THE INTELLIGENT TERMINAL LEADERSHIP PROGRAMME
From released product to collected value.
10-Session Programme · 25 Minutes Per Day · Free · For leaders of Nigerian terminals, depots and distribution
WEEK 2 OF 2 · APPLICATIONS, TRUST, CAPITAL AND THE LEADERSHIP MANDATE · SESSIONS 6–10 · WORK AT YOUR OWN PACE
Part Three · Applications That Change the OperationSESSION 6
Loading Gantries, Dispatch Performance and Equipment Integrity
Why This Matters
The loading gantry is where product availability, truck readiness, customer approval, measurement and release documentation meet, and where a delay created somewhere else becomes visible. Faster loading helps only when the complete dispatch can proceed. Today you follow the dispatch journey from confirmed order to accepted delivery and work through the book's gantry arithmetic, where 100 theoretical cycles cannot produce 100 dispatches when only 75 eligible trucks are ready. Then we turn to equipment integrity, where a predictive model can direct attention but never certifies an asset, and where one commercial assumption moves an illustrative payback from 1.5 years to 15. You will leave knowing what actually constrains your movements, and who holds the integrity decision your plan must respect.
25-Minute Module
0:00 – 5:00
Measure the whole dispatch journey
- A truck may occupy a bay while waiting for a decision that should have been resolved before it entered. Measure the dispatch journey, from confirmed order to accepted delivery. A narrow loading-time metric can shift delay into the queue, the gate or the customer site.
- Value is lost three ways. Queues arise when trucks arrive before their orders, credit, product or documents are ready. Rework arises when identifiers, quantities or destinations disagree. Capacity is lost when a bay cannot operate or the next prepared truck is unavailable.
- Timestamp distinct stages: arrival, readiness confirmation, bay entry, loading start, loading completion, document release and exit. Like the legs of a relay, a slow total says nothing about which leg lost the time.
- THE DOCTRINE
Avoid measuring everything as loading time. Each stage has a different responsible owner and available response.
- With owned and contracted vehicles, customer collections, several products and destinations, record the actual conditions for the route rather than assume that more loading capacity solves the distribution problem.
If your dispatch report has one time column, you are measuring the bay, not the journey.
5:00 – 11:00
Priority, capacity arithmetic and what AI can do
- Where prioritisation is disputed, establish an approved rule reflecting customer commitments, operating restrictions and commercial authority. A model must not silently give priority to the loudest message or the largest order.
Illustrative gantry (Chapter 9) 45-minute cycle 36-minute cycle Six bays for ten hours 3,600 bay-minutes 3,600 bay-minutes Theoretical complete bay cycles 80 100 Eligible trucks actually ready 75 75 Completed dispatches possible No more than 75 No more than 75 - These are simplified capacity calculations, not a promise of 20 additional dispatches. With only 75 eligible trucks ready, shorter cycles cannot produce 100 dispatches; the benefit may instead be shorter waiting and better service. Value the outcome that actually occurred.
- AI can forecast arrivals and readiness by product and time window, compare sequences under approved priorities, flag orders likely to arrive with missing requirements and prepare an exception brief before the truck reaches the gate. A model that labels a truck late without showing the unresolved order condition has limited value.
- Pilot one product or shift against a baseline that includes peak demand, equipment restriction and incomplete documentation. Measure waiting, complete cycle time, dispatches meeting commitment, rework and safety-relevant interruptions, and check whether one stage's gain created delay elsewhere.
- THE DOCTRINE
Do not reward speed achieved by bypassing approved controls.
Ask what the bay does in the minutes a shorter cycle frees. If the answer is wait, the constraint is elsewhere.
11:00 – 16:00
The board's question, and integrity versus availability
- The board sees service against commitment, lost bay availability, queue causes and eligible demand left unserved, with theoretical capacity separate from demonstrated capacity. The question is whether another bay, better maintenance or earlier order readiness would remove the actual constraint. Without a gantry of your own, apply it to the counterpart you depend on.
- Availability and integrity differ. A pump can run while its condition requires intervention; a tank may appear available in the schedule while an inspection restriction makes it unavailable.
- THE DOCTRINE
Predictive methods can support attention and planning. They do not certify fitness for service, extend an inspection interval or replace the competent authority responsible for an engineering decision.
- PRODUCT AND MOVEMENT · THE WORKED CASE
- Product and movement. In a fictional inland depot, the schedule shows a tank available for tomorrow's receipt, but the integrity record carries an inspection restriction. The schedule follows the approved status, not the other way round.
OWNERSHIP AND VALUE · THE PARALLEL- Ownership and value. The commercial team has an urgent customer commitment that depends on that tank. Leadership prevents commercial pressure from redefining the approved status and takes the commercial consequence through its own route: revise, defer or escalate the commitment.
When a schedule and an integrity record disagree, the integrity record stands until its competent owner changes it.
16:00 – 21:00
From warning to feasible action
- Link each asset to its service, criticality, history, condition evidence, inspection findings, maintenance actions and restrictions, and keep the original maintenance description: a label such as pump failure should not erase whether a seal, motor, power supply or operating condition was involved.
- Each warning must connect to an available maintenance action, resources and an accountable owner. A warning without a feasible response is risk information, not a realised reliability benefit; leadership funds the response, changes the plan or accepts a documented restriction within authorised limits.
Illustrative reliability case (Chapter 10) As assumed If only half the downtime constrained profitable delivery Gross contribution protected (120 hours × 50,000 litres per hour × ₦6 per litre) ₦36 million ₦18 million Less maintenance (₦12 million) and recurring system (₦4 million) costs ₦20 million net ₦2 million net Simple payback on ₦30 million upfront 1.5 years 15 years - Illustrative, excluding tax, discounting and other cash effects. The book's lesson: the commercial constraint assumption matters more than the attractiveness of the prediction.
- Pilot an asset class with dependable records, meaningful consequence and a practical response; evaluate false alarms, missed events, warning time and action taken, in time order and including periods with no failure. The board sees overdue integrity actions, constrained availability, recurring failures and verified benefits, not better model scores beside a growing backlog.
Before you admire a prediction, ask what the downtime would have cost if nobody had predicted it.
21:00 – 25:00
Write it down, and Section 6
- Before you close the laptop, write the dispatch or movement journey for your decision and mark which stages you can timestamp today.
- Then name the asset class your decision most depends on, and the role that holds its integrity decision.
- Section 6 of your Terminal Game Plan is The real dispatch constraint and the integrity authority our decision must respect.
Find the constraint before you fund the cure.
Key Terms
Dispatch journey
The sequence from confirmed order to accepted delivery, timestamped by stage so each delay traces to the owner who can remove it.
Demonstrated operating capacity
What the gantry actually achieves under real demand, readiness and restrictions, as distinct from theoretical cycles.
Approved status
The condition the established integrity process has placed an asset in. Neither commercial pressure nor an analytical score may redefine it.
Economically constrained downtime
Downtime that would actually have stopped profitable delivery; only this carries the contribution a reliability case may claim.
Self-Check — Before You Submit
1. In the illustrative gantry, a 36-minute cycle gives 100 theoretical cycles against 80, but only 75 eligible trucks are ready. What should the board expect?
- A. 20 additional dispatches
- B. No more than 75 dispatches; any benefit may be shorter waiting and better service
- C. 100 dispatches once dispatchers adapt
- D. Revenue rising with the extra cycles
B. No more than 75 dispatches; any benefit may be shorter waiting and better service — Shorter cycles cannot produce 100 dispatches from 75 ready trucks. The figures are simplified capacity calculations; the tempting answers confuse theoretical with demonstrated capacity.
2. In a fictional marine terminal, a model shows a pump's readings as stable. The commercial director asks to defer its inspection to keep a line available for a cargo. What is right?
- A. Defer it, because the model shows the pump is healthy
- B. Defer it if the vendor confirms the model's accuracy
- C. Refer it to the competent engineering authority within the approved integrity programme
- D. Let another month of model scores decide
C. Refer it to the competent engineering authority within the approved integrity programme — Predictive methods do not extend an inspection interval or certify fitness for service. A vendor's accuracy claim transfers no engineering authority, and commercial pressure must not redefine approved status.
3. In the reliability case, what happens if only half the downtime would have constrained profitable delivery?
- A. Gross ₦18 million, net ₦2 million, payback 15 years
- B. Net halves to ₦10 million, payback 3 years
- C. Nothing, because the prediction is equally accurate
- D. The upfront investment falls to ₦15 million
A. Gross ₦18 million, net ₦2 million, payback 15 years — The ₦16 million of costs stays, so ₦18 million gross leaves ₦2 million net and a 15-year payback. Halving the net figure is the tempting error; costs do not halve with the benefit.
Today’s Assignment
Name the actual constraint on your decision's movement or dispatch, the stage owners, and for one asset class how a warning reaches a feasible maintenance action within your integrity authority.
- The journey. Your decision's dispatch or movement journey in your company, with the stage timestamps you hold and those you lack, and each stage's owner by role. Name your approved prioritisation rule, or say there is none.
- The actual constraint. Answer Chapter 9's question: another bay, better maintenance or earlier order readiness? Separate theoretical from demonstrated capacity, with evidence.
- The integrity route. For one asset class, how a warning reaches a feasible maintenance action, who holds the engineering decision, and whether its downtime actually constrains profitable delivery.
- The honest part. Where your company measures everything as loading time, or lets commercial pressure lean on approved status. Without a gantry, use the counterpart you depend on. No AI in use yet is normal.
- Four hundred to nine hundred words.
YOUR 90-DAY PLAN
Section 6 — The real dispatch constraint and the integrity authority our decision must respect.
How this is marked — the rubric
| Traces the journey stage by stage with timestamps held and missing and an owner by role for each | 25 |
| Names the actual constraint against the book's three options, separating theoretical from demonstrated capacity with evidence | 30 |
| Traces one asset class from warning to feasible action within a named engineering authority, testing the commercial constraint | 30 |
| Names an honest weakness in dispatch measurement or protection of approved status | 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
The whole dispatch journeyTheoretical versus demonstrated capacityApproved prioritisation rulesIntegrity within engineering authorityTesting the commercial constraint
Parts Three and Four · Distribution, Quality and SafetySESSION 7
Product Distribution, Quality Release and Process Safety
Why This Matters
A truck leaving your gate on time is not a delivered order, and a full tank is not necessarily saleable product. This session follows your decision past the terminal, from release to accepted delivery and the commercial record that follows, because the customer sees the final service and finance needs evidence before it can invoice. Then it turns to product quality and process safety, where the line between information that supports review and authority that permits action matters most. You will name who releases product in your business, and write down which safeguards and approvals no AI application will replace. That paragraph may be the most important one in your plan.
25-Minute Module
0:00 – 5:00
Dispatch is not delivery
- A vehicle can leave on time and still miss the commitment because the receiving site is unready, the route changes or the order evidence is incomplete. The customer sees the final service, not the terminal's internal efficiency.
- THE DOCTRINE
Dispatch is not delivery. Distribution intelligence should connect terminal release to accepted delivery and the commercial record that follows.
- Identify where responsibility changes and what evidence confirms each handover, above all when another party provides transport.
- Measure route conditions, travel variability, customer access and receiving capacity for your actual network. A national average is not a route plan, and a scheduling application cannot create permission or readiness by assuming a transport resource is available.
Ask where your company's responsibility ends on each route. If nobody can say, that handover has no evidence.
5:00 – 11:00
The evidence chain from order to invoice
- Link seven records: order, customer readiness confirmation, transport assignment, terminal release, route events, delivery acknowledgement and invoice. Keep estimated and actual arrival separate, and record changes to destination, quantity and acceptance conditions with approval.
- Use location information proportionately, with authorised access and an established purpose. It may support arrival estimation. It does not prove the quantity delivered or the quality accepted; those need custody and receiving evidence.
- PRODUCT AND MOVEMENT · THE WORKED CASE
- Product and movement. In a fictional distribution network, tracking shows a truck at an industrial customer's gate in the morning, but the receiving tank is unavailable until the next day. A predicted early arrival is then of little use.
OWNERSHIP AND VALUE · THE PARALLEL- Ownership and value. Finance cannot invoice on the location signal; it needs the delivery acknowledgement and accepted custody evidence. Cash waits on distribution evidence, and nobody is presumed at fault because the record is incomplete.
- AI can forecast arrival ranges, flag commitments at risk and compare allocation options alongside receiving readiness, preparing actionable exceptions: confirm the slot, revise an authorised dispatch, arrange a resource or escalate a blocked delivery.
Illustrative distribution case (Chapter 11) Figure Ten repeat trips avoided at ₦150,000 each ₦1.5 million gross avoided transport cost Still to deduct and verify Coordination cost; that the trips would otherwise have occurred; which party receives it under the contract - Additional sales need evidence of additional demand served. Quicker delivery of the same sale is not new revenue.
Before anyone counts the ₦1.5 million, ask whose contract it lands in.
11:00 – 15:00
The distribution pilot and board measures
- Pilot one route family and product. Compare predicted with actual arrival at the customer decision deadline, record exceptions caused by terminal, transport, customer and commercial instructions, and test failed communication and missing delivery acknowledgement.
- Measure whether alerts led to useful action, customers received dependable updates and the evidence supported timely invoicing.
- The board sees delivery against commitment, failed delivery causes, unresolved acceptance records and cash waiting on distribution evidence. If customers or contractors carry the product, apply this to the counterpart whose acknowledgement you depend on.
Missing acknowledgements are not administration. They are cash waiting on distribution evidence.
15:00 – 21:00
Quality release and process safety
- An available quantity is not necessarily saleable or releasable. Connect product identity, sampling, laboratory results, specification, hold status and release decision. A certificate attached to the wrong batch can be more misleading than a visibly missing one.
- THE DOCTRINE
Distinguish information that supports review from authority that permits action. A model may summarise a certificate. It does not become the person authorised to release the batch.
- Where aviation fuel is in scope, use the applicable controlled quality-assurance standards; the Joint Inspection Group publishes standards and guidance for aviation fuel operations [5]. AI should flag ambiguity, not invent a specification or acceptance limit.
- API Standard 2350 addresses overfill prevention in petroleum storage tanks, and an AI inventory forecast is not an independent overfill safeguard [6]. The HSE safety bulletin on high-level switches following Buncefield illustrates why the reliability and correct use of specific protective equipment require attention; keep that assurance separate from any analytical application [7].
- A missing signal should produce an explicit degraded-information state, not a smooth forecast, and the response follows the approved facility process. Start with a read-only quality-document review testing wrong-batch certificates, missing results, superseded documents and product on hold.
- THE DOCTRINE
Do not average unacceptable control failures into an overall accuracy score. A model useful for routine document review may still be unsuitable for a release-related decision.
If a forecast could ever be mistaken for a safeguard, write down in plain words that it is not one.
21:00 – 25:00
Write it down, and Section 7
- The board sees open quality restrictions, repeated evidence defects, movements prevented by correct checks and overdue corrective actions, with safety-critical assurance reported through the established process.
- Before you close the laptop, mark the link in your evidence chain most often missing when finance tries to invoice, and write the role that releases product in your business.
- Section 7 of your Terminal Game Plan is From release to accepted delivery: our quality authority and the safeguards AI will not replace.
Your plan is stronger for what it refuses to automate.
Key Terms
Accepted delivery
The receiving party's acknowledgement of quantity and quality, supported by custody and receiving evidence. Dispatch and location signals do not establish it.
Batch identity
The controlled link between a batch, its samples, results, specification, hold status and release decision.
Independent overfill safeguard
Engineered arrangements and procedures established through the competent engineering and operating process; API Standard 2350 addresses overfill prevention [6]. An AI forecast is not one.
Degraded-information state
An explicit status when a signal is missing, so the approved process governs the response rather than a smooth forecast.
Self-Check — Before You Submit
1. A fictional depot's tracking data shows a truck stopped at the customer's site for two hours. What does that establish?
- A. The quantity delivered
- B. That the customer accepted the product
- C. Grounds to raise the invoice
- D. It may support arrival estimation; acceptance needs custody and receiving evidence
D. It may support arrival estimation; acceptance needs custody and receiving evidence — A location signal does not prove the quantity delivered or the quality accepted. Invoicing on it is the tempting shortcut; the invoice needs acknowledgement and accepted custody evidence.
2. How should a board treat the illustrative ₦1.5 million from ten avoided repeat trips at ₦150,000 each?
- A. As gross avoided cost, before coordination cost, verified and subject to contract terms
- B. As new revenue for the terminal
- C. As net profit to the terminal
- D. As a sector benchmark for distribution savings
A. As gross avoided cost, before coordination cost, verified and subject to contract terms — The book calls it gross avoided transport cost, before coordination cost, with the trips verified and the contract deciding who benefits. Net profit to the terminal ignores both the costs and the contract.
3. A fictional terminal's quality-document assistant scores highly overall but missed two wrong-batch certificates in testing. The sponsor wants it to approve release decisions. What should happen?
- A. Approve it, because the overall score is high
- B. Keep it for evaluated routine review, assess the missed critical defects separately, and keep release with authorised quality personnel
- C. Withdraw all AI from quality permanently
- D. Allow release approvals automatically once the score improves
B. Keep it for evaluated routine review, assess the missed critical defects separately, and keep release with authorised quality personnel — Unacceptable control failures are never averaged into an overall score, and authority stays bounded by the evaluated use. Withdrawing everything overreacts: the model may still suit routine review.
Today’s Assignment
Trace your decision from terminal release to accepted delivery and the evidence invoicing needs, name the quality and release authority that bounds it, and state which safeguards and approvals no AI application in your business will replace.
- The chain. Your decision in your company through the seven links from order to invoice: who records each and where responsibility changes hands, and what finance needs before it can invoice.
- Quality authority. By role, who holds batch release and quality holds for your product, and how certificates stay attached to the right batch. If aviation fuel is in scope, how the applicable standards are identified.
- The safeguards paragraph. Which safeguards and approvals no AI application will replace, including overfill prevention, and what your systems show when a signal is missing.
- The honest part. Where your company treats dispatch as delivery, or a certificate could attach to the wrong batch. No AI in use yet is normal.
- Four hundred to nine hundred words.
YOUR 90-DAY PLAN
Section 7 — From release to accepted delivery: our quality authority and the safeguards AI will not replace.
How this is marked — the rubric
| Traces release to accepted delivery with each handover and the evidence invoicing needs | 30 |
| Names the quality and release authority by role and how batch identity is preserved | 25 |
| States which safeguards and approvals AI will not replace, including overfill prevention and a degraded-information state | 30 |
| Names an honest weakness in delivery evidence or batch identity | 15 |
| 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
Dispatch is not deliveryThe order-to-invoice evidence chainProportionate use of location dataInformation versus release authoritySafeguards AI does not replace
Part Four · Trust, Control and AuthoritySESSION 8
AI Governance, Cybersecurity and Controlled Autonomy
Why This Matters
Every application you approve brings new dependencies: data access, model behaviour, suppliers, interfaces and the people who review its outputs. A policy file does not govern any of them. This session gives you the governance record, field by field, and the discipline of keeping analysis access apart from operational control. Then it takes on the harder question an agent raises: not whether its answer is good, but what it is authorised to do, with which evidence and under whose accountability. You will set permissions action by action and decide what happens when a task stops halfway. Pop Quiz 2 opens at the end.
25-Minute Module
0:00 – 4:00
Governance is assigned responsibility
- Governance means assigning responsibility for an application's dependencies and making the required controls part of daily operation.
- THE DOCTRINE
A policy file alone does not govern a system. The terminal needs an application register, a decision owner, authorised access, evaluation evidence, change control and a response when the application fails.
- Keep the NIST AI Risk Management Framework's govern, map, measure and manage functions as an organising reference, as in Session 2, and tailor controls to the application rather than assume that adopting the framework establishes adequacy [2].
- The approval record explains why AI is appropriate compared with a simpler rule, a repaired instrument or an improved workflow. No system gains authority merely because its supplier labels it advanced.
If your register cannot name an owner for an application, you have found your first governance action.
4:00 – 10:00
The governance record, field by field
- Write one record for each application. In plain language, its fields ask:
Field What it must state Purpose The business decision the application supports Users Who receives the output and acts on it Evidence sources The authorised records it may read Permitted outputs A summary, a recommendation or a draft Prohibited actions What it must never do, such as release product or amend stock Consequence of error What happens if it is wrong, late, absent or treated as authoritative Fallback The operating route when it is unavailable or wrong Owners Sponsor, operating owner, technical owner and the independent reviewers the actual risk needs - Retain test cases from the real operation: missing information, unusual products, incorrect documents and attempted misuse. A technically accurate recommendation can still be unsafe if its recipient mistakes it for an approved instruction.
- Changes to models, prompts, retrieval sources or interfaces can change behaviour. Record the change, reviewer and retesting, and do not treat an update as routine because the supplier releases it automatically.
Write the prohibited actions first. They are easier to agree and harder to forget.
10:00 – 15:00
Operational technology and access
- Operational technology, or OT, is the systems that monitor and control physical equipment. NIST SP 800-82 Rev. 3 discusses OT security, including the distinctive performance, reliability and safety needs of OT environments; design terminal applications with competent OT and cybersecurity personnel [8].
- THE DOCTRINE
Keep analysis access distinct from operational control. Use only authorised interfaces and the minimum access required for the task.
- Document what happens if a supplier service becomes unavailable, a credential is compromised or an interface returns stale information, and keep an operating route that does not depend on one external analytical service.
- The first pilot runs read-only with a clear purpose, testing inappropriate requests and unsupported conclusions. The board sees applications, owners, authority levels, control issues, changes and incidents, including applications stopped or restricted after evaluation: evidence of governance working, not an embarrassment.
Ask whether the application can write to anything that moves product. For a first pilot, the answer should be no.
15:00 – 21:00
A bounded agent and permission by action
- An agent prepares or executes a sequence of digital tasks, so you govern the sequence, not merely one answer. Describing a system as autonomous answers none of the real questions.
Action The book's dispatch-brief agent, initially Read Authorised order, stock-status and transport records Prepare Assembles tomorrow's dispatch brief and lists unresolved conditions Propose Drafts requests for review Approve Not the agent; approval and execution may sit with different people or systems Execute Cannot release product, amend stock, change customer credit or issue a movement instruction - Any wider permission needs separate evaluation and approval, explicit limits on record types, transaction scope and destinations, and a controlled preview of consequential changes.
- PRODUCT AND MOVEMENT · THE WORKED CASE
- Product and movement. In a fictional inland depot, an order changes mid-preparation and a stock-status record proves stale. If a request was interrupted, establish whether the action occurred before retrying, or duplicate reservations and tickets can follow.
OWNERSHIP AND VALUE · THE PARALLEL- Ownership and value. The brief lists a customer with an unresolved credit condition; finance decides. Time saved counts only when work is actually removed, and reporting time saved is not dispatch reliability gained.
- THE DOCTRINE
The system should stop when required evidence or authority is missing and hand the decision to its owner. A confident narrative is not a recovery mechanism.
- Begin with preparation and human review, and prove recovery after interruption before any automatic administrative execution. The board sees approved actions, attempts outside permission, duplicate or partial executions and verified net time saved.
The autonomy ladder moves only as the evidence supports it, and leadership keeps accountability for every permission decision.
21:00 – 25:00
Write it down, and Section 8
- Before you close the laptop, write the eight governance fields for the application your decision might use, even if it is still a proposal, and mark each of the five actions permitted, approval-dependent or prohibited.
- Section 8 of your Terminal Game Plan is The governance record and permission by action for our application.
- Pop Quiz 2 opens now: five questions on Sessions 6 to 8, closed book.
Permission is granted per action, never per system.
Key Terms
Governance record
One record per application: purpose, users, evidence sources, permitted outputs, prohibited actions, consequence of error, fallback and owners.
Operational technology (OT)
Systems that monitor and control physical equipment, with distinctive performance, reliability and safety needs [8].
Permission by action
Separate authority for read, prepare, propose, approve and execute, with approval and execution able to sit with different people.
Controlled preview
A consequential change shown to an authorised person before it takes effect, so it can be inspected and stopped.
Self-Check — Before You Submit
1. A supplier seeks wider authority for its application because the new version is its most advanced. What should the approval record show?
- A. Why AI beats a simpler rule, repaired instrument or improved workflow, with the new authority separately evaluated
- B. The supplier's description of the new version
- C. That adopting the NIST AI RMF establishes adequacy
- D. How many users prefer the new version
A. Why AI beats a simpler rule, repaired instrument or improved workflow, with the new authority separately evaluated — No system gains authority because its supplier labels it advanced, and new authority needs separate evaluation. Citing the NIST AI RMF is tempting, but the book says adoption does not establish adequacy.
2. A fictional terminal's scheduling agent, approved to submit loading-slot reservations, times out while sending one. The user asks it to try again. What comes first?
- A. Resend, since each step appeared valid
- B. Establish whether the reservation was made before attempting it again
- C. Give the agent execute permission to confirm the slot
- D. Accept the agent's narrative of what happened
B. Establish whether the reservation was made before attempting it again — After an interruption, establish whether the action occurred before retrying, or duplicates follow. A confident narrative is not a recovery mechanism, and widening permission rewards the failure.
3. After evaluation, a fictional depot restricts one AI application to read-only and stops another. How should the board report treat them?
- A. Omit them to avoid embarrassment
- B. Report only applications still running
- C. Report them only as write-offs
- D. Include them as evidence of governance working
D. Include them as evidence of governance working — Chapter 13 asks the board to see applications stopped or restricted after evaluation. Treating them only as failures discourages the honest evaluation the board needs.
Today’s Assignment
Write the governance record for the application your decision might use, set its permission for each action, and name its read-only first pilot and the recovery rule for an interrupted task.
- The governance record. All eight fields for the application in your company, with owners by role.
- The comparison. Why AI rather than a simpler rule, repaired instrument or improved workflow. If the simpler route wins, say so.
- Permission by action. Read, prepare, propose, approve and execute, each permitted, approval-dependent or prohibited, with the approver.
- Access and recovery. How analysis access stays apart from operational control, the read-only first pilot, and how an interrupted task is recovered without repeating an action.
- The honest part. Where your business runs a tool or supplier service with no owner, no change record or excess access. No AI in use yet is normal: work from the proposal in front of you.
- Four hundred to nine hundred words.
YOUR 90-DAY PLAN
Section 8 — The governance record and permission by action for our application.
How this is marked — the rubric
| Completes all eight fields with owners by role and a consequence of error covering wrong, late, absent and over-trusted output | 30 |
| Justifies AI against a simpler rule, repaired instrument or improved workflow | 20 |
| Sets permission for each of the five actions with an approver, keeping analysis access apart from control | 30 |
| Names a read-only first pilot and a recovery rule, with an honest weakness | 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 — Dispatch, Distribution, Safety and Control 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
The governance recordAnalysis access versus controlPermission by actionFailure and recoveryHonest value of agents
Part Five · Capital and PeopleSESSION 9
Working Capital, Terminal Economics and Local Capability
Why This Matters
A profitable sale can consume more cash than the business can fund. That is the uncomfortable truth behind this session. AI proposals often blur revenue, contribution, profit, cash released and financing cost avoided, and a board that approves a blurred number cannot later tell what it bought. You will work through the book's three illustrative cash cases and read each for what it is not. Then you will price the investment completely and separate its benefits without counting anything twice. Finally, the people: a routine that depends on one analyst is not yet a capability. Practice Exam 2 opens at the end.
25-Minute Module
0:00 – 5:00
Five numbers that are not the same
- In plain words: revenue is what the customer is invoiced; contribution is what remains after costs that move with volume; profit is what remains after all costs; cash released is money no longer tied up in stock or receivables, a balance-sheet and cash-timing effect; financing cost avoided is borrowing cost saved when that cash reduces funded borrowing.
- THE DOCTRINE
A profitable sale can consume more cash than the business can fund. A clear financial model makes the operating change and its cash consequence inspectable.
- A demand forecast contributes evidence; it does not set the risk appetite or authorise a purchase. The executive chooses the service and funding position using scenarios and accountable assumptions.
- A single days-of-stock measure can hide restricted inventory. Keep usable owned stock separate from custody product and stock on hold: Session 3's ownership ledger, now priced.
When a proposal says savings, ask which of the five numbers it means.
5:00 – 11:00
Three illustrative cash cases
Illustrative case (Chapter 15) Arithmetic What it is not Inventory capital 2 million litres a day × (10 − 8) days = 4 million litres × ₦900 = ₦3.6 billion potential capital release ₦3.6 billion of annual profit Financing cost ₦3.6 billion × an assumed 24% for a full year = ₦864 million simplified effect Certain: it depends on borrowing terms, duration and use of the cash Receivables ₦300 million daily invoiced sales × 5 days = ₦1.5 billion cash released (simplified steady state) Additional revenue; it needs actual collection evidence and excludes any shift in sales mix or payment terms - Service must stay within the approved requirement; a reduction achieved by missing deliveries is not dependable, so test delayed replenishment, higher demand and product on hold. If invoices wait for delivery acknowledgement, improve that record flow before assuming collection prediction will solve the cash problem.
Illustrative movement (Chapter 15) Cash flow Cumulative position Day 0, product purchase −₦400 million −₦400 million Day 3, logistics −₦12 million −₦412 million (maximum funding need) Day 10, customer receipt +₦440 million +₦28 million (final surplus) - Before financing and other cash effects. A later receipt changes funding duration and potentially financing cost, not the sales value. Model doubtful collection separately, never delaying and removing the same receipt unexplained. Finance verifies signs, units and dates.
Ask to see the day your cash position is lowest. That number, not the margin, decides whether you can fund the movement.
11:00 – 16:00
The investment decision without double counting
- The complete cost: data preparation, integration, specialist review, training, recurring service, support and the operating response. A software licence is not the full cost. Compare simpler alternatives and test a downside.
- Separate additional contribution, cost avoided, working capital released and financing-cost effect. Never count a stock release as both annual income and cash released, or the same constrained volume in both reliability and dispatch benefits.
- PRODUCT AND MOVEMENT · THE WORKED CASE
- Product and movement. In a fictional refinery-connected facility, a reliability application and a dispatch application both claim the same constrained litres on one line. The volume is counted once, where the evidence supports it.
OWNERSHIP AND VALUE · THE PARALLEL- Ownership and value. The same facility cuts owned stock days. Finance records working capital released and, separately, a financing-cost effect tested against actual borrowing, never annual income as well.
- THE DOCTRINE
Simple payback is a screening measure. A larger investment may require discounted cash analysis, tax treatment and detailed funding review.
- The board sees usable owned inventory, restricted capital, receivable duration, unbilled completed movements and funding need, with AI benefits reconciled to finance-approved evidence.
An intelligent terminal earns the right to hold less capital by making its service and replenishment decisions more dependable.
16:00 – 21:00
A routine that survives one absence
- Local capability means the people responsible understand the evidence, can challenge a recommendation and can sustain the routine through staff changes. Training after implementation is too late if users never shaped the workflow.
- Skills follow responsibility: executives frame decisions and challenge economics; terminal managers connect recommendations to readiness and authority; measurement specialists examine quantity evidence; engineering evaluates asset implications; finance reconciles value and cash; digital specialists learn the operational consequence of their design choices.
- Train on the terminal's own controlled examples: find missing evidence, an unauthorised action and a double-counted benefit. Define supplier ownership of records, configurations, documentation and support, and how the business continues if the contract ends.
- THE DOCTRINE
Do not make acceptance of recommendations the primary performance measure. The aim is competent challenge, not automatic compliance with the assistant.
- Assign a deputy for each critical role, test the routine with the usual analyst absent and have another shift reproduce the result. The board sees capability by role, dependence on individuals and examples of appropriate challenge.
The leadership standard is a business that can operate, question and sustain the system it has approved.
21:00 – 25:00
Write it down, and Section 9
- Before you close the laptop, write your decision's cash consequence as a short timeline or working-capital effect, labelled with what it is not, and name the one person your routine would stop without.
- Section 9 of your Terminal Game Plan is The economics of our decision without double counting, and the team that sustains it.
- Practice Exam 2 opens now: fifteen questions on Sessions 1 to 9, closed book.
Name the number and the person. Both must survive scrutiny.
Key Terms
Contribution
What a sale or movement earns after the costs that vary with it; distinct from revenue and from profit.
Cash released
Money no longer tied up in inventory or receivables: a cash-timing effect, not annual profit or additional revenue.
Financing-cost effect
Borrowing cost avoided when released cash reduces funded borrowing; it depends on actual terms and duration.
Maximum funding need
The deepest cumulative cash position before the customer pays; ₦412 million in the book's illustrative timeline.
Self-Check — Before You Submit
1. A proposal says cutting owned stock from ten to eight days on 2 million litres of daily demand, at an assumed ₦900 per litre, adds ₦3.6 billion to annual profit. What is wrong?
- A. Nothing; the arithmetic is right
- B. The release is 2 million litres, not 4 million
- C. ₦3.6 billion is a potential capital release; the simplified annual financing-cost effect at an assumed 24% is ₦864 million
- D. It should be divided by ten days
C. ₦3.6 billion is a potential capital release; the simplified annual financing-cost effect at an assumed 24% is ₦864 million — The arithmetic is right, which is the trap: ₦3.6 billion is a balance-sheet and cash-timing effect, not annual profit. The ₦864 million financing-cost effect is simplified and depends on the borrowing terms, duration and use of the released cash.
2. In the illustrative timeline, the ₦440 million arrives on day 20 instead of day 10, nothing else changed. What changes?
- A. Funding duration and potentially financing cost, not sales value
- B. The final surplus falls to ₦14 million
- C. Maximum funding need rises to ₦852 million
- D. The movement becomes unprofitable
A. Funding duration and potentially financing cost, not sales value — A later receipt changes funding duration and potentially financing cost, not the sales value. Cutting the surplus confuses timing with value; doubtful collection is modelled separately.
3. A fictional depot's managing director proposes measuring staff on the share of AI recommendations they accept. What is better?
- A. Adopt it to drive adoption
- B. Apply it to junior staff only
- C. Suspend the assistant until acceptance rises
- D. Measure whether decisions use evidence and outcomes are reviewed, with a route to report errors
D. Measure whether decisions use evidence and outcomes are reviewed, with a route to report errors — Chapter 16 warns that an acceptance measure discourages appropriate challenge. The aim is competent challenge, so measure evidence-based decisions and reviewed outcomes.
Today’s Assignment
State the cash consequence of your decision, separate its benefits without double counting, list the complete cost, and name the people, deputies and supplier terms that let the routine survive without one analyst.
- The cash consequence. A movement-level timeline with maximum funding need, or a working-capital effect, for your decision in your company, with assumed figures marked as assumed.
- The benefits. Each placed in one of the four categories with its counterfactual, and a double-counting check against Section 6's dispatch and reliability claims.
- The complete cost. Every cost line, the simpler alternative and a downside case.
- The team. Skills by role, a deputy for each critical role, and supplier terms on records, configurations, documentation and support.
- The honest part. Where your business reports cash released as profit, or depends on one person. No AI in use yet is normal.
- Four hundred to nine hundred words.
YOUR 90-DAY PLAN
Section 9 — The economics of our decision without double counting, and the team that sustains it.
How this is marked — the rubric
| States the cash consequence with correct signs and units, and says what it is not | 30 |
| Separates benefits into the book's categories with a counterfactual and a double-counting check | 25 |
| Lists the complete cost with a simpler alternative and a downside case | 20 |
| Names roles, deputies and supplier terms that remove single-person dependence, with an honest weakness | 25 |
| 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.ASSESSMENT NOW OPEN
Practice Exam 2 — Full Syllabus opens now. Sit it before you start the next session — it is how you find out whether the doctrine landed, and the practice papers may be resat as often as you like.
Sit Practice Exam 2 →Closed book, no notes, and Coach Ritchie switches himself off while you sit it.Capabilities Covered
Revenue, contribution, profit and cashMovement-level cash timelinesComplete cost and benefit categoriesAvoiding double countingCapability without single-person dependence
Capstone · The Leadership MandateSESSION 10
The Terminal Game Plan
Why This Matters
Nine sessions ago you named your company, its terminal archetype and one unresolved decision. Every section since has been written around it. Today you assemble those sections into one Terminal Game Plan, lay them across Chapter 17's ninety days, complete the Appendix A pilot charter and the Appendix B scorecard and acceptance gates on evidence, and sign it for your executive committee. Chapter 17 is clear that the first 90 days are a management sequence, not a guarantee of a fully transformed terminal. The plan's job is to make the next decision inspectable. Then you sit the Final Exam, and we close where the book closes.
25-Minute Module
0:00 – 5:00
The ninety days, stage by stage
- The game plan starts with one decision, a supported baseline and an accountable operating response, not a long list of applications.
Stage (Chapter 17) By the end of the stage Days 1 to 30: establish the operating truth. Map the four ledgers and handovers, evidence, authority and response capacity; agree baseline and valuation; compare a simpler alternative The sponsor holds a signed pilot charter, evidence sources, accountable owners, evaluation cases and an agreed stop condition Days 31 to 60: test the decision and response, bounded, with human review; test incomplete data, changed instructions and disruption; review weekly Leadership knows where it helps and fails, whether the response is practical, and the recurring cost Days 61 to 90: decide, do not merely extend. Reconcile results and finance-approved benefits; test recovery and dependence on individuals Scale, revise, continue within a bounded scope or stop - If day-30 items are missing, resolve them rather than expand the demonstration. A repaired instrument, clear cut-off or documented decision right can be an excellent result of the AI review.
- THE DOCTRINE
Do not change the benefit definition simply because early results are disappointing. A justified new definition is recorded as a revised evaluation with a clear reason.
- Expansion names the next products, tanks, routes or users and the evidence required. Approval at one site is not approval for every terminal archetype.
Agree the stop condition before you start. Later, it becomes a contest over who sponsored the technology.
5:00 – 9:00
Twelve priorities and your starting point
- Chapter 17's twelve priorities, in brief: controlled identity; reconciled stock; visible ownership and release restrictions; custody evidence closed before invoicing; loss investigated without verdicts; confirmed constraints and forecast ranges; accountable ship and shore handovers; whole-journey dispatch readiness; integrity within engineering authority; access and autonomy bounded by action; reconciled service, capital and benefits; a local team that can challenge and sustain the routine.
- PRODUCT AND MOVEMENT · THE WORKED CASE
- Product and movement. A marine terminal may begin with receiving and berth readiness, an inland depot with dispatch and replenishment, a refinery-connected facility with tank allocation and constrained movement.
OWNERSHIP AND VALUE · THE PARALLEL- Ownership and value. A third-party storage operator may begin with entitlement and custody evidence, a distribution network with customer readiness and delivery acceptance.
- The start follows the unresolved decision, not the appeal of a technology. Mark the priorities your plan addresses, and the one it most neglects.
If your plan does not address all twelve priorities, say which you are deliberately leaving for later.
9:00 – 14:00
Appendix A: working forms, not authority
- THE DOCTRINE
A completed form supports judgement; it does not confer technical or operating authority.
- A1, the pilot charter: the decision in one sentence naming product, facility and time horizon; operating owner, sponsor and evidence owners; baseline; application and simpler alternative; authority; evaluation; costs; stop or revise conditions; approvals.
- A2 runs daily movement readiness, confirmed separated from forecast. A3 records inventory exceptions, where an AI confidence score is never a finding. A4 reviews custody packs, where presence alone does not establish validity.
- A5, the benefits ledger: a recommendation with no action remains analysis, and a predicted receipt never enters the actual cash column. A6 records the day-90 decision on the scope actually tested. Working records 1 to 8 add a data dictionary and a change and incident record.
Fill the forms with evidence you could show an auditor. A marked gap is more honest than a guess.
14:00 – 20:00
Appendix B: what the board sees and decides
- B1, the scorecard, has seven areas: inventory; custody and quality; scheduling and service; equipment; distribution; capital and cash; AI control and capability. For every red or amber issue, state the decision required, options, consequence of delay and accountable owner.
- B2, decision rights by consequence: the sponsor approves purpose, resources and continuation; the operating owner, workflow and response; measurement, the technical basis; engineering keeps integrity authority; quality approves release; finance approves credit, valuation and posting. Digital specialists inherit no release or engineering authority. Record local stop rights too.
B3 acceptance gate The question Input readiness Are identities, measurement bases, cut-offs and source records adequate? Technical performance Evaluated on representative cases, including failures and missing evidence? Workflow readiness Can a responsible person act within the decision window? Authority Are permitted, approval-dependent and excluded actions explicit? Safety and security Have the competent owners accepted the controls for the actual scope? Economics Full cost, counterfactual and downside, without double counting? Capability Can the terminal sustain, support and recover the routine? - THE DOCTRINE
Do not average the gates so that an attractive financial return cancels an unacceptable authority or safety problem. Record each gate, reviewer and unresolved action separately.
- B4 checks units, signs, constrained volume, who receives the benefit and a downside and delay case before the meeting. The board decision working record holds the decision requested, options, economics, consequences, conditions and resolution.
Put the open gates on the summary page, above any financial figure.
20:00 – 25:00
Assemble, review, sign and close
- Assemble Sections 1 to 9 in order, noting in one line wherever a later session changed an earlier one.
- Rehearse Chapter 17's executive review: can the owner explain one recommendation, its evidence, the authorised action and the result; finance one benefit without double counting; the technical team one failure and recovery; staff what became easier and harder? If not, do not scale on a polished presentation.
- Sign and date the plan. Then sit the Final Exam: fifteen questions, twelve from the paper and three rebuilt from your own earlier mistakes, closed book, pass mark 70.
- THE DOCTRINE
Begin with a decision that matters. Establish the evidence and the simpler alternative. Bound the application's authority. Test the response and reconcile the result. Expand when the operating routine earns that decision.
- The book's closing standard: a terminal becomes intelligent when accurate inventory supports dependable commitments, safe movements preserve the product, and completed service becomes collected value. AI prepares choices; leadership supplies purpose, authority and accountability. Neither substitutes for the other.
The game plan is complete when leadership can make an inspectable decision about the next stage.
Key Terms
Terminal Game Plan
Sections 1 to 9 laid across Chapter 17's ninety days, with the A1 charter, B1 areas, B3 gates and a board decision record, signed.
Stop condition
The event, agreed before the pilot starts, that stops or restricts the application, with who may suspend it.
Acceptance gate
One of seven B3 conditions, each recorded with its reviewer and unresolved action, never averaged.
Day-90 decision
Scale, revise, continue within a bounded scope or stop, on reconciled results and finance-approved benefits.
Self-Check — Before You Submit
1. At day 45 of a fictional depot's pilot, results disappoint. The sponsor proposes measuring a different benefit that looks better. What does Chapter 17 require?
- A. Change it quietly; the pilot is still learning
- B. Stop the pilot immediately
- C. If justified, record a revised evaluation with a clear reason
- D. Extend to day 120 under the original definition
C. If justified, record a revised evaluation with a clear reason — The benefit definition is not changed because early results disappoint; a justified change is recorded as a revised evaluation. Stopping at once is premature: day 90 decides on reconciled evidence.
2. A fictional terminal's case shows a strong return, but the authority gate is open. What should the board record?
- A. Each gate separately with reviewer and unresolved action; the return cancels nothing
- B. An average score that passes on economics
- C. Approval now, closing the gate after scaling
- D. A failed economics gate
A. Each gate separately with reviewer and unresolved action; the return cancels nothing — B3 forbids averaging so that a financial return cancels an authority problem. Approving now and fixing later is the tempting compromise, but some requirements are mandatory conditions for continuation.
3. A pilot ran well at a fictional coastal marine terminal. The group wants it at its inland depots next month. What is right?
- A. Roll out; one approval covers the group
- B. Roll out to the largest depot only
- C. Abandon it, since depots differ
- D. Specify the next scope and evidence required; one site's approval is not approval for every archetype
D. Specify the next scope and evidence required; one site's approval is not approval for every archetype — Differences in operating model and consequence can require a new review, so expansion names its scope and evidence. Abandoning it overreacts; the depots may benefit after review.
Today’s Assignment
Assemble Sections 1 to 9 into one signed Terminal Game Plan for your company's executive committee, with the A1 pilot charter, the ninety days, the B1 scorecard areas, the B3 gates, the day-90 decision rule and the board decision working record.
- One-page summary: decision, product, facility, owner and sponsor by role, baseline, stop condition, open gates and the decision you ask for.
- Sections 1 to 9 in order, with one-line corrections where a later session changed an earlier one.
- The A1 pilot charter, every field completed or honestly marked missing, and the ninety days stage by stage with the evidence and decision at each review.
- The B1 areas relevant to your decision with evidence and owner, and each B3 gate separately with reviewer and unresolved action.
- The day-90 rule and the board decision working record: decision requested, options, consequence of delay and accountable owner.
- Sign and date it, naming the commitment you are least confident your company will keep. With no AI in use yet, the plan still stands.
YOUR 90-DAY PLAN
The assembled, signed Terminal Game Plan.
How this is marked — the rubric
| Sections 1 to 9 assembled and consistent, one-page summary, signed, least-certain commitment named | 20 |
| A1 pilot charter complete, with simpler alternative, authority, stop condition and approvals | 20 |
| The ninety days with evidence and decision at each review, and a day-90 rule naming the four outcomes | 20 |
| Relevant B1 areas with evidence and owner; each B3 gate separate with reviewer and unresolved action | 25 |
| Board decision record with decision requested, options, consequence of delay and accountable owner | 15 |
| 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.FINAL EXAM
Your Final Exam is sat in this session. 15 questions, of which 3 are rebuilt from questions you got wrong on the practice papers. Sit it once your capstone is submitted.
Sit the Final Exam →Closed book, no notes, and Coach Ritchie switches himself off while you sit it.Capabilities Covered
Ninety-day sequencingThe pilot charterBoard scorecard and gatesDecision rights by consequenceThe day-90 decisionClosing the loop