30-Session Programme · 25 Minutes Per Day · Free · For leaders in Nigeria’s oil and gas sector
WEEK 5 OF 6 · CONTROLLED AUTONOMY, CAPITAL AND PEOPLE · SESSIONS 21–25 · WORK AT YOUR OWN PACE
Part Five · Controlled AutonomySESSION 21
Agentic AI and the Autonomy Ladder
Why This Matters
Every session so far has been about AI that informs a human decision. This week the book turns to AI that acts. Chapter 14 calls itself the most important chapter for the next three years, and its Leadership Charge says why: leaders will get hurt not by being too cautious, but by letting autonomy outrun their ability to govern it. Agentic AI is coming whether or not your company is ready. Some of it will be genuinely valuable; some will be a liability with an interface. Today you learn to tell them apart, and you draw the line in writing for three uses your business is actually considering.
25-Minute Module
0:00 – 4:00
AI that acts, not AI that informs
- Everything since Session 2 has produced intelligence for a person to act on: a prediction, an optimisation, a detection, a forecast. Agentic AI takes the action. It gathers data, drafts documents, prepares procurements and carries multi-step workflows to completion with limited human intervention.
- The value is real: less manual drag, shorter cycle times, scarce experts freed for judgement rather than process. For a Nigerian operator facing cost pressure and capability constraints, the book calls that leverage real and attractive.
- The danger is the same property: it acts. An agent that drafts a wrong report wastes time. One that takes a wrong action in a procurement, a financial system or an operational-technology environment causes real harm.
- THE DOCTRINE
The responsible posture is neither breathless enthusiasm nor reflexive rejection, but disciplined ambition: capture the value where errors are low-consequence and reversible, and keep a firm human hand where they are not.
The question is not whether to engage, but with what discipline, in what sequence and with what guardrails.
4:00 – 9:00
What an agent actually is
- A plain picture first. A chatbot is an adviser who writes you a note; you decide and you act. An agent is closer to a contractor handed a set of keys: it reads the situation, plans, uses the systems it was given, checks the result and goes again.
- The book's definition: an agent pursues a goal in a loop. It perceives data, documents and system state, reasons and plans, acts by calling tools and systems, and observes the result. Multi-step action through tools is what separates it from a model or a chatbot.
- The governance insight: an agent's power comes from its tools and permissions. One that can only read certain data and draft a document is low-risk. One that can execute transactions, change configurations or touch operational systems is high-risk and must be contained accordingly.
- THE DOCTRINE
The safety of agentic AI is not primarily a property of the model. It is a property of the guardrails that wrap the agent: bounded scope, approval gates, limits and audit.
Ask what it can touch before you ask how clever it is.
9:00 – 15:00
A dial, not a switch
- Chapter 14 is the Autonomy Ladder made concrete. Autonomy is not binary; it is a spectrum climbed gradually, workflow by workflow.
Level What the AI does Where the human sits Copilot Drafts, searches, summarises, analyses In full control and does the acting; most of today's value is here Supervised agent Executes steps itself Approves each consequential action: the agent proposes, the human authorises Bounded autonomy Acts within strict, predefined limits on lower-risk workflows Reviews outcomes rather than each step Full autonomy over OT and safety decisions Would act on high-consequence physical or safety decisions with no human in command Permanently out of bounds: a deliberate boundary, not a temporary limitation - Placement turns on two questions: what does an error cost, and how hard is it to reverse? A tender draft, a report, a compliance pack or a data summary is low-consequence and easily reversed. A production setpoint change, a well or facility intervention or a safety-system action is not. The book's Figure 2 is illustrative, and most of your business sits between those poles.
- IN THE FIELD · THE WORKED CASE
- Field. A fictional drilling and well services company is offered an agent that would change setpoints on its rig-site systems without a person approving each change. High consequence, hard to reverse: it sits at the ceiling, and the answer is no.
IN THE BOARDROOM · THE PARALLEL- Boardroom. Its bid team is offered a supervised agent that assembles compliance matrices and drafts technical responses, with a manager approving before submission. Same map, same two questions, opposite answer.
- THE DOCTRINE
Treat autonomy as a dial you turn gradually, workflow by workflow, never a switch you flip, and hold a permanent human-in-command ceiling over every operational-technology and safety decision.
If nobody can say how an error would be reversed, you have already answered the autonomy question.
15:00 – 21:00
Guardrails, reliability and the realistic horizon
Guardrail stack, base upward What it does Bounded scope Access only to the systems and data explicitly permitted Evaluation and testing Measured against defined cases before deployment, monitored in production Audit and observability Every action logged, traceable and reviewable Approval gates Explicit human sign-off before consequential actions execute Kill switches and limits Hard stops, plus spend, scope and rate limits Human oversight A person can always inspect, override and stop the agent - The book's image: the model is the engine; the guardrails are the brakes, the steering and the seatbelts. An agent without the stack should never reach production.
- Why judgement stays human. Today's AI can be confidently wrong, can be manipulated through its inputs, and does not possess judgement in the human sense. These are reasons to bound agents, not to avoid them.
- The realistic trajectory, labelled indicative phasing, not a forecast: copilots now; supervised agents in back-office and commercial work over the next year to two years; bounded autonomy on lower-risk operational workflows over the following two to three years. Safety-critical operational autonomy is not on this horizon.
- The hype filter. A vendor promising autonomous operation of a hazardous facility in the near term is selling a risk. One showing a supervised agent that compresses a tender cycle under human approval is offering present value.
Three questions cut through any agentic pitch: where on the map, what contains it, and who is in command.
21:00 – 25:00
Before you close the laptop
- List three candidate AI uses in your company, at least one back-office and one operational.
- For each, write the consequence of an error and how it would be reversed, then place it on the dial.
- That is Section 21 of your plan: three uses placed on the Autonomy Ladder, with guardrails.
Draw the line yourself, in writing, before an incident draws it for you.
Key Terms
AI agent
A system that pursues a goal in a loop: it perceives, plans, acts by calling tools and systems, and observes the result.
Supervised agent
An agent that executes steps itself but needs human approval for each consequential action.
Bounded autonomy
An agent acting within strict, predefined limits on lower-risk workflows, with humans reviewing outcomes rather than each step.
Human-in-command ceiling
The permanent boundary above which no agent acts on operational-technology or safety-critical decisions without a human in command.
Self-Check — Before You Submit
1. A fictional marginal field operator's operations director wants an agent to change a production setpoint automatically, because it performed well for three months assembling reports. What is the right decision?
- A. Approve, because the agent has demonstrated reliability
- B. Approve, provided a kill switch and an audit trail are in place
- C. Refuse: a setpoint change is high-consequence and hard to reverse, so a human must decide
- D. Defer until the vendor shows improved model accuracy
C. Refuse: a setpoint change is high-consequence and hard to reverse, so a human must decide — The book holds a permanent human-in-command ceiling over operational and safety decisions. Guardrails are the tempting answer but do not lift the ceiling, and nor does better accuracy: the ceiling reflects what accountability is, not how mature the technology is.
2. Which two dimensions does the book use to decide where agentic autonomy is safe?
- A. The consequence of an error and how hard it is to reverse
- B. The accuracy of the model and the maturity of the vendor
- C. The cost of the agent and its speed
- D. The volume of data and the frequency of the task
A. The consequence of an error and how hard it is to reverse — Consequence and reversibility place each workflow on the map. Model accuracy tempts, but the book says safety is engineered through guardrails and placement, not assumed from capability.
3. According to the book, what is the safety of agentic AI primarily a property of?
- A. The sophistication of the model
- B. The guardrails that wrap the agent
- C. The vendor's certification
- D. The volume of training data
B. The guardrails that wrap the agent — The book says safety is not primarily a property of the model but of the guardrails. A sophisticated model with broad permissions and no containment is still an uncontained risk.
Today’s Assignment
Write Section 21 of your plan: three of your company's candidate AI uses placed on the Autonomy Ladder, each with where it must stay human-gated and the guardrails it needs.
- Name three candidate uses. If your company has no AI in use yet, that is normal: use a proposal in front of you, a vendor's pitch, or the loss you named in Section 1.
- For each, state the consequence of an error and how hard it would be to reverse, in operational terms.
- Place each on the spectrum and state the highest rung it may reach. Wherever a use touches an operational-technology or safety decision, write the human-in-command ceiling explicitly.
- For each, describe the guardrail stack in practice, naming by role who approves its consequential actions and who can stop it.
- The honest part: where would your company today be tempted to grant autonomy faster than it could govern it? Aim for 400 to 900 words.
YOUR 90-DAY PLAN
Section 21 — Three uses placed on the autonomy ladder, with guardrails.
How this is marked — the rubric
| Places three named uses on the spectrum using consequence and reversibility, with a highest permitted rung for each | 30 |
| States the human-in-command ceiling explicitly wherever a use touches operational or safety decisions | 25 |
| Specifies the guardrail stack for each use, with approvers named by role, not department | 25 |
| Names one honest place where autonomy would outrun governance in the company today | 20 |
| Total · pass mark 70 | 100 |
Submitting session 21
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: 21, 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
Agent versus chatbotAutonomy as a dialConsequence and reversibilityGuardrail stack designFiltering agentic hype
Part Five · Controlled AutonomySESSION 22
Where Agents Pay First, and the Questions Every Board Should Ask
Why This Matters
Last session drew the map. This one decides where to spend. Chapter 14 is specific about where agents pay first in an energy company, and it is not the wellhead. It is the tender room, the compliance pack, the procurement chase and the management report: document-heavy, multi-step work, low in physical consequence. For service companies, faster and cleaner bids are, in the book's words, a survival issue. Today you choose one such workflow in your business, design a 90-day pilot that keeps the agent contained, and answer the seven questions every board should ask. A board that cannot get answers is not governing its agents. It is hoping.
25-Minute Module
0:00 – 5:00
Three sources of value, four honest costs
- Agentic value comes from removing manual drag from document-heavy, repetitive work; compressing cycle time, so a bid or report takes hours rather than days; and consistency and completeness, because a well-governed agent does not skip a compliance item under deadline pressure.
- The honest case includes four costs: the technology and the engine beneath it; the guardrails, evaluation and audit; developing the supervising workforce; and the risk cost of errors, bounded by keeping agents in the safe space.
- THE DOCTRINE
The economic case for agentic AI is strongest in high-volume, document-heavy, low-consequence, reversible work, where the drag removed is large and the risk is small.
- Refusing to engage is no safer. Competitors using supervised agents will become faster and cheaper in tendering, compliance, reporting and procurement, and the book warns that the gap compounds until it is structural.
Price the supervisors and the guardrails into the case, or the case is fiction.
5:00 – 10:00
Where agents pay first
Workflow What the supervised agent does What the human keeps Tendering and bid assembly Compliance matrices, draft technical responses, completeness checks, the Nigerian Content documentation Approval before submission Compliance and regulatory reporting Gathers data, populates templates, checks requirements, prepares the audit-ready pack Validation Procurement support and expediting Chases status, compares options, prepares requisitions The consequential decisions Reporting and analysis assembly Summarises data into a coherent narrative The conclusions Knowledge retrieval and synthesis Draws on governed procedures, lessons, records and contracts Verification Back-office process automation Rules-based finance, HR and administration tasks Approval; the lowest-risk, fastest-payback applications - The book ties compliance assembly to the measurement-based reporting regimes now emerging, citing NUPRC's directive on standardised templates and measurement-based methane and GHG reporting. Where procurement delay is a production threat, faster procurement support is direct operational value.
- The pattern: high value, low physical consequence, reversible errors, a human in the approval loop. This is also where your company builds the muscle of governing agents.
The board does not need to chase operational autonomy. It needs to capture the safe value first.
10:00 – 15:00
Foundation, containment and the 90-day pilot
- Agents inherit everything beneath them. Ungoverned data gives confident, wrong answers; broad permissions allow wrong actions. Governed data, access controls, lineage and a model-and-agent inventory are now prerequisites, not good practice.
- The architecture is containment: defined tools and permissions, logged actions, gated consequential steps, capped scope, spend and rate, governed data only, and strict OT/IT segmentation. The agent must never have a path to a high-consequence action that bypasses the guardrails, and gets the minimum scope its task requires.
- The pilot. Days 1 to 30: select a high-value, low-consequence, reversible workflow; assess its data and permissions; design guardrails, approval points and the success metric. Days 31 to 60: build within containment, with an evaluation harness, audit and approval gates. Days 61 to 90: run under supervision on real work, measure value and errors, study failure modes, refine.
- Each pilot needs a business owner, a governance owner, a baseline, a measured outcome and an executive sponsor who understands that the discipline, not the autonomy, is the achievement.
- THE DOCTRINE
An agentic pilot that creates value but cannot be governed is not a success; it is a risk that happened to pay off this time.
The deliverable is a governed supervised agent, not an autonomous system.
15:00 – 21:00
People, the risk surface and the board's seven questions
- The workforce shift is from doing tasks to supervising agents: judging output, catching characteristic failures, knowing when to intervene. Without it, people either fail to trust agents or trust them too much.
- The new risk surface has two sharp edges: manipulation through untrusted inputs, and OT, where no agent may have an ungoverned path to systems with physical consequences. The chapter's sources name the NIST AI Risk Management Framework, ISO/IEC 42001 and NIST SP 800-82.
- Sovereign capability. The book aligns indigenous capability to build, supervise, secure and govern agents with the National AI Strategy, and argues NCDMB's mandate should explicitly anticipate it. Sponsorship cannot be delegated to IT: where autonomy is allowed is a business and risk decision.
Board question What a credible answer shows Inventory What agents we run, what each can touch, who owns each Ceiling Where a human must decide, held absolutely for operational and safety decisions Guardrail What bounds each agent's scope, gates its actions, logs it and can stop it Data and permission What each agent inherited, and whether it is the minimum required Value and control Measured value captured within proven controls Capability People able to supervise and govern the agents Preparedness Foundations built ahead of need - Board metrics cover value against a baseline and discipline: guardrail-stack coverage, approval adherence, audit coverage, inventory completeness, and agent errors and near-misses.
- THE DOCTRINE
A high productivity gain with low guardrail coverage is a warning, not a success. Value without control is risk.
- IN THE FIELD · THE WORKED CASE
- Field. In a fictional marine and logistics company, an expediting clerk supervises an agent chasing spare-parts status, checking each requisition and logging every correction.
IN THE BOARDROOM · THE PARALLEL- Boardroom. Its board sees the approval-adherence rate and the corrections log beside the cycle-time gain. Same pilot from each end: the clerk is the guardrail, and the board can see her working.
Refuse to be satisfied with a reassuring non-answer to any of the seven.
21:00 – 25:00
Before you close the laptop
- Choose one safe-space workflow in your company and write its value case in one line, against a baseline.
- Test it against the book's seven failure modes: autonomy without guardrails, ignoring the ceiling, over-trust, ungoverned data and permissions, neglecting the supervising workforce, chasing operational autonomy for its own sake, and reflexive rejection.
- That is Section 22 of your plan: our first bounded agentic pilot and the board's questions answered.
Autonomy is earned. Make your company earn every rung.
Key Terms
Containment architecture
The engineered environment around an agent: defined permissions, logged actions, approval gates, hard limits and human override, with no path to a high-consequence action that bypasses them.
Agent inventory
The register of every agent the company runs, what each can touch, who owns it and how it performs.
Evaluation harness
The defined test cases an agent is measured against before and during production, so its behaviour is known rather than assumed.
Graduated trust
The principle that an agent earns expanded scope only through demonstrated reliability under supervision, never by default.
Self-Check — Before You Submit
1. A fictional EPC company's agentic pilot on tender assembly has cut bid preparation time against its baseline. But audit logs cover only some of the agent's actions, and no governance owner was named. Should it scale?
- A. Yes, because the value against the baseline is proven
- B. Yes, to one more workflow, fixing the gaps along the way
- C. No, stop the programme: agents are too risky for tendering
- D. Not yet: complete the audit coverage, name the governance owner and show the value within proven controls
D. Not yet: complete the audit coverage, name the governance owner and show the value within proven controls — A pilot that creates value but cannot be governed is a risk that happened to pay off. Scaling on value alone is the tempting error; stopping altogether is reflexive rejection in a workflow the book places in the safe space.
2. Where does the book say agentic AI pays first in an energy company?
- A. Autonomous optimisation of production setpoints
- B. Document-heavy, low-consequence work such as tendering, compliance reporting, procurement support and reporting
- C. Autonomous well-intervention planning
- D. Safety-system monitoring with automatic override
B. Document-heavy, low-consequence work such as tendering, compliance reporting, procurement support and reporting — The safe space is high value, low physical consequence and reversible errors with a human approving. The operational options tempt because the stakes look larger, but they sit at or near the ceiling.
3. Why does the book treat the data and permission foundation as a prerequisite for agentic AI?
- A. Because an agent inherits every weakness of the data and permissions beneath it
- B. Because regulators require it before any agent is deployed
- C. Because agents cannot run without a cloud platform
- D. Because a foundation makes the model more accurate
A. Because an agent inherits every weakness of the data and permissions beneath it — The agent amplifies whatever is beneath it. The regulator option tempts, but the book makes no such claim; its argument is inheritance, not compliance.
Today’s Assignment
Write Section 22 of your plan: a bounded 90-day agentic pilot in low-consequence work in your company, with the seven board questions answered for it.
- Choose one high-value, low-consequence, reversible workflow in your company. If your company has no AI in use yet, choose the workflow where manual drag hurts most. State why an error there is cheap to catch and reverse.
- Write the value case against a baseline: today's cycle time, effort or error rate, and the outcome you expect.
- Set out the three 30-day phases, the data and permissions the agent would inherit, its minimum scope, and every approval point by role. Name the business owner, governance owner and sponsor.
- Answer the seven board questions. Where your company cannot answer one today, say so and state what it would take.
- The honest part: which failure mode is your company most likely to commit? Aim for 400 to 900 words.
YOUR 90-DAY PLAN
Section 22 — Our first bounded agentic pilot and the board's questions answered.
How this is marked — the rubric
| Selects a genuinely low-consequence, reversible workflow with a baselined value case | 25 |
| Designs the three phases with minimum scope, approval points by role and named owners and sponsor | 30 |
| Answers all seven board questions specifically, admitting those the company cannot yet answer | 30 |
| Names the failure mode the company is most exposed to and how the pilot guards against it | 15 |
| Total · pass mark 70 | 100 |
Submitting session 22
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: 22, 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
Agentic value caseContainment designAgentic pilot designSupervising agentsBoard oversight of agents
Part Five · Capital and PeopleSESSION 23
The $60 Billion Question
Why This Matters
Every capability in this programme needs money. Chapter 15 turns to what the others assume: capital. NNPC has framed a roughly $60 billion energy-investment ambition by the end of the decade, and every dollar of it competes globally. The book's claim is uncomfortable and specific: capital now prices credibility and emissions, not only reserves, so whether your company's numbers can be defended is becoming a question of what money costs you. The same logic applies wherever someone decides whether to trust your numbers, whether a lender, a client or a joint-venture partner. Today you trace how your company's data credibility reaches its access to capital and its cost.
25-Minute Module
0:00 – 5:00
The ambition is a capital question
- Every target in Nigeria's energy strategy, from higher oil production and the Decade of Gas to reserves growth and decarbonisation, rests on capital: wells, facilities, gas infrastructure, pipelines, processing, and the digital and human capability to run them.
- The anchor: NNPC has framed an energy-investment ambition of roughly $60 billion by the end of the decade alongside its production targets, as the chapter's sources record from Reuters. The book's Figure 1, showing where that investment must flow, is labelled an illustrative split.
- THE DOCTRINE
The binding constraint on the national ambition is not whether the hydrocarbons exist, because they do, but whether the capital can be mobilised to develop them, and on what terms.
- The honest framing: AI does not create capital and does not substitute for a sound business. It shapes whether your company can attract capital, and on what terms, by building the credibility capital now prices.
If your AI case never mentions financing, you have presented half of it.
5:00 – 10:00
Two new lenses: credibility and emissions
- Capital has added two lenses to reserves and returns. Credibility: whether claims about production, reserves, costs and forecasts can be defended with evidence rather than asserted with optimism. Emissions: whether you measure, report, reduce and can prove your footprint, particularly flaring and methane.
- A credible operator with measured, improving emissions is a lower-risk proposition, and lower risk means cheaper, more available capital. The book calls this increasingly a hard, priced effect, not a soft, reputational one.
- In a capital-constrained environment, that difference can decide whether a project proceeds or stalls.
- IN THE FIELD · THE WORKED CASE
- Field. A fictional marginal field operator reports production from estimates nobody has reconciled, and characterises its flaring loosely.
IN THE BOARDROOM · THE PARALLEL- Boardroom. Its finance director meets a lender, who asks how the forecast was built and how emissions were measured. Optimism without lineage is priced as risk. One chain: the terms in the boardroom were set by the data in the field.
Capital will not take your word for it, and it should not.
10:00 – 16:00
The credibility flywheel and the assurance chain
- The book describes a credibility flywheel: a self-reinforcing loop that compounds once it turns.
Link What happens 1. Trusted data and governance The foundation 2. Defensible reporting and disclosure Production, reserves and emissions figures that can be traced, reproduced and defended 3. Investor and lender confidence Capital trusts what it can verify 4. Lower cost of capital Lower-risk operators are funded on better terms 5. Investment and growth Cheaper capital funds more investment 6. Reinvest in capability Growth reinvested in capability strengthens the data and governance at the base - AI turns the flywheel at its base. Without data quality, lineage and analytics, the reporting is not defensible and the flywheel does not turn.
- ESG-grade reporting is an assurance chain: measure, assure, disclose, finance. As NUPRC moves operators toward measurement-based methane and GHG reporting, the question is whether the report can be traced, validated, reproduced and defended. AI validates emissions data, detects implausible readings, reconciles measurements and manages the lineage.
- THE DOCTRINE
The data that makes an emissions number defensible is the same data that makes an operational decision sound. Build the foundation once and let it pay in both performance and financing.
An assurance chain is engineered for traceability, not assembled at deadline.
16:00 – 21:00
The worked logic and the gradient
- The book's worked logic from a data investment to a lower cost of capital is a chain of reasoning, not a calculation. The book attaches no figures to it, and neither will I; its gradient chart is labelled conceptual and illustrative.
The foundation produces Over time it builds Financing outcome Accurate production reporting A track record of forecasts that hold up Confidence of lenders whose covenants depend on those forecasts Traceable emissions measurements ESG disclosures that withstand scrutiny Access to climate-conscious capital that conditions on such credibility Quantified reserves uncertainty A reserves statement investors can trust Support for the company's valuation - Together these move the operator down the cost-of-capital gradient, from the opaque end (estimates, weak data, loosely characterised emissions) to the credible end (trusted data, defensible reporting, quantified reserves, evidenced ESG).
- THE DOCTRINE
The data investment pays through a channel that never appears on the operational P&L: the terms on which the entire business is financed. The financing case for the foundation can be larger than the operational case.
- AI also sharpens the capital decision itself where the most money meets the most uncertainty: project sanction and sequencing, portfolio allocation, reserves and forecast credibility, asset valuation. A modest improvement in a billion-dollar sanction decision, the book says, is worth far more than a large one in a routine operating decision.
A CFO who can trace the whole chain can fund what a purely operational framing would starve.
21:00 – 25:00
Before you close the laptop
- Write down which link of your company's flywheel fails first today.
- Name the one number a lender, partner or client would most want to verify, and whether your company could trace it to source this week.
- That is Section 23 of your plan: our credibility flywheel and what investors or lenders can verify today.
- Pop Quiz 5 opens now: five questions, closed book, covering Sessions 21 to 23.
Investors fund the operator they can measure.
Key Terms
Credibility (as capital prices it)
The degree to which a company's claims about production, reserves, costs and forecasts can be defended with evidence rather than asserted with optimism.
Credibility flywheel
The compounding loop in which trusted data produces defensible numbers, which build confidence, which lowers the cost of capital, which funds growth that strengthens the data.
Assurance chain
ESG-grade reporting as four linked stages (measure, assure, disclose, finance), traceable end to end so every figure can be reconstructed on demand.
Cost-of-capital gradient
The book's conceptual picture of the cost of capital as a slope, not a fixed number, running from the opaque operator to the credible one.
Self-Check — Before You Submit
1. In the credibility flywheel, what does defensible reporting and disclosure directly build?
- A. A lower operating cost per barrel
- B. Higher production
- C. Investor and lender confidence
- D. Regulatory approval of reserves
C. Investor and lender confidence — The flywheel runs from trusted data to defensible reporting to confidence, then to a lower cost of capital. Lower operating cost tempts, but the flywheel is a financing mechanism, not an operational one.
2. A fictional upstream operator will seek finance in six months. Its CFO proposes compiling a polished emissions disclosure from existing spreadsheets just before the meeting. What does the book point to instead?
- A. Engage a communications adviser to present the numbers well
- B. Engineer emissions reporting as an assurance chain now, with validation, reconciliation and lineage
- C. Leave emissions out until the financing closes
- D. Publish the spreadsheet figures as they are
B. Engineer emissions reporting as an assurance chain now, with validation, reconciliation and lineage — The book says ESG reporting is an assurance chain engineered for traceability, not a disclosure assembled at deadline. Better presentation tempts, but it treats credibility as a communications task, and capital sees through presentation to substance.
3. How does the book present its worked logic from a data investment to a lower cost of capital?
- A. As a calculation of the saving per naira invested
- B. As proof that AI creates capital
- C. As a forecast of the national investment ambition
- D. As a chain of qualitative links that together move the operator down the cost-of-capital gradient
D. As a chain of qualitative links that together move the operator down the cost-of-capital gradient — The book traces the chain without attaching figures, and labels its gradient chart conceptual and illustrative. A precise calculation is the tempting misreading, and the book says plainly that AI does not create capital.
Today’s Assignment
Write Section 23 of your plan: your company's credibility flywheel, and what investors, lenders or clients can verify about it today.
- Name who judges your company's numbers: lenders, partners, offtakers or, for a contractor, the clients and banks who weigh your delivery record.
- Walk the six links of the flywheel for your company, stating honestly whether each turns today and what evidence shows it.
- Apply the worked logic (forecasts, emissions, reserves) to your business. If your company holds no reserves, name the numbers your clients or lenders scrutinise instead.
- Name the number capital or a client would most want to verify, its owner by role, and whether it could be traced to source on demand. Do not invent figures; where you cannot quantify a link, describe it, as the book does.
- The honest part: where does your company present optimism as evidence today? Aim for 400 to 900 words.
YOUR 90-DAY PLAN
Section 23 — Our credibility flywheel and what investors or lenders can verify today.
How this is marked — the rubric
| Identifies who judges the company's numbers and walks all six flywheel links with evidence | 30 |
| Applies the worked logic to the company's own forecasts, emissions or reserves, or their service-business equivalents | 25 |
| Names the most scrutinised number, its owner by role, and whether it is traceable today | 25 |
| States honestly where the company presents optimism as evidence, without invented figures | 20 |
| Total · pass mark 70 | 100 |
Submitting session 23
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: 23, 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 5 — Autonomy and Capital 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 5 →Closed book, no notes, and Coach Ritchie switches himself off while you sit it.Capabilities Covered
Capital as the binding constraintCredibility and emissions lensesCredibility flywheelESG assurance chainCost-of-capital framing
Part Five · Capital and PeopleSESSION 24
The Financing Case and the Integrity of the Numbers
Why This Matters
Last session showed why credibility lowers the cost of capital. This one is about the work: which numbers, which disclosures, whose name sits against each, and how your company proves it can produce a defensible figure on demand. Chapter 15 is direct that financial credibility cannot be assembled at the moment capital asks for it. It is built into the operating model in the months and years before, and tested suddenly: at a loan, a raise, a rating review or a partner negotiation. It is also a CFO and board capability, not an IT matter. Today you list the numbers your business must defend, assign owners, and design the capital-credibility pilot.
25-Minute Module
0:00 – 5:00
The methods, and where the case is won
Method What it gives capital Credible production and reserves forecasting A range with confidence rather than a single optimistic line; a track record of delivery Economic scenario modelling and stress-testing The business tested under price, FX, cost and emissions assumptions before commitment Portfolio optimisation and capital allocation Projects ranked by risk-adjusted value: analysis rather than politics Reserves credibility through quantified uncertainty Probabilistically honest reserves, more defensible than deterministic optimism ESG and emissions assurance Measurement-based, traceable emissions numbers Auditable capital-decision support Sanction, allocation and valuation against quantified ranges with stated assumptions - The case is won in six places: project sanction, reserves disclosure, M&A and asset valuation, lender and covenant reporting, ESG disclosure to capital, and partner confidence in joint ventures. The book calls defensible lender reporting a renewable financing asset.
- THE DOCTRINE
The data foundation is the financing foundation. An operator cannot have credible numbers for capital while having untrustworthy data for operations; they are the same data.
- So the data investments the operational side proposes are also financing investments. In the book's words, they pay twice: in performance and in capital.
Stated assumptions are what turn a range into an auditable decision.
5:00 – 10:00
The problem, seen from the capital end
- From the capital end, the problem is the gap between the data the company has and the credibility capital requires. It raises the hurdle every investment must clear and shrinks the set of viable projects: in the book's words, a first-order destroyer of value.
- The architecture of financial credibility is the chain of trust seen from the capital end: governed data, validation and lineage, analytics, then the disclosures and decisions capital assesses. A weak link anywhere breaks credibility at the point capital is watching.
- The transparency dividend. Transparency is an asset only if what it reveals is credible; with weak data it is a threat. The book notes that the sector's numbers have historically been a concern for some investors, so demonstrable transparency is a powerful signal. Its sources cite measurement-based reporting and the Extractive Industries Transparency Initiative's Nigeria profile.
- Capital assesses the country through the companies seeking it, so inaction at company level aggregates into a national financing constraint.
- IN THE FIELD · THE WORKED CASE
- Field. A fictional EPC and construction company reports project progress from site returns each project manager compiles differently. Nobody can reconstruct last quarter's completion figure.
IN THE BOARDROOM · THE PARALLEL- Boardroom. It pursues a large contract whose client and bank will scrutinise its delivery record. The figures are presented confidently and cannot be traced: the book's exposure dressed as confidence, in a contractor's form.
Transparency pays only when what it reveals can withstand scrutiny.
10:00 – 15:00
Human judgement, governance and ownership
- Capital decisions remain human decisions, informed by AI. AI quantifies uncertainty, models scenarios, ranks options and assures data; which risks to accept and which projects to fund stays with experienced finance and executive leadership, owned by the board. Confident wrongness and hidden assumptions are exactly the failures that destroy value here.
- THE DOCTRINE
Govern every capital-facing number as you govern a financial statement: clear ownership, versioning, validation and the lineage that lets it be defended.
- This is a CFO and board capability. Can we defend our reserves? Is our emissions reporting traceable? Do our forecasts hold up? Is our cost of capital improving? A board that does not connect these to the data foundation is governing its financeability without understanding what produces it.
- Sponsorship usually sits with the CFO, working with the CEO and the board. A CFO who sees AI only as an operational cost underfunds the capabilities that lower the cost of capital.
- Sovereign financial credibility. The book places financial-analytical, data-governance and ESG-assurance capability among the high-value capabilities the Local Content 2.0 agenda exists to develop at home, citing NCDMB, and aligns it with the National AI Strategy. Credibility produced by foreign-controlled analysis is a dependency at your most strategic interface.
If the CFO thinks the data programme belongs to IT, the financing case has no sponsor.
15:00 – 21:00
Sequence, pilot, rhythm and scoreboard
- Sequence. Trusted data first; then defensible reporting, quantified reserves, assured emissions and auditable capital decisions in order of financing value. A portfolio-allocation model built before the foundation produces numbers scrutiny will not survive, which the book calls worse than no programme.
Days Phase Output 1 to 30 Diagnose The decision or disclosure to strengthen, its data assessed, the credibility gap defined 31 to 60 Build The governance, lineage and analytics that make the chosen number defensible 61 to 90 Deliver For example a reserves estimate with quantified uncertainty, an emissions report with full lineage, or a stress-tested sanction analysis - Each pilot needs a finance owner, a data owner, a baseline and an executive sponsor, typically the CFO.
- THE DOCTRINE
A financing analysis that cannot survive scrutiny is not credibility; it is exposure dressed as confidence.
- Rhythm. The regular review checks the foundation: trusted data, forecasts holding against reality, traceable emissions. The periodic review checks the financing case: cost-of-capital trend, lender relationships, ESG assessment.
- Scoreboard, each against a baseline: capex efficiency, forecast credibility against delivery, reserves defensibility, ESG disclosure quality, emissions-data traceability, capital-decision cycle time, investor and lender confidence, relative cost of capital. The book's scoreboard status is illustrative. Capital prices the portfolio, not the best asset in it.
Credibility is built continuously and tested suddenly.
21:00 – 25:00
Before you close the laptop
- List the numbers your company must defend, and the disclosure or decision each feeds.
- Check your plan against the book's six failure modes: assembling credibility at financing time, presenting optimism as forecast, treating ESG as disclosure rather than assurance, governing capital-facing models loosely, framing AI only as an operational cost, and neglecting human judgement in capital decisions.
- That is Section 24 of your plan: the numbers we must defend, their owners, and our capital-credibility pilot.
The cost of capital is earned in the field, assured in the data and realised in the boardroom.
Key Terms
Quantified uncertainty
Expressing a forecast, reserves estimate or capital case as an honest range with stated assumptions, not a single optimistic number.
Stress-testing
Running the business model under adverse price, FX, cost and emissions scenarios before committing capital.
Capital-facing numbers
The reserves, forecasts, emissions figures and analyses that lenders, investors and partners rely on, governed to the standard of a financial statement.
Transparency dividend
The advantage of a company whose numbers are trustworthy enough that opening its books builds confidence rather than exposing weakness.
Self-Check — Before You Submit
1. A fictional gas company's board wants to begin its capital-credibility programme with a portfolio-allocation model before a fundraise. Its emissions data has never been validated. What do you advise?
- A. A 90-day pilot making one high-value disclosure, such as the emissions reporting, defensible, with a finance owner, data owner, baseline and the CFO as sponsor
- B. Build the portfolio model first, as the most sophisticated capability
- C. Delay all credibility work until after the fundraise
- D. Present an external ESG rating in place of validated data
A. A 90-day pilot making one high-value disclosure, such as the emissions reporting, defensible, with a finance owner, data owner, baseline and the CFO as sponsor — The book puts trusted data first, built around the highest-value disclosures such as emissions reporting. The sophisticated model is the trap: built before the foundation, it produces numbers scrutiny will not survive.
2. Which of these is an activity metric rather than a financing outcome?
- A. Production-forecast credibility against subsequent delivery
- B. Relative cost of capital
- C. The number of reports and models produced
- D. Investor and lender confidence
C. The number of reports and models produced — The book says reports produced and models built are activity, not value. Forecast credibility can look like a process measure, but it tests whether forecasts held up against delivery.
3. Who does the book say should make a company's capital decisions?
- A. The AI model, because it removes bias
- B. The data team, because it owns the models
- C. The vendor that built the tool
- D. Experienced finance and executive leadership, owned by the board and informed by AI
D. Experienced finance and executive leadership, owned by the board and informed by AI — Capital decisions stay human, informed by AI rather than made by it. Letting the model decide tempts, but confident wrongness and hidden assumptions are exactly what destroy value in capital decisions.
Today’s Assignment
Write Section 24 of your plan: the numbers and disclosures your company must be able to defend, who owns each, and your 90-day capital-credibility pilot.
- List five to eight numbers or disclosures your company must defend to lenders, partners, investors, clients or regulators, and the decision or relationship each feeds.
- Name an owner for each, by role, and state whether the number can be traced to source, reproduced and defended today.
- Choose one for the pilot and set out the diagnose, build and deliver phases with a finance owner, data owner, baseline and sponsor.
- State how that number will be reviewed in the regular and periodic rhythm.
- The honest part: which of the six failure modes does your company commit today? Aim for 400 to 900 words.
YOUR 90-DAY PLAN
Section 24 — The numbers we must defend, their owners, and our capital-credibility pilot.
How this is marked — the rubric
| Lists the numbers the company must defend, each linked to the decision or relationship it feeds | 25 |
| Assigns an owner by role to each number and assesses its traceability honestly | 25 |
| Designs a three-phase pilot on one number with finance owner, data owner, baseline and sponsor | 30 |
| Names the failure mode the company commits today and the review rhythm that will prevent it | 20 |
| Total · pass mark 70 | 100 |
Submitting session 24
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: 24, 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
Financing-case methodsDefensible disclosuresGoverning capital-facing numbersCapital-credibility pilotCFO and board sponsorship
Part Five · Capital and PeopleSESSION 25
Local Content 2.0: Owning the Intelligence Layer
Why This Matters
The first generation of local content delivered jobs, fabrication, logistics and services. Chapter 16 argues the next must deliver something harder: ownership of the intelligence layer. Every capability in this programme, from predictive operations and twins to optimisation and surveillance, depends on people who can build, run and govern it. If those people sit with a vendor, so does the capability. This is not a training-department topic. It is whether your company's own people can interpret what the asset is telling them, govern the models you depend on, and reduce your dependence on vendors for routine intelligence. Today you map, role by role, the capability your company needs.
25-Minute Module
0:00 – 5:00
From steel and manpower to the intelligence layer
- Chapter 16 sharpens the book's test for any AI programme: does it change field decisions, reduce delay, protect production and strengthen accountability, and are the people who run it Nigerian?
- The production ambitions you met earlier (NNPC's 2 million barrels per day by 2027 and 3 million by 2030, as Reuters reported, and NUPRC's 2025 target of at least 2.1 million barrels per day) are, in the book's words, execution tests, and execution at that scale needs a workforce that can operate modern digital tools.
- NUPRC describes 323 developed fields, 265 production processing stations, 31 export terminals and 5,284 kilometres of pipelines. Chapter 16's point is that every interface in that system is run by people whose digital capability now determines performance.
- Two of the book's small handoff failures are pure capability: a model nobody understands is quietly ignored, and a vendor is called for a task the team should be able to do itself.
- THE DOCTRINE
Treat AI capability as an operating discipline, not a training-department line item. The real value is intelligence converted into disciplined execution by people who own it.
An advanced platform your engineers cannot interpret produces dashboards, not decisions.
5:00 – 10:00
The economic problem is a capability gap
- Most companies have more data than they can use. The problem is data that is scattered, late or unvalidated, and people not equipped to turn it into decisions. That gap between what the field knows and what the workforce can act upon becomes cost, downtime, deferment, compliance exposure and dependence on outsiders.
- The symptoms: digital skills gaps, vendor dependency, weak AI literacy, limited data engineering capacity, poor integration between technical and digital teams, and insufficient executive understanding of AI operating models, often managed, the book says, through outsourcing, one-off training and hope.
- THE DOCTRINE
Every task outsourced that could have been done in-house is margin leaving the country. Every model nobody local can maintain is a dependency and a risk.
- The book's tests of whether capability work is strategic: does it reduce vendor spend on routine work, improve field decisions, strengthen compliance and governance, build models the company owns, and create career pathways that retain talent?
- Traceability is a capability before it is a technology. Your people must be able to say where data came from, who owns it, when it was captured, whether it was validated and how it links to a decision. As NUPRC moves operators toward measurement-based methane and GHG reporting, credible MRV needs exactly those people.
Experience with the previous generation of tools does not transfer automatically to this one.
10:00 – 15:00
The maturity curve and the six programmes
- Technical staff used to operate equipment and interpret reports. Now they must also interrogate models, validate machine recommendations, structure data and govern automated decisions.
- The maturity curve: awareness of what AI can and cannot do, where the book says most operators honestly sit today; literacy, using AI tools safely; application, AI embedded in real field workflows; and ownership, Nigerian teams building, governing and improving their own models rather than renting them.
Programme What it builds AI literacy Shared understanding from field to boardroom; the cheapest, highest-leverage investment Role-based capability maps Which skills each role needs, at awareness, working, proficient or expert level; the most important design decision Data engineering academies Pipelines, definitions and quality controls; the scarcest and most outsourced capability in most Nigerian operators Prompt engineering for operations Framing questions, checking answers against ground truth, spotting confident wrongness Field analytics training Turning production, pressure, maintenance and inspection data into decisions on the company's own assets AI governance training Validity, reliability, safety, security, accountability, transparency and model risk, backed by the NIST AI RMF and ISO/IEC 42001 - THE DOCTRINE
A training programme that does not change how someone does their job is a certificate, not a capability.
- The book names NCDMB's mandate for capacity development, R&D and industrialisation as the natural institutional home for this agenda.
The point is not to make everyone a data scientist. It is to give every role the right depth.
15:00 – 21:00
Who needs what: capability by role
Role Capability profile in the book Engineers The deepest field-analytics capability: interrogate models, validate recommendations, trust their own judgement against a model. The priority population. Field supervisors Practical tool fluency: read alerts, validate them against the field, act. Their buy-in disproportionately determines adoption. Maintenance teams Act on remaining-useful-life estimates and criticality scores; feed disciplined work-order and failure-coding data back HSE teams Field analytics and especially governance of how AI-derived safety and emissions information is produced and defended Commercial teams Strong literacy and prompt-for-operations fluency, to get reliable answers and stress-test them BD teams Literacy and prompt fluency, with enough governance awareness to represent the company's AI capability honestly Project managers Governance and role-mapping: what analytics can and cannot promise, and orchestrating the teams that act on it Executive leadership The deepest governance literacy of all: ask the right questions, set risk appetite, sponsor credibly - The book labels its role-based capability map illustrative. Three patterns recur: differentiation (depth matching the role), integration (technical and digital teams sharing enough language to collaborate) and governance (those overseeing AI, especially leaders, equipped to govern it).
- The claim about your own seat: a board that genuinely understands AI is, in the book's words, the single strongest predictor of whether a capability programme succeeds, because capability follows sponsorship.
- IN THE FIELD · THE WORKED CASE
- Field. A fictional fabrication yard installs an AI-enabled inspection tool. Supervisors were shown it once; when an alert disagrees with what they see, they ignore it and call the vendor.
IN THE BOARDROOM · THE PARALLEL- Boardroom. Its executive committee approved the purchase without asking who validates the model, who owns the data, or who is being developed to replace the vendor. The book calls those operating discipline questions. Same gap, seen from the top.
Map the executive row first. Capability follows sponsorship.
21:00 – 25:00
Before you close the laptop
- List the roles in your company that materially affect performance, starting from the book's eight and adjusting for your company type.
- For each, write the target depth (awareness, working, proficient or expert) and where it stands today.
- That is Section 25 of your plan: the capability each role in our company needs.
- Practice Exam 5 opens now: fifteen questions, closed book, covering Sessions 21 to 25. The pass mark is 70.
Decide whether your company will own its operational brain or rent it. Then choose ownership.
Key Terms
Intelligence layer
The capability that increasingly determines performance, from AI-enabled operations and data engineering to analytics, AI governance and training. Local Content 2.0 is about owning it.
AI workforce maturity curve
The climb from awareness to literacy, application and ownership, where Nigerian teams build, govern and improve their own models.
Role-based capability map
A definition, for each role, of which skills are needed and to what level: awareness, working, proficient or expert.
Operational traceability
The workforce's ability to say where data came from, who owns it, when it was captured, whether it was validated and how it links to a decision.
Self-Check — Before You Submit
1. The HR director of a fictional drilling and well services company proposes the same two-day AI course for every employee, from rig crews to the board, measured by certificates issued. What do you do?
- A. Approve it, because broad coverage raises literacy fastest
- B. Redesign it around a role-based capability map that sets each role's depth and ties training to real decisions
- C. Replace it with a data science course for everyone
- D. Defer training until the company buys an AI platform
B. Redesign it around a role-based capability map that sets each role's depth and ties training to real decisions — The book maps capability to roles because the right depth differs sharply, and a course that changes nobody's work is a certificate, not a capability. Broad coverage tempts, but one blanket course measured by certificates is activity.
2. What are the four stages of the book's AI workforce maturity curve, in order?
- A. Literacy, awareness, ownership, application
- B. Training, certification, deployment, scale
- C. Awareness, application, literacy, ownership
- D. Awareness, literacy, application, ownership
D. Awareness, literacy, application, ownership — The climb runs from awareness to literacy, application and ownership. The near miss puts application before literacy, but people must use tools safely before AI can be embedded in field workflows.
3. Which role does the book say needs the deepest governance literacy of all?
- A. Engineers
- B. Field supervisors
- C. Executive leadership
- D. Commercial teams
C. Executive leadership — Executives need the deepest governance literacy to ask the right questions, set risk appetite and sponsor credibly. Engineers tempt, but the book gives them the deepest field-analytics capability, not governance literacy.
Today’s Assignment
Write Section 25 of your plan: the capability each role in your company needs, mapped by depth against where it stands today.
- List the roles in your company that materially affect performance. Start from the book's eight and adjust honestly for your company type.
- For each, state the target depth (awareness, working, proficient or expert) in the programmes that matter for it, and the honest current level.
- Place your company on the maturity curve, with evidence, and name the stage it can reach next.
- Name your company's three most expensive capability gaps: routine work paid to vendors, a decision delayed because nobody internal could interpret a model, a model nobody local can maintain.
- Assess the executive row first and name the sponsor by role. The honest part: where does your company outsource by reflex? Aim for 400 to 900 words.
YOUR 90-DAY PLAN
Section 25 — The capability each role in our company needs.
How this is marked — the rubric
| Maps the company's performance-critical roles with a target depth and an honest current level for each | 35 |
| Places the company on the maturity curve with evidence and names the next stage | 20 |
| Names the three most expensive capability gaps in vendor spend, delayed decisions or unmaintainable models | 25 |
| Assesses the executive row honestly and names the sponsor by role | 20 |
| Total · pass mark 70 | 100 |
Submitting session 25
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: 25, 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 5 — Controlled Autonomy, Capital and People 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 5 →Closed book, no notes, and Coach Ritchie switches himself off while you sit it.Capabilities Covered
Owning the intelligence layerMaturity curve diagnosisRole-based capability mappingOperational traceabilityCapability sponsorship