset manager if still open at close of day.
OUTSIDE ENERGY · THE PARALLEL- A hospital reviews theatre utilisation and cancellations monthly. They are explained accurately, month after month, and the pattern persists.
- Under exception management a list at risk breaches a threshold the afternoon before, routes to the named theatre coordinator, and escalates to the clinical director by handover.
- Watch for detection without closure. High detection with low closure means visibility was bought without accountability.
Accept the system's default thresholds and you have delegated what counts as abnormal in your business.
19:00 – 25:00
Decision latency, the rhythm, and the honest test
- One number sits underneath all of this: decision latency, the elapsed time from a validated signal to an action taken. Measure it for three decisions you make repeatedly and review the trend monthly.
Cadence Runs on The question Daily Exceptions What changed, what is overdue, what escalates before close of business Weekly Trends and accountability Which risks recur, which interventions closed, which owners carry open items Monthly Performance and learning How are outcome metrics trending against baseline - THE DOCTRINE
If your management rhythm depends on a month-end report to know what is wrong, you do not have visibility. You have history.
- Apply that test now. When did you last learn something important from a live signal rather than a report or a phone call? If you cannot remember, that is your finding.
- Write the six to ten signals you would run your own weekly meeting from, and the owner of each one's exceptions. That is Section 8.
- Pop Quiz 2 opens after this session: five questions on Sessions 6 to 8 — sequence, the operations-excellence lens and visibility. Ten minutes, closed book, and sit it before Session 9.
The test of an operating picture is whether it ever changed a decision inside the same week.
Key Terms
Operational visibility
The current, trusted state of the operation across its critical domains, surfaced by exception. Reporting records what already happened.
Exception management
Comparing live signals against defined thresholds and pushing each breach to a named owner with a response deadline and an escalation path.
Decision latency
The elapsed time from a validated signal to an action taken: the master variable of an intelligent operating model.
Exception closure rate
The share of raised exceptions resolved on time by their named owner. High detection with low closure means accountability is missing.
Self-Check — Before You Submit
1. Your monthly pack has shown, three months running, that procurement delay was the largest cause of lost output. Each month it is explained and nothing changes. What is the first structural fix?
- A. Commission a deeper analysis of procurement performance
- B. Escalate the issue so procurement receives executive attention
- C. Define a delay threshold for production-critical items, name the owner, set a response deadline and escalation, and review closure weekly
- D. Procure a dashboard tracking lead times in real time
C. Define a delay threshold for production-critical items, name the owner, set a response deadline and escalation, and review closure weekly — The delay is not unknown; it has been reported three times. What is missing is a threshold, an owner and a deadline.
2. Which statement best distinguishes visibility from reporting?
- A. Visibility shows what is happening and what needs attention now; reporting records what already happened
- B. Visibility uses live data feeds; reporting uses manual data entry
- C. Visibility is for operational managers; reporting is for the board
- D. Visibility is produced by AI systems; reporting is produced by people
A. Visibility shows what is happening and what needs attention now; reporting records what already happened — The distinction is the decision window, not the technology or the audience. A manual signal reviewed daily against a threshold is visibility; an automated monthly pack is history.
3. An exception system raises many exceptions each week and very few are closed on time. What does this most likely indicate?
- A. Thresholds are too tight and should be widened until volume falls
- B. The organisation has visibility without accountability — ownership and deadlines are not enforced
- C. The underlying data is too poor for exception management
- D. More automation is needed so exceptions close themselves
B. The organisation has visibility without accountability — ownership and deadlines are not enforced — Detection without closure is the classic failure: signals arrive but no owner is held to a deadline.
Today’s Assignment
Design the one-page operating picture you would run your own weekly leadership meeting from, with the exception thresholds and the owner of each signal.
- List the six to ten signals. For each: the signal in plain language, its source record, how current it would be when you read it, and why it earns a place on a page of ten. Signals you cannot source stay on the list, marked as gaps.
- Set each exception threshold — the value or condition at which it should interrupt someone — and say who set it and on what basis.
- Give each signal an owner by role, a response deadline, and who it escalates to when the deadline passes. No signal goes on the page without all three.
- Nominate three recurring decisions, estimate today's decision latency for each in hours or days, and say how you would measure it properly from next month.
- The honest part: state what your leadership team learns only from the month-end pack, and what that lateness cost last year, labelling estimates as illustrative.
- Aim for 400 to 900 words plus the table of signals. This becomes Section 8 of your charter.
YOUR CHARTER
Section 8 — Our operating picture and the exceptions that will interrupt us.
How this is marked — the rubric
| Six to ten signals from the student's own operation, each with a stated source and currency, with unsourceable signals marked as gaps | 30 |
| Every signal carries a specific threshold, an owner named by role, a response deadline and an escalation path | 30 |
| Three named recurring decisions with an estimated current decision latency and a method for measuring it properly | 20 |
| States what is learned only at month end and what that lateness cost, with estimates labelled illustrative | 20 |
| Total · pass mark 70 | 100 |
Submitting session 8
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: 8, 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 2 — Sequence and Visibility 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 2 →Closed book, no notes, and Coach Ritchie switches himself off while you sit it.Capabilities Covered
Exception designDecision-latency measurementOperating rhythmSignal selectionAccountability enforcement
The Operating Model · OwnershipSESSION 9
Who Owns AI: The Programme Office, Decision Rights and the Business
Why This Matters
Of everything in the operating model, one decision shapes the rest: who owns AI. Get it wrong and you will still be reviewing pilots in two years with nothing banked. I have watched companies with better technology than their competitors lose, purely because the programme sat in the wrong place on the organisation chart. The structure is not complicated — a sponsor with real authority, a small empowered programme office, a governance committee, named data owners — but each part fails in its own way without the others. This is the least glamorous section of your charter and the one most likely to save it.
25-Minute Module
0:00 – 6:00
Where the programme sits decides what it produces
- AI cannot belong to the technology function alone. Not because IT is weak — it is often the most competent function in the building — but because only the business can answer for the outcome.
- THE DOCTRINE
An AI programme owned by IT produces tools. An AI programme owned by the business produces performance.
- So the test of your design is one question. When a system produces a recommendation nobody acts on, whose performance review does that appear in? If it is the head of technology, the design is wrong.
- The opposite error is to declare that the business owns it and give the business no structure, no capability and no forum. That produces enthusiasm, a scatter of experiments, and no compounding.
Ownership is not a sentence in a strategy document. It is a name, a budget and a decision point.
6:00 – 13:00
The four parts, and how each fails alone
- An executive sponsor with real authority: budget, access, and the standing to remove a blocker in a week rather than a quarter. Sponsorship is leadership of the redesign, not endorsement.
- A small but empowered AI programme office connecting technology, operations, finance, commercial, quality or safety, compliance, human resources and procurement. It connects and standardises rather than building.
- An AI governance committee deciding what is allowed, at what consequence, on whose signature. Named data owners, one per critical domain, accountable for definition, quality and access.
- A committee without a sponsor stalls: nothing it decides can be funded. A sponsor without a programme office cannot execute: nobody runs the work between meetings. Data owners without a committee are names in a policy.
- ENERGY · THE WORKED CASE
- An operator's predictive-maintenance alerts arrived in an inbox watched by the digital team, which could see equipment degrading but could not authorise a shutdown, a spare or a crew.
- The fix was structural: the operations director sponsored it, each prediction routed to the maintenance planner as named owner, the reliability engineer owned the equipment data.
OUTSIDE ENERGY · THE PARALLEL- A manufacturer's quality model, owned by IT, identified defect patterns accurately for nine months. No supervisor changed a parameter, because nothing in their objectives referred to it.
- The fix: the plant director sponsored it, line supervisors owned the corrective actions, the quality manager owned the defect register, and value was measured in scrap rate.
Keep the office small enough to know every name. A department defends itself instead of connecting others.
13:00 – 19:00
Decision rights, written down
- Most disputes in an AI programme are unwritten decision rights surfacing under pressure. Four rights cover nearly all of it.
The right The question it settles Where it belongs Who proposes Who puts a candidate forward, in what form Any function head, via the programme office Who funds Who commits money, at which threshold The sponsor, within a delegated limit Who approves autonomy What a system may touch and act on The governance committee, by consequence Who can stop it Who switches a live system off without a meeting A named owner and the risk function - THE DOCTRINE
An unwritten decision right is discovered during an incident, which is the worst possible moment to negotiate one.
- The fourth is most often missing and is the one a board will ask about. Someone must be able to stop a live system today, alone, and everyone must know who.
Test each right against a real past dispute. If it would not have settled it, rewrite the line.
19:00 – 25:00
How the work gets done, and your wrap
- You cannot get into the detail yourself. The mechanism that reaches the truth fast is a rapid-assessment team: small, cross-functional, time-boxed to seven to fourteen working days, on one problem. It assesses and recommends; it does not build.
- Charter it before it deploys: sponsor, members, the single question, scope in and out, timebox, data needed, and your decision date.
- Require the same one-page report every time — readiness score, top three findings with evidence, value at stake, the recommended bounded pilot, risks, the decision ask, and Go, Hold or Stop. Identical structure lets you compare six opportunities in an afternoon.
The executive — you The team Set direction and priorities Get into the operational detail Sponsor the team, remove blockers Score the diagnostic honestly Provide access, data, resources Quantify the value at stake Decide: fund, hold or stop Design the pilot, report in standard format - THE DOCTRINE
You own the what and the whether. The team owns the how and the how-much. A team that cannot fill in the value-at-stake line has not finished.
- Then hold the rhythm: weekly to unblock, monthly to fund, quarterly to scale or stop. Write four real names — sponsor, office lead, committee chair, one data owner — and the four rights in a line each. That is Section 9.
Your two jobs here are air cover and the decision date. Miss the second and honest findings stop coming.
Key Terms
Executive sponsor
The senior leader who owns the outcome, holds the budget and the authority to remove blockers, and decides fund, hold or stop.
AI programme office
A small, empowered team connecting technology, operations, finance, commercial, quality, compliance, human resources and procurement under the sponsor.
Decision rights
The written record of who proposes, who funds, who approves autonomy and who can stop a live system. One line each, by named role.
Rapid-assessment team
A cross-functional team, time-boxed to seven to fourteen working days on one problem, reporting one page ending in Go, Hold or Stop.
Self-Check — Before You Submit
1. A governance committee has met monthly for nine months, approved several proposals in principle and funded nothing. No pilot has started. What is the most likely structural cause?
- A. Committee membership is too narrow
- B. There is no executive sponsor with budget authority and no programme office running the work between meetings
- C. The proposals are not technically mature enough
- D. The organisation needs a chief digital officer
B. There is no executive sponsor with budget authority and no programme office running the work between meetings — A committee can decide but not commit: without a sponsor holding budget, approval never becomes money.
2. In the division of labour between the executive and a rapid-assessment team, which pairing is correct?
- A. The executive scores the diagnostic; the team sets priorities
- B. The executive designs the pilot; the team funds it
- C. The executive owns the what and the whether; the team owns the how and the how-much
- D. The executive gathers evidence; the team removes blockers
C. The executive owns the what and the whether; the team owns the how and the how-much — The executive sets direction and decides fund, hold or stop; the team quantifies value and designs the pilot. Other pairings swap the two jobs.
3. Which decision right is most often left unwritten, and is the one a board will eventually ask about?
- A. Who can stop a live system, alone, without convening a meeting
- B. Who may propose a candidate use case
- C. Who funds a pilot within a delegated limit
- D. Who chairs the governance committee
A. Who can stop a live system, alone, without convening a meeting — The other three follow the organisation chart and settle early. The authority to switch something off is written down only after an incident.
Today’s Assignment
Draw your own organisation's AI ownership map — sponsor, programme office, governance committee and data owners — and write the four decision rights in one line each.
- Name your executive sponsor: the person, the role, and the authority they hold today over budget and blockers. If it is you, say what you will stop doing to make room.
- List the programme office by function, one named person each, marking functions you cannot staff as gaps. Say who leads it and how much of their time this is.
- Name the governance committee by role, and one data owner for each critical domain your priority decision depends on.
- Write the four rights in one line each: who proposes, who funds and to what limit, who approves autonomy by consequence class, and who can stop a live system alone.
- Test them against one real dispute or near-miss you have had. Say whether your written version would have settled it, and rewrite the line if not.
- The honest part: name the function most likely to resist this, the thing they will be right about, and your answer to it.
- Aim for 400 to 900 words. This becomes Section 9 of your charter.
YOUR CHARTER
Section 9 — Who owns this: sponsor, office, committee, data owners, decision rights.
How this is marked — the rubric
| Sponsor, office, committee and data owners all filled with real named people in current roles, with unstaffed functions marked as gaps | 30 |
| All four decision rights written, one line each, held by a named role, including who can stop a live system alone | 30 |
| Decision rights tested against a real past dispute and amended or defended on that evidence | 20 |
| Names the function most likely to resist, what it will be right about, and the answer to it | 20 |
| Total · pass mark 70 | 100 |
Submitting session 9
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: 9, 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
Decision-rights designOwnership architectureSponsor disciplineCross-functional structureEvidence-based assessment
The Operating Model · The RoadmapSESSION 10
The Twelve Pillars, Sequenced Into One Roadmap
Why This Matters
This is the midpoint, and the session where everything so far becomes one roadmap. The twelve pillars are the readiness system behind this whole programme, and they are not a menu. They are an interlocking, deliberately ordered set in which the dull foundations make the exciting capabilities both possible and safe. Today I take them out of oil and gas language entirely, so a bank, a manufacturer, a hospital or a ministry can score itself against the same twelve. Then you will do exactly that, zero to five, with evidence. Most executives find their scores lower than expected.
25-Minute Module
0:00 – 5:00
A system, not a menu
- Twelve pillars define readiness. Four are foundational — executive fluency, the data foundation, operational visibility and governance, the last of which the doctrine names as a foundation and which runs through pillars 9 and 10 rather than sitting as a numbered pillar of its own. The rest are the capabilities that create visible value.
- THE DOCTRINE
The twelve pillars are not a menu. They are an interlocking system with a deliberate order: governance and data first, autonomy last.
- The relationship is causal, not a matter of taste. The foundations make the high-value pillars possible and safe. Attempt the exciting ones without them and the programme will stall, leak money, or create risk faster than value.
- Which is why a low score on a dull pillar matters more. A two on the data foundation caps every pillar above it; a two on process models costs you one capability.
Read the next step with a pen. Look for your own organisation in the third column.
5:00 – 15:00
The twelve pillars, out of energy language
- The first six carry the foundation and the core operating decision.
Pillar In energy In any business 1. Executive fluency The board can judge where AI pays Executives judge value, data need, risk and return without a vendor translating 2. Data foundation Production data with owners and definitions Every critical domain — customer, claim, patient, shipment, contract — has a data contract and an owner 3. Operational visibility Output and equipment health, by exception The live state of the operation surfaced by exception, not a monthly deck 4. Predictive reliability Predictive maintenance on critical equipment Early warning on anything whose unplanned failure stops revenue 5. The core operating decision Decline, constraint and intervention ranking The recurring decision that sets output: pricing, routing, scheduling, credit 6. Living asset or process models A twin synchronised with the asset A model of how your process behaves, current enough to decide from - The second six carry commercial performance, protection, trust and capability.
Pillar In energy In any business 7. Commercial and bid intelligence Tender intelligence, bid or no-bid discipline How you win work: proposals, pricing, renewals, a memory of every bid lost 8. Security and loss prevention Pipeline surveillance, theft detection Wherever value leaks: fraud, shrinkage, leakage, diversion, unbilled volume 9. Audit-grade compliance Measurement-based emissions reporting Any number filed externally, reconstructible on demand 10. Cybersecurity and trust The enterprise-to-field boundary The boundary between analytical systems and systems of record or action 11. Capability and reskilling Owning the intelligence layer, not renting it Capability mapped to roles, data-engineering and governance tracks first 12. Partnerships and ecosystem Operators, service firms, regulators, financiers Partners chosen for implementation discipline: slide to number - Three cautions. Pillar five is scored too generously: the test is whether you make that decision from evidence, continuously, with the constraint identified as it moves. Pillar nine pays twice, because the lineage that makes a model trustworthy makes a filing defensible. Pillar eleven holds the trap of the transition: data engineering and governance, the two scarcest skills, are the ones most often outsourced.
Most of these twelve are indifferent to your industry.
15:00 – 20:00
The pillars do not change. The sequence does.
- Every organisation eventually builds all twelve. Where you start depends on your economics and where value leaks fastest.
- ENERGY · THE WORKED CASE
- An upstream operator with a large sunk asset base starts at predictive reliability, optimisation of the core production decision and loss prevention: barrels are recoverable from assets already paid for.
- A marginal-field operator holding the same twelve starts at cash-sensitive visibility, decline monitoring and lean models, because on thin margins candidate selection decides survival.
OUTSIDE ENERGY · THE PARALLEL- A bank starts at the data foundation, audit-grade reporting and governance: its regulator already demands reconstructible numbers.
- A retail distributor starts at visibility, the core operating decision — demand forecasting and inventory — and loss prevention, because stock-outs and shrinkage are where its margin moves.
- What does not change is the precedence rule: whatever you start with, the foundation underneath it is built with it, not after it.
- The roadmap is two questions: where does value leak fastest, and what foundation must that pillar have first?
Your competitors hold the same twelve pillars. Your advantage is the honesty of your scores.
20:00 – 25:00
Score it: zero to five, with evidence
- Now the exercise, which is also the assignment. Score your organisation on each pillar from 0 (absent) to 5 (mature), recording the evidence beside every score. A score without evidence is an opinion.
- Three rules keep it honest. Write the evidence first and the number second. A pillar that depends on one person's memory or spreadsheet is not above a two. And no pillar scores higher than the data foundation beneath it.
- THE DOCTRINE
Your three lowest scores are not weaknesses to hide. They are your roadmap — and each one gets an owner before you leave the room.
- Then ask the harder question of those three: which is a foundation? A low foundational score is fixed first, however exciting the alternative looks.
- That is Section 10: our scorecard and the three foundations we will fix first. Week 3 turns to trust, control and the board.
- Practice Exam 2, the midpoint paper, opens after this session: fifteen questions across Sessions 1 to 10. Sit it before Week 3 — it is the honest read on whether the first half landed, and you may resit it as often as you like.
Score it alone first, then with your team. The gap between the two is itself a finding.
Key Terms
Foundational pillars
Executive fluency, the data foundation, operational visibility and governance. They generate no headline value themselves and cap what every other pillar can achieve.
High-value pillars
Optimisation, living process models, commercial intelligence, loss prevention and autonomy. They produce the visible returns, and only once the foundations exist.
Readiness score
A 0 to 5 rating of one pillar — 0 absent, 5 mature — recorded with the evidence that justifies it.
Precedence rule
No pillar can be trusted above the foundation beneath it: your optimisation or autonomy score cannot exceed your data and governance scores.
Self-Check — Before You Submit
1. Your board is enthusiastic about a living process model. Your honest scores: data foundation 1, visibility 1, governance 2, process models 0. What do you recommend?
- A. Fund the model now — board enthusiasm is scarce
- B. Fund the foundations first and stage the model behind a data and visibility milestone, using the board's enthusiasm to fund the unglamorous work
- C. Decline it and propose a different high-value capability
- D. Commission a vendor proof of concept
B. Fund the foundations first and stage the model behind a data and visibility milestone, using the board's enthusiasm to fund the unglamorous work — A model built on a foundation scored 1 drifts from reality and is distrusted. Convert the enthusiasm into funding for the precondition.
2. Why does a low score on a foundational pillar matter more than a low score on a high-value pillar?
- A. Foundational pillars cost more and carry more financial risk
- B. Regulators require foundational pillars; high-value pillars are discretionary
- C. A weak foundation caps every other pillar; a weak high-value pillar costs one capability
- D. Foundational pillars take longer to build
C. A weak foundation caps every other pillar; a weak high-value pillar costs one capability — The asymmetry is structural: fluency, data, visibility and governance are preconditions, so a two on data limits everything above it.
3. An upstream operator and a bank both score themselves against the twelve pillars. What should differ between their roadmaps?
- A. The pillars themselves — some do not apply outside asset-heavy industries
- B. Nothing — both build the twelve in the published order
- C. Only the technology choices
- D. The entry point and the order, chosen by where value leaks fastest, while the pillars and the precedence rule stay the same
D. The entry point and the order, chosen by where value leaks fastest, while the pillars and the precedence rule stay the same — The twelve apply to everyone; only the entry point changes. Calling a pillar inapplicable is how a company excuses itself from loss prevention.
Today’s Assignment
Score all twelve pillars for your own organisation from 0 to 5 with evidence, then name the three that must be fixed before anything else is funded.
- Score each of the twelve pillars 0 to 5 for your own organisation, writing the evidence first and the number second: the system, the document, the meeting, the named owner — or the absence of all four.
- Apply the precedence rule and say whether it forced you to lower a score. If optimisation or autonomy exceeds data and governance, justify it or correct it.
- Name your three lowest-scoring pillars and state for each whether it is foundational or high-value.
- Name the three you will fix first before anything else is funded — they need not be the three lowest — each with an owner by role and a date for the first deliverable.
- The honest part: score your executive team on pillar one with the evidence, then name the pillar your organisation over-rates itself on.
- Aim for 400 to 900 words plus the scorecard. This becomes Section 10 of your charter.
YOUR CHARTER
Section 10 — Our twelve-pillar scorecard and the three foundations we will fix first.
How this is marked — the rubric
| All twelve pillars scored 0 to 5 for the student's own organisation, each with specific evidence naming a system, document, meeting or owner | 35 |
| Precedence rule applied explicitly, with any score lowered or defended on evidence | 15 |
| Three foundations to fix first named with a reason, each with an owner by role and a date for the first deliverable | 30 |
| Scores pillar one honestly with evidence and names one pillar the organisation over-rates itself on | 20 |
| Total · pass mark 70 | 100 |
Submitting session 10
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: 10, 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 2 — Midpoint 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 2 →Closed book, no notes, and Coach Ritchie switches himself off while you sit it.Capabilities Covered
Readiness scoringCross-sector translationSequencing by company typeEvidence disciplineRoadmap construction