Build an AI strategy pack for a fictional company
Score ideas, build a business case, list the risks and plan a pilot for Riverside Logistics.
What you will have at the end
A five-part strategy pack: scored use cases, a business case with a cautious case, a risk register, a pilot plan and a one-page memo for the leadership team. This is a ready sample for case interviews.
What you need
- A spreadsheet
- A document editor
- Any chat AI tool to challenge your thinking
All names, numbers and data in this lab are made up for practice. Never use real customer or personal data. Your ticks are saved in this browser.
The scenario
Riverside Logistics is a mid-sized delivery company with 300 staff. The chief executive wants an AI plan for the next 12 months. The budget is modest and the board is cautious.
You are the AI transformation strategist. You will pick where to start, make the business case, name the risks and plan a pilot.
All numbers are made up for practice. Use them as given, and say which ones you would verify in real life.
Starter material
Copy these into your own files or chat. Change anything you like.
What you know about the company.
Staff: 300. Delivery drivers: 120. Customer support team: 18. Back office: 40. Support gets about 2,400 contacts a month. About half ask where a parcel is. The rest are address changes, damage claims and quotes. Average handling time: 6 minutes per contact. Fully loaded cost of a support hour: 28.00. Claims team: 6 people. Each claim takes about 25 minutes of reading photos, emails and the policy. Dispatch planners build routes by hand for 40 minutes each morning. Data: parcel tracking system (good), support emails and chat (messy), claims folders (scanned photos and PDFs). Rules: customers are in several countries. Claims involve personal data. The board wants no customer-facing mistakes about money.
The ideas people have suggested.
1. Chatbot for "where is my parcel" questions 2. AI draft replies for support agents 3. Read claim photos and emails and prepare a claim summary 4. Route planning assistant for dispatch 5. Predict late deliveries 6. Auto-answer quote requests 7. Meeting notes and summaries for managers 8. Search across company policies and manuals 9. Screen job applicants for driver roles 10. Detect fraud in claims
Score each idea 1 to 5. For risk, 5 means low risk.
IDEA | VALUE | FEASIBILITY | DATA READY | RISK (5 = low) | TOTAL | QUICK WIN / CORE BET / PARK | OWNER 1 | | | | | | |
Template for the risk register.
RISK | LIKELIHOOD | HARM | CONTROL | OWNER | REVIEW DATE Example: Wrong claim summary leads to a wrong payout | Medium | High | Person approves every payout; sample checks | Claims lead | monthly
The steps
Do them in order. Each step says how you know it is done.
- 1
Read the company and ask five questions
- Read the profile. Write five questions you would ask the chief executive before choosing anything.
- Write the one business goal you will serve, in a sentence.
- Write what would make you say no to a project.
Done when: Five questions and a one-line goal.
- 2
Score the ten ideas
- Fill in the scorecard for all ten ideas, with a one-line reason for each score.
- Mark two quick wins, one core bet and the ideas to park.
- Say why you parked the risky ones, such as idea 9 (hiring) and idea 10 (fraud).
Done when: A full scorecard and a shortlist of three with reasons.
- 3
Check the data for your top idea
- Run the data readiness check: findable, owned, current, clean, allowed, permissioned, reachable, examples.
- Write the fix plan for the two weakest lines.
- Decide: go, fix first or stop.
Done when: A readiness check with a decision and a fix plan.
- 4
Build the business case
- Use the numbers in the profile. Work out the cost today for one use case, such as support handling time.
- Write a low, expected and high benefit. Be cautious: say what happens to the saved hours.
- List the full cost: build, monthly running, review time, training. Use the calculators on this site for running costs.
- Work out payback in months for the cautious case, and add a stop rule.
Done when: A one-page case with three benefit cases, a cost list, a payback and a stop rule.
- 5
Write the risk register and rules
- List at least eight risks: privacy, wrong answers, fairness, vendor, people, cost.
- Give each a control, an owner and a review date.
- Set the risk tier of each shortlisted idea, and who approves it.
Done when: A risk register with owners and review dates, and tiers for the shortlist.
- 6
Plan the pilot and the people side
- Write a pilot plan: team, length, baseline, success number, stop rule and decision date.
- Write a short change plan: the message to staff, training and who the champions are.
- Write how you will report: the scorecard fields and the rhythm.
Done when: A pilot plan and a change plan, each on one page.
- 7
Write the one-page memo
- Write the memo for the leadership team: recommendation, three reasons, numbers, risks, ask and next step.
- Read it aloud. Cut every sentence that a manager would skip.
- List the three numbers you would verify first in real life.
Done when: A one-page memo that opens with the recommendation and the decision you need.
Success checklist
How to know the lab is really done
Show it in an interview
- Present the memo in three minutes, with the numbers and the biggest risk.
- Explain why you parked two ideas, and what would change your mind.
- Say which assumptions you would test first, and how.
Common mistakes
- Picking the exciting idea instead of the best-scored one
- Counting saved hours as saved money
- Ignoring data readiness
- Forgetting the people side
Want a bigger challenge?
- Do the same exercise for a company you know.
- Build the scorecard in a spreadsheet with formulas.
- Add a twelve-month roadmap with quarterly reviews.
Next step: practice the interview questions for this role →
Other labs
A small agent that looks up orders and opens tickets, with checks, limits and a test set.
A researcher, a writer and a reviewer pass work to each other. You add limits and test the failures.
Play the forward deployed engineer for a fictional clinic: discover, scope, prototype, test and hand over.