Appendix 5.1 · Strategy Course

Oristella Strategy Course

Student programme with teaching notes assignments and worked examples

This course uses Oristella as a case for designing and testing a competitive strategy for an AI-enabled professional service. Students turn strategic claims into choices, operating requirements, financial reasoning and evidence tests. They may recommend proceeding, narrowing the scope or stopping. The quality of their reasoning determines the assessment.

Proposed delivery: eight weeks, 72 hours total. Each week has one hour of concept teaching, two hours of case workshop and six hours of independent work. Intended entry level: final-year undergraduate or postgraduate students with basic business arithmetic. A short prerequisite exercise covers percentages, contribution and cash versus profit. This is a curriculum proposal, not an accredited qualification.

Learning outcomes

1. Define a buyer, an urgent job, relevant alternatives and a defensible competitive position.

2. Connect strategic choices to activities, resources, exclusions and a staged delivery plan.

3. Measure automation, learning, review capacity and total cost on a comparable accepted output.

4. Separate learning economies, network effects, retention and contractual switching costs.

5. Explain the relationships between reusable information, customer rights, cash, debt, profit and recognised assets.

6. Defend a 90-day investment and delivery decision with traceable evidence, sensitivity analysis and explicit stop conditions.

Assessment

Eight module submissions contribute 5% each, 40% in total. The final decision paper, model and presentation contribute 45%; individual oral defence contributes 15%. Team size is three to four; each student keeps a contribution record and defends one calculation and one strategic choice. Proposed pass threshold: 50% overall and at least 50% in the individual defence. A material unsupported claim or unexplained calculation error must be corrected before a pass.

Students disclose AI assistance, retain source references and independently check important calculations. Simulated customer interviews must be labelled as simulations. No external contact, confidential mining dataset or live operational action is needed for this course.

Teaching sequence and submission standard

Week	Module	Submission
1	Choose a market and a position	Prepare a two-page market and positioning brief for one initial Oristella offer.
2	Design the activity system	Create one activity map and a four-stage release plan.
3	Measure automation and expert capacity	Build a task measurement sheet and evaluate a proposed four-system supervision model.
4	Price delivery and continuing service	Prepare a one-page offer and a 12-month unit-economics and cash schedule.
5	Govern knowledge reuse and network effects	Produce a module register, a licence flow and an experiment for shared learning.
6	Build the market and university channel	Design a 90-day sales experiment and a university teaching pilot.
7	Plan delivery and regional expansion	Prepare an operating responsibility matrix and a regional expansion decision.
8	Defend the investment and information model	Submit the final decision paper, evidence register, model and presentation.

Workshop rhythm: 20 minutes to inspect the claim; 40 minutes of group analysis; 30 minutes of peer challenge; 30 minutes to revise and state the decision. The concept session introduces the definitions and worked example. Independent work prepares the assessed output and a source register.

Every submission identifies the decision, source evidence, assumptions, calculation method, uncertainty and next test. Students may challenge the strategy and receive full credit for a well-supported decision to narrow, defer or reject the proposal.

1 Choose a market and a position

Outcome: Produce a specific buyer and job definition and defend it against two plausible alternatives.

Source mapping: strategy theses 1, 2, 7 and 13; appendices 1, 2, 3, 4.

A competitive position connects a particular customer need to a distinctive way of serving it. In the Oristella case, junior gold companies in Guinea are a starting segment, but a licence deadline, missing specialist capacity and available source data make the buying situation more specific. A company without adequate data may need fieldwork before a documentation service can deliver value.

Separate price, supplier cost and customer outcome. A lower price can destroy the supplier’s margin; a lower cost can preserve margin without being passed on. Quality must refer to acceptance criteria and the severity of errors. “Highest quality” across an undefined market is not a testable specification.

Map the buyer’s actual alternatives: a consultant, internal specialists, separate software, a smaller scope or delay. Investigate who controls the budget and bears the cost of failure. The five-forces exercise concerns the profitability of this service business, not merely the gold price or the profitability of a mine.

End with a choice. A useful position states whom we serve, which job we deliver, the promised result, the economic mechanism and what we exclude. The course exercise is a hypothesis until customer evidence supports it.

Assignment

Prepare a two-page market and positioning brief for one initial Oristella offer.

1. Select one buyer and one buying trigger.

2. Compare three alternatives on equal scope, time, cost and acceptance.

3. Write a 60-word positioning statement and three explicit exclusions.

4. List five critical assumptions and one disconfirming test for each.

Required output: Buyer and job definition; Alternatives comparison; Positioning statement; Assumptions and tests.

Worked example

Illustrative positioning: We serve a junior operator in Guinea that needs an auditable licence-extension package from existing, authorised data. We coordinate preparation and specialist review under an agreed scope and fee. New field investigations and regulatory approval are outside the promise. The buyer accepts the package against a checklist agreed before work starts.

Acceptance checks: The buyer and approver are named by role. Comparisons use identical scope and input quality. At least one finding could overturn the recommendation.

Reference reading: 1.

2 Design the activity system

Outcome: Translate the position into an activity map and a bounded first release.

Source mapping: strategy theses 2, 9, 12 and 13; appendices 6, 7, 8, 12, 13.

List the activities required to deliver the chosen promise: qualification, rights clearance, data validation, modelling, professional review, delivery and support. Then explain the links. A standard input contract can reduce validation effort; reusable procedures can reduce preparation time; independent acceptance can prevent low-quality work from entering the reusable library.

This is an activity system only when the links reinforce the position. A list of tools is insufficient. Identify at least one trade-off: a broad range of bespoke projects may generate revenue while slowing a repeatable first release. Evaluate that trade-off explicitly.

Convert the first release into a demonstrable workflow. Define its starting input, final accepted output, responsible human, failure paths and excluded decisions. A documentation matrix records preparation progress. It does not prove that software, integrations or recovery work.

Use stage gates to control expansion. A missed test may justify repairing the workflow, changing the offer or stopping. Farafina is a useful test environment; acceptance by an independent paying client remains a separate gate.

Assignment

Create one activity map and a four-stage release plan.

1. Draw 8–12 activities with at least five justified reinforcing links.

2. Identify three exclusions and the opportunity cost of each.

3. Assign one accountable role to each gate.

4. Define pass, repair and stop decisions for each gate.

Required output: Activity map; Trade-off register; Four-stage release plan; Responsibility and acceptance matrix.

Worked example

Example link: a common data dictionary reduces repeated mapping work; validation catches incompatible units before modelling; specialist acceptance determines which corrected methods enter the next release. A failed coordinate-system test prevents model release even if the document is otherwise complete.

Acceptance checks: The map explains causal links. The plan has observable outputs rather than completion percentages. Every gate has an owner and a failure response.

Reference reading: 1, 2, 5.

3 Measure automation and expert capacity

Outcome: Calculate labour substitution and feasible supervision capacity without confusing them with headcount reduction.

Source mapping: strategy theses 3, 4 and 12; appendices 10, 13, 15.

Measure a comparable accepted task. Baseline labour includes preparation, review and corrections. Assisted labour includes data preparation, AI operation, human review and rework. Labour substitution equals one minus assisted hours divided by baseline hours. This ratio is meaningful only when scope and acceptance quality are comparable.

An expert-to-system ratio is a queueing and capacity question. The expert needs time for both routine review and exceptions. Available hours must allow for management, training and peak workload. A system count without a workload definition cannot establish capacity.

An experience curve requires comparable observations across completed assignments. Lower hours can reflect an easier task, better inputs or a changed model rather than learning within Oristella. Record these factors and use a held-out evaluation set when testing a release.

Choose automation by total economic contribution subject to constraints. More automation can reduce preparation work while raising exceptions or computing costs. An inflection point in a curve is not necessarily the maximum. Evaluate several feasible levels and retain required human sign-off.

Assignment

Build a task measurement sheet and evaluate a proposed four-system supervision model.

1. Use the supplied synthetic task data and calculate the baseline and assisted hours.

2. Calculate substitution and total delivery cost.

3. Test four and five systems against review capacity and a 20% capacity buffer.

4. Specify a controlled pilot that could validate or reject the result.

Required output: Transparent calculation; Capacity decision; Measurement protocol; Quality and exception rules.

Worked example

Synthetic example: baseline 120 hours; assisted preparation 24, review 16, rework 8, total 48. Labour substitution is 60%. At $50 per hour plus $600 computing, assisted cost is $3,000 versus $6,000 baseline. Cost saving is 50%, not 60%. With 80 review hours per month and 20 hours per system, raw capacity is four systems; a 20% buffer permits three whole systems.

Acceptance checks: Review and rework are included. Labour and cost ratios are separate. The conclusion respects quality and capacity limits.

Reference reading: 5.

4 Price delivery and continuing service

Outcome: Construct contribution, break-even and cash calculations for a defined offer.

Source mapping: strategy theses 5 and 10; appendices 4, 11, 16.

Start with the contract. A fixed fee requires a defined scope, acceptance procedure and change control. Recurring fees need a continuing benefit and corresponding obligations. A new customer may need more onboarding and support than an established customer.

Contribution is revenue less the costs that vary with delivery. Fixed development and central administration are then deducted to estimate operating result. A growing customer count can improve fixed-cost absorption until a new support or infrastructure capacity band is required.

Cash, revenue and profit differ. An advance can provide cash before revenue is earned. A receivable can arise when revenue is earned before cash arrives. Debt proceeds increase cash and a liability; principal repayment reduces both without becoming an operating expense.

Build at least three scenarios using explicit drivers. In Oristella’s adjacent workbook, 15% monthly revenue growth and flat operating spending are assumptions. They need a customer, workload and capacity explanation before they become a convincing growth case.

Assignment

Prepare a one-page offer and a 12-month unit-economics and cash schedule.

1. State what the fixed fee and recurring fee each cover.

2. Separate direct delivery, support and fixed cost.

3. Calculate contribution and break-even volume.

4. Stress price by minus 10%, delivery cost by plus 20% and cash collection by one month.

Required output: Offer scope; Unit economics; Cash timing; Sensitivity conclusion.

Worked example

Synthetic exercise: a $10,000 assignment with $4,000 direct cost contributes $6,000. Monthly fixed cost of $24,000 requires four such assignments to break even. At a $9,000 price and $4,800 cost, contribution is $4,200 and break-even is six whole assignments. A one-month payment delay requires separate working capital even when the work is profitable.

Acceptance checks: Costs are not counted twice. Break-even uses contribution, not revenue. The cash schedule includes timing and opening cash.

Reference reading: 1.

5 Govern knowledge reuse and network effects

Outcome: Design a rights-aware learning loop and distinguish its value from asserted network effects.

Source mapping: strategy theses 6, 8, 9 and 14; appendices 5, 9, 14, 17.

Reusable knowledge can reduce repeat work. Shared infrastructure can spread fixed cost. A network effect requires a benefit to a participant from other participants or their contributions. These mechanisms can coexist, but each needs a separate explanation and measurement.

Separate the master module, its versions, customer licences and deployed copies. A non-exclusive licence can leave the original resource with Oristella. Customer possession of a copy does not establish unrestricted reuse or training permission. Record rights before a dataset contributes to shared learning.

Retention can come from useful enterprise models, trained people and integrations. Describe portability and exit arrangements alongside switching cost. The client should understand the value of staying and the terms of leaving.

Token records can link authorised grants and reported usage. They cannot prove that every off-chain copy has been found. Accounting records must distinguish cost, carrying value, a licence price and a proposed commercial valuation. Issuing a free accounting token does not itself create revenue or a monetary asset.

Assignment

Produce a module register, a licence flow and an experiment for shared learning.

1. Register one master module, two versions and two customer licences.

2. Separate owner, custodian, authorised user and physical deployment.

3. Describe permitted and prohibited reuse.

4. Design a held-out test for whether approved new knowledge helps existing clients.

Required output: Object and licence register; Rights matrix; Learning experiment; Accounting and token boundaries.

Worked example

Example: module GEO-001 version 2 is licensed to clients A and B. The module remains one master resource; there are two licences and two deployments. Client A permits operational processing but prohibits central model training. That data stays out of the shared training set. A synthetic carrying cost of $50,000 is not multiplied by the number of licences.

Acceptance checks: Rights and custody are distinct. No duplicated asset values arise from copies. The experiment isolates a measurable benefit to existing clients.

Reference reading: 3, 4.

6 Build the market and university channel

Outcome: Design a measurable customer experiment and an independently assessed education pathway.

Source mapping: strategy theses 7 and 11; appendices 3, 11, 13, 15.

A lower service cost can broaden demand, but affordability is only one condition. A prospect also needs an eligible project, data, budget authority and a reason to act. Record the difference between a new customer previously unable to buy and a customer switching from another provider.

Build a funnel with defined stages: eligible prospect, qualified conversation, relevant demonstration, proposal, paid assignment and repeat use. Use consistent cohorts and dates. A demonstration is not a contract, and a related-party assignment is not independent market validation.

University collaboration can develop operators and create introductions. It needs a teaching agreement, permitted datasets, assessment criteria and a supervised route into practice. Keep enrolment, competence, employment and customer acquisition as separate measurements.

Begin with a small, falsifiable experiment. Specify who is approached, what they see, what counts as conversion, the time limit and how failure changes the offer. A student course is also allowed to reveal that the strategy needs revision.

Assignment

Design a 90-day sales experiment and a university teaching pilot.

1. Define five prospect qualification fields.

2. Set proposed funnel targets and a stop or revise rule.

3. Create a competence test for one module.

4. Explain how student participation is separated from customer sales and confidential project work.

Required output: Funnel definitions; Experiment plan; Competence test; Partnership and data requirements.

Worked example

Illustrative funnel: 10 qualified conversations, four demonstrations, two paid pilots and one repeat order. Conversation-to-paid conversion is 20%; demonstration-to-paid conversion is 50%. These are different denominators. A proposed university pilot assesses ten students on a permitted dataset; passing does not guarantee employment or authorise independent professional approval.

Acceptance checks: Denominators and periods are explicit. Targets are labelled as proposed. Education and sales outcomes are measured separately.

Reference reading: 1.

7 Plan delivery and regional expansion

Outcome: Make a resourced go or no-go decision for one regional delivery centre.

Source mapping: strategy theses 2, 9, 12 and 13; appendices 8, 12, 13, 14, 18.

The initial operating centre and a future franchise have different ownership, funding and control. Define who owns premises, equipment, software and customer relationships. A partner contribution is not automatically an asset owned by Oristella.

Plan power, connectivity, backup, offline operations, access control, local staffing and central support. Establish service limits and recovery targets from the customer’s operating need. A diagram of a centre is not evidence that these controls work.

Resource the human review bottleneck before promising more customers. Identify the capacity band that requires another specialist or infrastructure investment. Build partner support costs into the expansion case rather than treating a franchise fee as pure profit.

Use a release and risk register. Each risk has a cause, consequence, owner, control, test and residual exposure. Independent acceptance, restoration and company-separation tests are material gates. Government and university relationships remain proposed unless documented.

Assignment

Prepare an operating responsibility matrix and a regional expansion decision.

1. Assign responsibilities for premises, equipment, data, approvals and support.

2. List capacity needs and funding sources.

3. Specify five acceptance tests and their pass criteria.

4. State what evidence would postpone expansion.

Required output: Responsibility matrix; Capacity and funding plan; Acceptance tests; Expansion recommendation.

Worked example

Example release test: restore an approved dataset from backup into an isolated environment, verify its manifest and reproduce one accepted output. Record elapsed time and any missing records. Passing this test supports the restoration claim; it does not establish that every security control is effective.

Acceptance checks: Ownership and operating responsibility are separate. Capacity and support costs are visible. Every critical risk has an executable test.

Reference reading: 5.

8 Defend the investment and information model

Outcome: Reconcile strategic claims, financial consequences and evidence into an investment decision.

Source mapping: strategy theses 4, 10 and 14; appendices 16, 17, 18.

Begin with the decision: fund, fund a smaller test, defer or decline. Explain which evidence justifies it. A forecast may balance arithmetically while depending on unrealistic sales or capacity assumptions. Test both the arithmetic and the business mechanism.

Reconcile funding to debt, cash flow to closing cash and profit to equity. Debt principal, dividends and operating expenses are separate movements. The source package describes repayment followed by a transfer of existing founders’ shares; final contractual terms need to define that arrangement clearly.

Information assets require a controlled unit of account. Development expenditure is not capitalised merely because a useful file exists. Customer licences and deployed copies can be recorded without duplicating the master asset’s cost. Assess useful life and recoverability separately from token quantity.

Present the strongest counterargument to your recommendation. State which result would change your mind, who must obtain it and how much time and money the test requires. Your final strategy must remain coherent when an investor opens the presentation, appendices and workbook in any order.

Assignment

Submit the final decision paper, evidence register, model and presentation.

1. Trace every headline metric to a source or a clearly labelled assumption.

2. Reconcile cash, debt, equity and the information register.

3. Test downside sales, cost and collection timing.

4. Defend a staged funding recommendation and its stop conditions.

Required output: Decision paper; Eight-slide presentation; Traceable model; Evidence register and 90-day plan.

Worked example

Synthetic accounting example: opening cash 100 and equity 100; borrow 50; earn and collect 40; pay operating costs 25; repay principal 20. Closing cash is 145, debt is 30 and equity is 115. Profit is 15. The principal repayment changes cash and debt, not profit. This exercise is independent of the Oristella forecast.

Acceptance checks: The balance and cash bridge agree. Every figure uses a defined period and unit. The recommendation includes a credible opposing case and decision triggers.

Reference reading: 1, 4, 5.

Final project and marking rubric

Recommend a 90-day launch decision for one defined Oristella offer in Guinea. Submit an 8–10 page decision paper, an eight-slide presentation, a transparent calculation file and a source register. Include a one-page executive decision, market and alternatives, activity map, automation and capacity test, unit economics and cash, knowledge rights, operating controls, and a staged plan. Conclude with the strongest reason not to proceed and the test that would resolve it.

Criterion	Points	Assessment
Evidence and traceability	25	Claims link to reliable sources, periods and assumptions; contrary evidence is considered.
Strategic coherence	25	Buyer, promise, activities, exclusions and expansion choices reinforce each other.
Economic and operational reasoning	25	Calculations reconcile; capacity, cash timing and sensitivities affect the decision.
Execution and governance	15	Responsibilities, rights, acceptance, risks and stop conditions are actionable.
Clarity	10	A reader can follow the decision and reproduce the important calculations.

For each criterion: excellent work is supported, reproducible and considers contrary evidence; competent work is substantially correct with limited gaps; incomplete work relies on unsupported claims or misses a material dependency. Award points within the criterion maximum and explain the reason. Do not award a high score merely for agreeing with the company’s preferred conclusion.

Individual defence

Each student has ten minutes: three minutes for the decision, four minutes to explain a calculation or evidence choice, and three minutes for challenge questions. Defence marking: 40% factual and numerical command, 40% response to counterarguments and 20% clarity. The resulting defence score carries 15% of the course grade.