THE INTELLIGENT REFINERY LEADERSHIP PROGRAMME
AI earns its place in a refinery one decision at a time.
20-Session Programme · 25 Minutes Per Day · Free · For leaders of Nigeria’s modular and mega refineries
WEEK 1 OF 4 · LEADERSHIP AND REFINERY ECONOMICS · SESSIONS 1–5 · WORK AT YOUR OWN PACE
Orientation · The ThesisSESSION 1
Why AI Belongs in the Refinery's Decisions
Why This Matters
Most refinery leaders I meet have already been shown an AI demonstration. Fewer have been asked the harder question: which operating decision will this improve, and how will we see the improvement in the record? A refinery earns its result through connected decisions about feedstock, units, equipment, product and cash. When one of those decisions is weak, no dashboard repairs it. Over twenty sessions you will build a signed 90-Day Intelligent Refinery Plan for your own facility. Today you choose the one live refinery problem you will carry through all of it, and you describe it honestly, in operating language, before anyone attaches an AI label to it.
25-Minute Module
0:00 – 5:00
A refinery earns its result through decisions
- Feedstock must be selected and received. Units must run within their approved limits. Equipment must remain available. Products must meet their specifications and reach customers. Finance then has to establish whether those activities produced the margin the business expected.
- Each of those is a decision, and they are connected. A late crude barge changes the campaign; a product tank on quality hold changes dispatch; a critical pump waiting on an imported spare changes both.
- THE DOCTRINE
Artificial intelligence becomes useful when it improves one of the refinery's operating decisions and the improvement can be traced through the operating record.
If nobody can name the decision a proposed application improves, the proposal is not yet ready for your approval.
5:00 – 10:00
Start with the business problem
- If a production plan depends on feedstock that has not arrived, a prediction will not unload a vessel. If an operating constraint requires equipment work, a dashboard will not carry out that work.
- Better analysis helps management see the issue earlier and compare options. The value appears when an accountable person makes a better decision and follows it through.
The common discipline The question it puts to a proposal Define the decision What exactly must someone choose, and by when? Establish dependable evidence Which records does the choice rely on, and who owns them? Assign authority Who can act on the answer, physically and financially? Measure the outcome How will we know the decision improved? - MODULAR REFINERY · THE WORKED CASE
- Modular refinery. A smaller installation cannot simply copy the technology programme of a larger one. Its decisions sit close together: one crude receipt, one tank, one pump can carry much of the plan.
INTEGRATED (MEGA) REFINERY · THE PARALLEL- Integrated (mega) refinery. A larger operation has more units, more interfaces and more systems. The same four-part discipline applies; the decisions that matter must still be identified in its own operating model.
I would rather see one decision properly owned than ten applications launched.
10:00 – 15:00
Authority, the regulator and the evidence
- THE DOCTRINE
No model should receive authority merely because its recommendation sounds convincing. Decisions affecting operating conditions, product release or asset integrity keep their technical review and existing approval arrangements.
- The executive task is to establish those boundaries before procurement and to make them visible in the daily operating process.
- The setting, stated precisely: Nigeria's Petroleum Industry Act 2021 establishes the Nigerian Midstream and Downstream Petroleum Regulatory Authority (NMDPRA) and its statutory remit. A refinery's AI programme belongs within that operating and regulatory setting; it does not create a separate permission to operate.
- The International Energy Agency's Energy and AI report examines opportunities for AI across energy operations. It is useful context, not evidence that a particular refinery application will pay for itself.
- Said once, plainly: this programme is leadership training, not engineering, safety, legal or investment advice. Those judgements stay with your qualified specialists and your existing approvals.
Sector reports describe an industry. Your operating record describes your refinery. Approve against the second.
15:00 – 20:00
Read with a live refinery problem
- Choose an issue that already appears in your operating review: unreliable feedstock delivery, a recurring product-quality correction, a critical spare that repeatedly arrives late, or a discrepancy between tank movements and the daily production balance.
- Describe it in ordinary operating language before attaching an AI label. State what someone needs to decide, when the choice is required and how delay affects the business.
- Read with the people who own the evidence. A commercial manager can explain a crude differential but may not know the current unit limitation. A process engineer can assess a campaign but may not know whether treasury can fund it.
- The book's examples use expressly assumed inputs, so that you can challenge them. Replace them with your own verified volumes, product netbacks, contract terms and cash dates.
- THE DOCTRINE
A decision that changes when one assumption moves is telling management where further evidence is worth buying.
- The outcome may be approval, a smaller pilot, a conventional engineering fix or a decision to stop. The quality of the reasoning matters more than the number of applications launched.
The best problem to carry is the one your operating review keeps discussing and never closes.
20:00 – 25:00
Four weeks, the finish line, and Section 1
- Week 1: leadership and refinery economics. Week 2: margin, feedstock and production. Week 3: reliability, integrity, energy and supply. Week 4: enterprise execution, governance and the ninety days.
- The finish line is a signed 90-Day Intelligent Refinery Plan for your own facility: nineteen sections written one per session, assembled and signed in Session 20.
- House rules: write about your own refinery or business; bring a live problem; no number without a baseline; no decision without a named owner; report a blocker the same day. The pass mark is 70.
- Not at a refinery? If you supply, finance, regulate or advise one, choose the refinery decision your work touches most directly.
- Section 1 of your plan is Where we are today: the facility, the live refinery problem and the decision it needs.
If you cannot state the decision without phoning two colleagues, write that down. It is your first finding.
Key Terms
Live refinery problem
An issue already in the operating review, described in operating language, with the decision it needs, its deadline and the effect of delay.
The common discipline
Define the decision, establish dependable evidence, assign authority and measure the outcome. The test applied to every proposed application.
Assumed input
A figure used for teaching or early analysis that has not been verified from the refinery's own records. It must be labelled and later replaced.
90-Day Intelligent Refinery Plan
The signed plan this programme produces for your own facility, written one section per session around one live refinery problem.
Self-Check — Before You Submit
1. Your modular refinery's campaign is repeatedly delayed because crude barges arrive late. A supplier offers a model that predicts arrival times. What is the first question to settle?
- A. Whether the model's accuracy has been independently tested
- B. Whether a larger refinery already uses the same model
- C. Which decision the prediction would improve, and who can act on it
- D. Whether the IEA report supports AI in feedstock logistics
C. Which decision the prediction would improve, and who can act on it — A prediction will not unload a vessel. Value appears only if an accountable person can change a decision with it. Accuracy testing is the tempting answer, but an accurate forecast that nobody can act on improves nothing.
2. Which statement about the regulatory setting is accurate as the book states it?
- A. The Petroleum Industry Act 2021 establishes the NMDPRA, and an AI programme does not create a separate permission to operate
- B. An AI programme needs its own operating permission from the NMDPRA
- C. The IEA Energy and AI report sets requirements for Nigerian refineries
- D. Approved AI applications may bypass existing approval arrangements
A. The Petroleum Industry Act 2021 establishes the NMDPRA, and an AI programme does not create a separate permission to operate — The AI programme sits inside the existing operating and regulatory setting. Inventing a separate AI permission sounds cautious, but it misstates the law; the IEA report is sector context, not a Nigerian requirement.
3. After reading with the evidence owners, your team concludes that the live problem is best solved by a conventional engineering fix, not AI. How should you treat that result?
- A. As a failure of the programme
- B. As a reason to find a different AI use case quickly
- C. As a matter to defer until a supplier proposal arrives
- D. As a legitimate outcome of disciplined reasoning
D. As a legitimate outcome of disciplined reasoning — The outcome may be approval, a smaller pilot, a conventional engineering fix or a decision to stop. Hunting for another AI use case is the tempting reflex, but it measures the programme by applications launched rather than by the quality of the decision.
Today’s Assignment
Name your facility and the one live refinery problem you will carry through this programme, and state the decision it needs as things stand today in your company.
- The facility. Modular refinery, integrated (mega) refinery, or a company that supplies, finances, regulates or advises refineries. Do not name a real facility if confidentiality requires; describe its configuration in a sentence. Not at a refinery? Use the refinery decision your work touches most directly.
- The live problem. One issue already in your operating review, in ordinary operating language, with no AI label.
- The decision. What someone must decide, by when, and how delay affects the business. Name the person who decides today, by role.
- The other accounts. Which functions would give a different account of this problem today, and where they disagree.
- The honest part. One paragraph on what your company currently gets wrong about this problem. Send it as Section 1, four hundred to nine hundred words.
YOUR 90-DAY PLAN
Section 1 — Where we are today: the facility, the live refinery problem and the decision it needs.
How this is marked — the rubric
| Describes one specific problem from the student's own operating review in operating language, without an AI label | 25 |
| States the decision required, its deadline and the effect of delay, with the current decision-maker named by role | 30 |
| Names at least two functions that would give different accounts and states where they disagree | 20 |
| The honest paragraph names a current failing in the student's own company, not the industry | 25 |
| Total · pass mark 70 | 100 |
Submitting session 1
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: 1, 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-first thinkingThe common disciplineRegulatory setting stated preciselyChoosing a live problemPlan authorship
The Toolkit · Six FrameworksSESSION 2
The Reader's Toolkit for the Refinery
Why This Matters
You will be asked to approve refinery technology before anyone has located the weakness in the decision it is meant to improve. Today I give you six tools for that meeting. They are my management tools, carried through The Intelligent Series. They are not official standards or methods endorsed by an educational institution. Judge them by what your team can explain, execute and verify. You will finish by writing one decision record for your Session 1 problem and naming the framework that exposes its weakest link.
25-Minute Module
0:00 – 4:00
Six tools for one meeting
- Apply them to a real decision, not to a technology catalogue.
Framework The question it puts in the operating review Operating-Intelligence Loop Where does sense, decide, act, learn break? Data-Foundation Precondition Does the evidence have identity, unit, time basis, quality status and owner? Autonomy Ladder Inform, recommend or execute a narrowly approved task: which authority is proposed? Value-versus-Risk Map What does an error cost, and what control does that consequence require? Sequencing Discipline Have we proved it in a bounded setting before expanding? Operations-Excellence Lens Which refinery outcome improves, counted once? - None of the six asks which product to buy.
If a proposal cannot answer the first two questions, the other four can wait.
4:00 – 10:00
The loop and the evidence beneath it
- An exchanger's performance appears to deteriorate. A trend may suggest fouling, but the signal alone does not establish the cause. The team checks the measurement, examines process conditions and compares explanations. If intervention is justified, maintenance plans the work and operations makes the equipment available. The review then asks whether performance recovered and whether the diagnosis was useful.
- THE DOCTRINE
A programme that produces an alert but cannot arrange the response has completed only part of the loop. Locate the break before buying a more sophisticated predictor.
- Four breaks, four remedies: the evidence arrives late; the recommendation is unclear; the owner lacks authority; the money is unavailable.
- Dependable evidence does not require every historical record to be perfect. It requires the information used for the chosen decision to have an understood identity, unit, time basis, quality status and owner. A laboratory result must belong to the correct sample and campaign; a tank level to its calibration and movement period.
- Start with a data contract around the first use case: required fields, source, update frequency, permissible gaps and handling of corrections. If the feedstock-quality record is missing, the application should withhold its recommendation and show why, not quietly substitute a convenient average.
A convenient average conceals exactly the uncertainty an executive needs to see.
10:00 – 15:00
Authority and consequence
- The Autonomy Ladder. An application first provides information, then drafts or recommends for human review, and may later execute a narrowly approved administrative task such as a draft work request. Process controls, protective functions, financial commitments and product release demand separate scrutiny.
- THE DOCTRINE
Do not confuse an application's accuracy with its permission. A useful purchasing recommendation is not an approved purchase order.
- For each rung record what it can read, propose and write, who must approve, what happens when it is unavailable or wrong, and how a supervisor pauses it.
- The Value-versus-Risk Map. A search assistant that retrieves a duplicate invoice differs from a recommendation affecting a furnace, product certification or an inspection interval. Consequence decides the evidence, review and control. A large expected benefit does not make an unacceptable safety exposure acceptable.
- Count both kinds of error. A false alert consumes inspection capacity and can interrupt a campaign; a missed event leaves the exposure the application was meant to manage.
The map is a discussion aid, not a score that overrides technical judgement.
15:00 – 20:00
Sequence, the lens and the method
- Sequence: decision and data foundation, a bounded proof, verified response and economics, then a decision to expand. Small scope should still include difficult conditions: missing data, contradictory records, a changed campaign, a user who rejects the recommendation for a legitimate reason.
- MODULAR REFINERY · THE WORKED CASE
- Modular refinery. One product tank or one repeated planning decision may be enough to test the response chain. Keep the trial small, but not easy.
INTEGRATED (MEGA) REFINERY · THE PARALLEL- Integrated (mega) refinery. Extending to a second unit requires an assessment of that unit's equipment, instrumentation, process conditions and ownership. Reused software does not remove the need for a new data contract or acceptance test.
- The lens: deliverable production, product quality, equipment availability, energy performance, working capital, verified compliance and collected contribution. Avoided downtime and extra saleable output may be the same economic effect; count it once.
- Know which method you are using. A material balance is not automatically AI. Linear programming can optimise a plan without machine learning. Advanced process control and a safety instrumented system have defined engineering purposes. Machine learning estimates relationships from data; a physical simulator represents specified engineering relationships; a language model generates or interprets text and may produce unsupported statements; a hybrid needs validation for each part.
Ask every supplier to name the actual method. The answer tells you what validation to demand.
20:00 – 25:00
Carry one decision record, and Section 2
- The same record runs through the whole programme. Place controlled supporting documents behind it and preserve earlier versions when the recommendation changes.
Field What it must hold Decision requested The choice, in one sentence, with its deadline Feasible alternatives Including the current approach Evidence and uncertainty Which records, and what remains unknown Accountable owner A role, not a department Operating and cash consequences What changes on site and when cash moves Next review A date Condition that would change the choice A late delivery, a different assay, an invalid sensor, an unavailable spare - THE DOCTRINE
Record the condition that would change the choice. It makes uncertainty operational and gives the team a reason to revisit the decision before the refinery is committed to an obsolete assumption.
- Section 2 of your plan is the decision record for your live problem, and its weakest link.
Write the condition that would change your mind first. It is the field most often left blank, and the most useful.
Key Terms
Operating-Intelligence Loop
Sense, decide, act, learn. A programme creates value only when evidence leads to an owned decision, a resourced action and a recorded lesson.
Data contract
The agreed definition of the records, meaning, quality, freshness, ownership and failure behaviour required for one application.
Autonomy Ladder
Inform, recommend, then narrowly approved administrative execution. Each rung is a separate grant of authority, not a reward for accuracy.
Decision record
Decision requested, feasible alternatives, evidence and uncertainty, accountable owner, operating and cash consequences, next review, and the condition that would change the choice.
Self-Check — Before You Submit
1. An exchanger-fouling alert has been accurate for months, but cleaning is repeatedly deferred because the unit manager cannot release a production window without group approval. Where is the loop broken?
- A. The evidence arrives late
- B. The owner lacks authority
- C. The recommendation is unclear
- D. The model needs retraining
B. The owner lacks authority — The alert works; the response chain does not. Retraining the model is the tempting purchase, but a more sophisticated predictor will not give the unit manager the authority to act.
2. A supplier describes its crude-planning tool as AI. On inspection it is a linear programme over the refinery's yields and prices. What follows?
- A. It should be rejected because it is not real AI
- B. It needs the same validation as a language model
- C. Its label determines its rung on the Autonomy Ladder
- D. It may be useful, but should be bought and governed as what it actually is
D. It may be useful, but should be bought and governed as what it actually is — Linear programming can optimise a plan without machine learning. Rejecting it is the over-correction: many useful refinery tools are conventional operations research. Naming the method correctly tells you what to validate.
3. The feedstock-quality record for today's campaign is missing. What should an advisory blending application do?
- A. Make the gap visible and withhold the recommendation
- B. Use last month's average quality and proceed
- C. Use the supplier's nominal assay without comment
- D. Recommend the previous campaign's blend
A. Make the gap visible and withhold the recommendation — The data contract should define when the application withholds a recommendation. Substituting a convenient average looks helpful, but it conceals the uncertainty the decision-maker most needs to see.
Today’s Assignment
Write the decision record for your Session 1 problem and state which framework exposes its weakest link in your company.
- The record. All seven fields for your live problem: decision requested and deadline, feasible alternatives including the current approach, evidence and uncertainty, accountable owner by role, operating and cash consequences, next review date, and the condition that would change the choice. Leave unknown fields visibly blank and say why.
- The method. If any analytical tool already touches this decision, name its actual method: material balance, linear programming, simulator, machine learning, language model or hybrid.
- The weakest link. Choose the one framework of the six that exposes the weakest point in how your company handles this decision, and give the specific evidence for that choice.
- The honest part. One paragraph on where your company currently treats accuracy as permission, or counts one benefit twice. Send it as Section 2, four hundred to nine hundred words.
YOUR 90-DAY PLAN
Section 2 — The decision record for our live problem, and its weakest link.
How this is marked — the rubric
| All seven decision-record fields addressed for the student's own problem, with unknown fields marked blank and explained | 35 |
| Names one framework as exposing the weakest link and supports it with specific evidence from the student's company | 25 |
| States a concrete condition that would change the choice | 20 |
| Names the actual method of any existing tool, or states plainly that none is used | 20 |
| Total · pass mark 70 | 100 |
Submitting session 2
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: 2, 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
Loop diagnosisDecision-specific evidenceGranting authority by rungConsequence-led controlNaming the methodDecision-record authorship
Leadership · Chapter 1SESSION 3
Leading the Intelligent Refinery
Why This Matters
Most refinery leadership teams do not lack information. They lack an agreed way of turning several correct accounts into one executable choice. Production can hit its target while dispatch falls short and contribution weakens, and every head of function can be right. This session is about the operating model that resolves that: what the executive team must own, how to write an objective that does not reward the wrong behaviour, and who is accountable when an application enters daily work. You will name the four accountable roles for your live problem, and you will find out where today nobody has authority.
25-Minute Module
0:00 – 4:00
Three accounts of yesterday
- A managing director receives three accounts. Production reports that the crude unit achieved its target. Commercial reports that dispatch fell below plan. Finance reports that contribution is weaker than expected.
- All three may be correct. The refinery could have processed the planned quantity while a quality hold limited dispatch and an unfavourable product mix reduced the economic result.
- THE DOCTRINE
The leadership problem is the relationship between the accounts. An intelligent refinery needs an executive operating model that can carry a choice from analysis into accountable action.
I have sat in that review. The danger is not that someone is wrong; it is that everyone is right and nothing is decided.
4:00 – 9:00
What leadership must own
- Begin with the few decisions that regularly reach the executive team: which crude campaign to fund; what constraint limits profitable delivery this week; which equipment intervention deserves a production window; what must be resolved before a product is released; which digital application is worth sustaining after the trial.
- For each, identify the person accountable for the business result. That person need not perform every technical assessment. Process engineers, laboratory personnel and finance keep their responsibilities; the accountable owner ensures the assessments meet in a decision that can actually be executed.
- A digital transformation manager can recommend applications and evaluate suppliers. Only leadership can reconcile an intervention with production commitments, risk appetite, cash availability and the authority already assigned to the operating organisation.
Delegating the programme is fine. Delegating the refinery's priorities is not possible.
9:00 – 14:00
Make the operating objective explicit
- Describe success as an operating outcome with a boundary and period. Improve performance does not tell the team whether to increase throughput, reduce quality corrections or release cash. Reduce avoidable quality-hold time for one product while preserving release requirements is a decision the organisation can investigate.
A target for Needs alongside it, to prevent hidden substitution Throughput A quality condition and a delivery view Energy A consistent feed and production basis Response time An assessment of whether the response was appropriate - Ask what behaviour a measure could reward if pursued alone.
- Before approving an application, ask the operating team how it decides today. Sometimes the process already identifies the issue but cannot fund the response: that needs treasury and planning discipline. Sometimes information is fragmented across shift logs and commercial spreadsheets: that may justify better integration and analysis. They are different problems.
If the same objective could be met by shipping off-specification product faster, it is not yet an objective.
14:00 – 19:00
An executive review that resolves a choice
- The fictional refinery has product under a quality hold. The meeting should establish the affected quantity, the outstanding test, the expected release time, available tank capacity and the consequence for the next campaign. It should not begin with a general debate over why the dashboard is red.
- Operations may propose delaying the next feedstock change. Commercial may identify a revised dispatch window. The laboratory may confirm that a repeat sample is needed and explain why. Treasury may disclose a payment deadline that changes the feasible campaign. The record shows the option selected and the assumptions needing another review.
- THE DOCTRINE
A polished summary that declares the issue closed before the sample is tested can make the meeting faster while making the decision worse.
- MODULAR REFINERY · THE WORKED CASE
- Modular refinery. The same few people may attend in several capacities. Write down which hat each wears in the decision, so the laboratory's view is not absorbed into production's.
INTEGRATED (MEGA) REFINERY · THE PARALLEL- Integrated (mega) refinery. The hold may affect tank capacity shared with another unit's rundown. The record must show which other commitment the selected option changes.
An analytical assistant may prepare the record from authorised sources. Judge it by whether it preserved the unresolved questions.
19:00 – 25:00
Accountability, the pause, and Pop Quiz 1
Role What it does in practice Executive sponsor Resolves resources and cross-functional obstacles Process owner Carries the application into work Evidence owner Maintains the relevant records Technical custodian Supports and monitors the technical system - Agree how an application will be paused: who can stop recommendations, revoke access or revert to the established workflow. A supplier should not be the only party able to disable an unreliable service. A supervisor should not need to persuade a steering committee to stop using an obviously invalid recommendation.
- The review cadence follows the decision: campaign exceptions may need daily attention, benefit verification a monthly review, authority changes a formal gate. A weekly meeting without a defined decision only adds reporting work.
- THE DOCTRINE
Report recommendations received, accepted, rejected and acted on, and explain material rejections. Paired with the operating result, this shows whether the programme improves choices or merely supplies more to read.
- Section 3 of your plan is the operating objective, its boundary and the four accountable roles.
- Pop Quiz 1 opens now: five questions on Sessions 1 to 3, covering the thesis, the toolkit and leadership ownership. Sit it before Session 4.
If the owner cannot arrange the physical and financial response, repair that authority before commissioning more analysis.
Key Terms
Operating objective
Success described as an operating outcome with a stated boundary and period, specific enough for the organisation to investigate.
Hidden substitution
Meeting a target by shifting the loss elsewhere, such as throughput at the expense of quality or delivery. Prevented by pairing measures.
The four accountable roles
Executive sponsor, process owner, evidence owner and technical custodian, each with a practical, distinct responsibility for an application.
Pause arrangement
The agreed way to stop recommendations, revoke access or revert to the established workflow, without depending on the supplier or a committee.
Self-Check — Before You Submit
1. A morning review on a product under quality hold opens with twenty minutes on why the dashboard shows red. As chair, what do you redirect the meeting to establish?
- A. Who configured the dashboard thresholds
- B. Whether a better dashboard would have warned earlier
- C. Which function caused the hold
- D. The affected quantity, outstanding test, expected release time, tank capacity and next-campaign consequence
D. The affected quantity, outstanding test, expected release time, tank capacity and next-campaign consequence — The meeting exists to resolve a choice. Asking for a better dashboard is the tempting technology answer, but it leaves the hold, the tank and the next campaign undecided.
2. An advisory application's recommendations are plainly invalid after an instrument fault. The shift supervisor must wait for the monthly steering committee to stop using it. What has the programme failed to agree?
- A. The benefit verification cadence
- B. A pause arrangement
- C. The process owner's targets
- D. The supplier's service levels
B. A pause arrangement — The organisation should know who can stop recommendations, revoke access or revert to the established workflow. Supplier service levels feel relevant, but the supplier should not be the only party able to disable the service either.
3. Your team reports that the refinery already identifies avoidable tank-hold delays every week, but cannot pay for the extra handling needed to clear them. What does this suggest?
- A. The information is fragmented and needs integration
- B. An AI application should be approved to predict the delays
- C. The response is unfunded and needs treasury and planning discipline
- D. The objective has no period
C. The response is unfunded and needs treasury and planning discipline — An unfunded response and fragmented information are different problems. Buying better prediction is tempting, but the issue is already identified; the gap lies in funding the response.
Today’s Assignment
Write the operating objective for your live problem, its boundary and period, and name the four accountable roles in your company.
- The choice. Write the recurring management choice behind your live problem in a single sentence, with its deadline, the decision owner and the evidence needed from each function.
- The objective. State it as an operating outcome with a boundary and period. Name the measure, the paired condition that prevents hidden substitution, and the behaviour the measure could reward if pursued alone.
- How it is decided today. Is the weakness an unfunded response, fragmented information, or both? Give the evidence.
- The four roles. Executive sponsor, process owner, evidence owner and technical custodian, each by role. Say who can pause an application and how.
- The authority test. Can the decision owner arrange the physical and financial response? If not, state what authority must be repaired first.
- The honest part. Where in your company nobody has authority today. Send it as Section 3, four hundred to nine hundred words.
YOUR 90-DAY PLAN
Section 3 — The operating objective, its boundary and the four accountable roles.
How this is marked — the rubric
| States the objective with an explicit boundary, period and a paired condition against hidden substitution | 30 |
| Names all four accountable roles by role, not department, and a workable pause arrangement | 25 |
| Diagnoses the current weakness as unfunded response, fragmented information or both, with evidence | 20 |
| Answers the authority test honestly and names the authority gap in the student's own company | 25 |
| Total · pass mark 70 | 100 |
Submitting session 3
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: 3, 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 1 — Decisions Before Tools 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 1 →Closed book, no notes, and Coach Ritchie switches himself off while you sit it.Capabilities Covered
Reconciling functional accountsWriting a bounded objectivePreventing hidden substitutionRunning a decision meetingAssigning accountable roles
Leadership · Chapter 2SESSION 4
Modular and Integrated Operating Models
Why This Matters
Refinery size does not define the operating problem. I have seen a small facility treated as if it were a larger plant with fewer units, and a large one treated as if its crude capacity told you what every downstream step could do. Both mistakes lead to the wrong applications. A smaller installation lives with concentrated constraints and limited staffing; an integrated refinery lives with interfaces and shared resources. The doctrine is common; the implementation follows the real configuration. Today you build the configuration sheet that your AI proposals will be tested against, and separate what is fixed from what can change.
25-Minute Module
0:00 – 5:00
Size does not define the problem
- Two facilities with similar crude capacity can have different conversion capability, utilities, storage, feedstock access and product outlets. A modular installation may have a narrow processing scope or form part of a broader configuration. A large refinery may contain substantial complexity and still be constrained by a single external dependency.
- The U.S. Energy Information Administration describes refining through separation, conversion and treatment, with storage supporting incoming crude and outgoing products. That is useful language for technical context, not a Nigerian requirement.
- THE DOCTRINE
Establish the actual unit set from the refinery's design and operating records. Do not infer capability from the words modular or integrated.
Ask your team to describe the refinery without using its capacity figure. The description you get is the one that matters.
5:00 – 11:00
Begin with the physical configuration
- Map the installed units and their interfaces. Identify which streams can be processed on site, which require treatment or blending, which are sold as intermediates and which depend on an external outlet. Establish the approved operating envelope.
- An optimisation application should receive that configuration as a controlled input. It cannot reasonably recommend a processing route that does not exist.
- Nor is an installed unit available simply because it appears on the flowsheet. Maintenance status, catalyst condition, utilities and operating restrictions can change the practical route.
Nameplate capacity Usable capacity for the decision period A description of design What the connected units can actually do next week Often stated through the crude distillation unit Set by the conditions and connected units that can actually operate Changes rarely Changes with maintenance, catalyst, utilities and restrictions - THE DOCTRINE
Crude distillation capacity is a useful description of design. It is not the capacity of every subsequent processing or delivery step.
When a plan quotes capacity, ask which capacity, and for which period.
11:00 – 17:00
The smaller installation and the integrated refinery
- MODULAR REFINERY · THE WORKED CASE
- Modular refinery. A concentrated set of constraints dominates. A feedstock interruption, an unavailable product tank or a critical pump can affect much of the plan. Concentrate the digital programme on those dependencies rather than imitating the number of applications used elsewhere.
- A practical start: a campaign record linking crude receipt, laboratory evidence, production, tank movements and customer collection. A maintenance application may be worthwhile where the same asset family repeatedly limits production and the response can be supported.
INTEGRATED (MEGA) REFINERY · THE PARALLEL- Integrated (mega) refinery. More interfaces and alternative routes. A change at one unit can affect hydrogen demand, fuel gas, intermediate availability and the product pool.
- A local unit recommendation can be technically attractive while displacing a more valuable use of a common utility or intermediate. The plan must name shared constraints explicitly.
- MODULAR REFINERY · THE WORKED CASE
- Modular refinery. Limited staffing shapes the design. An application needing continuous specialist intervention may be unsuitable even after an impressive demonstration. Ask who reviews recommendations on the night shift, maintains the data mapping and arranges repairs when the service is unavailable.
INTEGRATED (MEGA) REFINERY · THE PARALLEL- Integrated (mega) refinery. Not every decision belongs at group level. Local teams act within approved boundaries. Escalate where a choice changes another unit's commitment, consumes a shared resource beyond allocation or changes a financial exposure.
- Good integration makes interfaces visible without removing competent local authority.
A broad enterprise platform may eventually be useful at a smaller site. Justify it by actual integration and support requirements, not by what a larger refinery bought.
17:00 – 21:00
A common doctrine, different implementation
- Both models need traceable evidence, clear decision rights and verified economics. A smaller installation can use controlled records and modest integration. A larger operation may require a historian, laboratory system, maintenance system, planning tools and an enterprise data environment that share identifiers and time bases.
- Put the burden of maintaining that environment in the business case. More data sources mean more mappings, permissions, quality checks and recovery requirements.
- THE DOCTRINE
A central platform should remove demonstrable work or improve defined decisions. It should not become a destination for every record simply because storage is available.
Storage is cheap. The people who keep the mappings right are not.
21:00 – 25:00
The configuration sheet, and Section 4
Sheet entry Second column: fixed or changeable? Installed units Fixed design information Available routes Changes with maintenance, catalyst and restrictions Major shared utilities Design is fixed; availability changes Product outlets Can change with contracts and logistics Current restrictions Changes during the planning period - Use the sheet to challenge the proposed AI portfolio. Select applications that address the facility's binding problems and can be supported by its workforce.
- Retain the same leadership doctrine across the company, while letting implementation follow each refinery's real configuration.
- Section 4 of your plan is your configuration sheet: fixed design facts versus what changes in the planning period.
The column that matters is the second one. Most proposals fail on a fact someone assumed was fixed.
Key Terms
Approved operating envelope
The approved conditions and limits within which the relevant equipment or process may operate.
Usable capacity
What the connected units can actually process and deliver in the decision period, as distinct from nameplate design capacity.
Shared constraint
A utility, intermediate or resource used by several units, where one unit's gain can displace a more valuable use elsewhere.
Configuration sheet
Installed units, available routes, shared utilities, product outlets and current restrictions, each marked as fixed design fact or changeable in the period.
Self-Check — Before You Submit
1. An optimisation supplier's first recommendation routes an intermediate through a unit that is on the flowsheet but currently restricted because of catalyst condition. What does this tell you?
- A. The model needs more historical data
- B. The unit should be returned to service quickly
- C. The application was not given the current configuration and restrictions as a controlled input
- D. The recommendation is acceptable because the unit is installed
C. The application was not given the current configuration and restrictions as a controlled input — An installed unit is not available simply because it appears on the flowsheet. More historical data is the tempting fix, but history cannot tell the model about today's restriction; the configuration input can.
2. In an integrated refinery, a unit team wants to raise its rate, which would draw more fuel gas than its allocation. Where should the decision sit?
- A. It should be escalated, because it consumes a shared resource beyond allocation
- B. With the unit team, as a local decision
- C. With the supplier of the optimisation tool
- D. At group level, as every decision should be
A. It should be escalated, because it consumes a shared resource beyond allocation — Local teams act within approved boundaries; escalation is needed when a choice consumes a shared resource beyond allocation. Sending everything to group level is the over-correction and removes competent local authority.
3. A modular refinery is offered an application that performed impressively in a demonstration but needs a data specialist to adjust it most days. The refinery has no night-shift specialist. What is the sound position?
- A. Approve it, since the demonstration was strong
- B. Approve it and hire a specialist later
- C. Ask for a broader enterprise platform first
- D. Treat it as possibly unsuitable until the night-shift review, mapping and repair questions are answered
D. Treat it as possibly unsuitable until the night-shift review, mapping and repair questions are answered — Limited staffing affects the operating design. An impressive demonstration is the tempting evidence, but it does not show who reviews recommendations at night or arranges repairs when the service fails.
Today’s Assignment
Prepare the configuration sheet for your facility and use it to test the live problem you have carried since Session 1 in your company.
- The sheet. Installed units, available routes, major shared utilities, product outlets and current restrictions. Mark each as fixed design information or changeable during the planning period. Not at a refinery? Build the sheet for the facility your work touches most directly.
- Capacity. State the nameplate figure your company quotes and the narrowest step that limits usable capacity in the next planning period, from your own records.
- Your operating model. Modular: name the concentrated dependencies and who reviews on the night shift. Integrated: name the shared constraints and the escalation triggers.
- The challenge. Test one current or proposed digital application against the sheet. Does it address a binding problem, and can your workforce support it?
- The honest part. One fact your company treats as fixed that actually changes. Send it as Section 4, four hundred to nine hundred words.
YOUR 90-DAY PLAN
Section 4 — Our configuration sheet: fixed design facts versus what changes in the planning period.
How this is marked — the rubric
| Completes all five sheet entries and correctly separates fixed design facts from changeable conditions | 35 |
| Distinguishes nameplate from usable capacity and names the limiting step from the student's own records | 20 |
| Names the operating-model specifics: dependencies and night-shift review, or shared constraints and escalation triggers | 20 |
| Tests one application against the sheet and identifies a fact wrongly treated as fixed | 25 |
| Total · pass mark 70 | 100 |
Submitting session 4
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: 4, 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
Configuration before capabilityNameplate versus usable capacityShared-constraint awarenessLocal authority and escalationPortfolio fit to workforce
Leadership · Chapter 3SESSION 5
Turning Operating Constraints into Decisions
Why This Matters
A refinery rarely lacks forecasts of what might happen. What it often lacks is a clear statement of which restriction is limiting the result this week, what that restriction is costing, and which feasible action improves the position. Raising a nominal target does not remove a late vessel, a full tank or an unfunded payment. Today you learn to find the binding constraint, record it so that it can be challenged, put a cost on it within a consistent boundary, and prepare responses that can actually be executed. This session closes Week 1.
25-Minute Module
0:00 – 4:00
Which constraint is binding now
- An unconstrained prediction can describe what might happen without helping the refinery choose what to do. The useful question is narrower: which feasible action improves the result under the restrictions that apply now?
- Those restrictions include physical capability, product specifications, contractual commitments, available people and cash.
- A unit can appear underutilised because feedstock is late, product cannot be dispatched, a utility is limited or another unit is unavailable.
- THE DOCTRINE
Raising the nominal target does not remove any of the conditions that are actually limiting the unit. Establish which constraint is binding during the relevant period.
When someone says the unit is underused, ask what is stopping it today. The answer is rarely the unit.
4:00 – 9:00
Build a constraint register
- Record the constraint, its evidence, its consequence and the person responsible for the next action.
Entry What it must identify Tank capacity The tank and usable space, the release condition for existing stock, the incoming stream affected Utility The resource, available supply and the units drawing on it Funding The amount and payment date - Give each entry a validity period. A maintenance restriction may remain until an inspection is completed; a feedstock assumption may change when a vessel schedule is revised; a quoted product netback may expire. Show when each assumption was last checked and what triggers its replacement.
- Do not combine hard limits with preferences. A minimum safety requirement is not interchangeable with a preferred production rate. A customer nomination may be flexible within a contract; an engineering limit may require a formal change process. Classifying the constraint prevents an optimiser treating every restriction as negotiable.
An entry without a validity period becomes a permanent assumption by default.
9:00 – 13:00
Show the cost of the constraint
- Use a consistent economic boundary. If tank availability limits dispatch, compare contribution from feasible deliveries, the cost of additional handling and the effect on the next campaign. If a utility shortage limits processing, consider the displaced production and the cost of a reliable alternative.
- The value of removing a constraint is conditional. It can be small if another restriction becomes binding immediately. A project promising a higher unit rate should show the connected delivery and product-quality conditions.
- THE DOCTRINE
A marginal value from a planning model may identify the next investigation. It is not a guaranteed investment return; it depends on the model, prices and current configuration.
- An actual project has costs, execution time, reliability consequences and cash requirements that a planning calculation may not represent.
Management is buying an operating improvement, not a local capacity figure.
13:00 – 20:00
A worked constraint decision
- The book's example, with inputs assumed for teaching purposes: 1,500 tonnes of a released product are available for immediate dispatch. Customer demand for the week is 2,000 tonnes, but current loading arrangements can move only 1,200 tonnes. Contribution after the costs included is assumed at $40 per tonne.
Step (assumed inputs) Figure Result Released stock beyond current loading 1,500 − 1,200 tonnes 300 tonnes available Additional contribution 300 × $40 $12,000 Temporary loading at $8,000 $12,000 − $8,000 $4,000 before tax, financing and excluded costs Temporary loading at $14,000 $12,000 − $14,000 Reduces stated contribution by $2,000 - Demand exists for more volume, but only 300 additional tonnes are available within the stated inventory. A model that proposes loading 800 additional tonnes has ignored available stock.
- If the next campaign requires the emptied tank, the avoided disruption may change the $14,000 decision. It must be shown separately and supported. It cannot be silently added to make the first calculation attractive. A team approving the $14,000 option may have a valid system reason; it records that reason and the incremental consequence.
- MODULAR REFINERY · THE WORKED CASE
- Modular refinery. One product tank and one loading arrangement can carry the week. The register entry for that tank, with its release status, may be the single most valuable record you keep.
INTEGRATED (MEGA) REFINERY · THE PARALLEL- Integrated (mega) refinery. The emptied tank may serve another unit's rundown or the next campaign. That system value is real only if it is shown on its own line with its own evidence.
Replace every assumed figure here with your own verified volumes, contribution and costs before you rely on the answer.
20:00 – 25:00
Executable alternatives, Section 5 and Practice Exam 1
- Prepare alternatives that include doing nothing, a temporary arrangement and a permanent change where relevant. Describe the approval, personnel, permits, contracts and cash each needs. An option with an unresolved prerequisite appears as conditional, not beside executable options with an unexplained optimistic date.
- AI can link dependencies, identify inconsistent assumptions and help examine alternative schedules. The functions responsible for physical and commercial feasibility still review the recommendation.
- THE DOCTRINE
The constraint register should remain intelligible without the model.
- After the response, measure: did the constraint reduce, what became binding next, did the expected delivery or cash improvement occur? That closes the Operating-Intelligence Loop.
- Section 5 of your plan is the binding constraint, its register entry and two executable responses.
- This closes Week 1. Practice Exam 1 opens now: fifteen questions across Sessions 1 to 5. Sit it before Week 2.
Two executable options beat five attractive ones that each depend on an approval nobody has asked for.
Key Terms
Binding constraint
The restriction actually limiting the agreed result during the relevant period. Removing a different one changes little.
Constraint register
A record of each constraint with its evidence, consequence, next-action owner, validity period and classification as hard limit or preference.
Marginal value
A planning model's estimate of the value of relaxing a restriction. A lead for investigation, not a guaranteed investment return.
Executable alternative
An option with its approvals, personnel, permits, contracts and cash identified. One with an unresolved prerequisite is shown as conditional.
Self-Check — Before You Submit
1. Using the book's assumed figures, a planning tool recommends a temporary loading arrangement to move 800 additional tonnes this week. What is wrong?
- A. Nothing; demand is 2,000 tonnes and loading moves 1,200
- B. Only 300 additional tonnes of released product are available
- C. The $40 per tonne contribution is too low
- D. The temporary arrangement should be permanent
B. Only 300 additional tonnes of released product are available — Released stock is 1,500 tonnes and current loading moves 1,200, leaving 300. The demand gap of 800 tonnes is the tempting figure, but demand does not create inventory; the model ignored available stock.
2. The temporary loading arrangement costs $14,000. The project team presents a positive result by adding an estimated benefit from avoiding disruption to the next campaign into the same line. What do you require?
- A. Approval, since the total is now positive
- B. Rejection, since the first calculation is negative
- C. A higher contribution assumption
- D. The avoided disruption shown separately and supported, with the reason and incremental consequence recorded
D. The avoided disruption shown separately and supported, with the reason and incremental consequence recorded — The system reason may be valid, but it cannot be silently added to make the first calculation attractive. Outright rejection is the over-correction: the $14,000 option may be right if its separate case is evidenced.
3. A planning model reports a high marginal value for relaxing a unit feed limit. A manager proposes treating it as the project's return. What is the correct reading?
- A. It is a lead for investigation, dependent on the model, prices and current configuration
- B. It is the expected return, net of project costs
- C. It is valid only for modular refineries
- D. It confirms the limit is a preference, not a hard limit
A. It is a lead for investigation, dependent on the model, prices and current configuration — Marginal values can identify the next investigation, but an actual project has costs, execution time, reliability consequences and cash requirements the model may not represent. Reading it as a return is the tempting shortcut to a weak business case.
Today’s Assignment
Identify the binding constraint on your live problem, write its register entry, and prepare two executable responses for your company.
- The binding constraint. Ask each function to bring one constraint and its evidence; select the one limiting the agreed result, and say why the others are not binding now.
- The register entry. Constraint, evidence, consequence, next-action owner by role, validity period, and whether it is a hard limit or a preference.
- The cost. Calculate it on a consistent economic boundary, using your company's verified figures where you have them and labelling any assumed figure as assumed.
- Two responses. Two feasible responses, plus doing nothing as the baseline, each with cost, date, approvals, personnel, permits, contracts and cash. Mark any conditional option.
- The measure afterwards. How you will know whether the constraint reduced, what became binding next and whether delivery or cash improved.
- The honest part. Where your company currently raises a target instead of removing a constraint. Send it as Section 5, four hundred to nine hundred words.
YOUR 90-DAY PLAN
Section 5 — The binding constraint, its register entry and two executable responses.
How this is marked — the rubric
| Identifies one binding constraint with evidence and explains why other constraints are not binding now | 25 |
| The register entry includes evidence, consequence, owner by role, validity period and hard-limit or preference classification | 25 |
| Costs the constraint on a consistent boundary with assumed figures labelled and separate benefits kept separate | 25 |
| Presents two executable responses and a do-nothing baseline with prerequisites, dates and a follow-up measure | 25 |
| Total · pass mark 70 | 100 |
Submitting session 5
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: 5, 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 1 — Leadership and Refinery Economics 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 1 →Closed book, no notes, and Coach Ritchie switches himself off while you sit it.Capabilities Covered
Finding the binding constraintConstraint register disciplineConsistent economic boundaryReading marginal valuesExecutable alternatives