Retail Strategy and the AI Bandwagon
A national retailer with 320 stores and 35 percent online sales has a five-year strategy focused on customer loyalty and supply-chain resilience. The board has read about generative AI and asks management to put AI in the strategy before the annual investor day. The CIO has a backlog of 18-month-old legacy replenishment systems, a data platform that covers only online transactions, and a new chief data officer who joined three months ago. Store managers still make manual stock transfers. A vendor is offering a generative AI customer-service assistant with a six-month payback claim. The CFO is sceptical; the CMO wants it for personalisation. You must advise the board on how to evaluate and sequence IT opportunities so they align with strategy rather than fragment it.
1. How should the board evaluate the AI proposal against the business strategy? What evidence would make it a yes, no, or not yet decision?
Read the worked answer
Evaluate against the five-year strategy, not the board's enthusiasm.
A 'yes' requires: a clear link to loyalty or supply-chain resilience; a quantified baseline (for example, cost per contact, resolution time, stock availability); evidence that the data needed is accessible and of sufficient quality; a realistic total cost including integration and change; an assessment of vendor lock-in and customer-data risk; and a benefits owner. A 'no' follows if the proposal is a solution looking for a problem, or if the risk exceeds appetite. 'Not yet' is likely if the data platform covers only online transactions and store data is incomplete: the assistant may work for online queries but fail on store stock or returns.
Use a stage gate: approve a small discovery or pilot, not enterprise rollout. Ask what would have to be true for the payback claim to hold, and test the assumptions. The board should also compare the proposal with alternatives such as improving existing search or store replenishment, which may deliver more strategic value.
2. How do you handle the CIO's legacy backlog and the CDO's new data platform in the same portfolio?
Read the worked answer
Treat the legacy backlog and the data platform as one portfolio, not competing pet projects.
- Map dependencies: AI and personalisation depend on clean product, customer and stock data; replenishment depends on store data and supply-chain integration.
- Sequence by strategic capability and risk reduction, not by who shouts loudest. Fund the data platform's store-data scope if it unlocks both loyalty and replenishment.
- Protect a small innovation envelope, but do not let it consume the remediation budget.
- Use architecture review to prevent point solutions that deepen fragmentation.
- Set portfolio-level criteria: strategic fit, risk reduction, dependency readiness, value, and total cost.
- Report trade-offs to the board: what is being delayed and why.
This avoids AI on a broken foundation. The CDO's new platform should be positioned as enabling infrastructure with explicit business outcomes, and the CIO's backlog should be prioritized by the same criteria. The board's role is to approve the portfolio shape and risk appetite, not to choose individual tools.
3. What governance behaviours would prevent innovation theatre?
Read the worked answer
Prevent innovation theatre through governance behaviours:
- Require every initiative to state the business outcome and baseline before funding; demos are not evidence.
- Use stage gates with kill criteria, and celebrate stopping weak ideas early.
- Assign a business owner for benefits, not just an IT delivery lead.
- Ask for evidence of user adoption and operational change, not just model accuracy.
- Separate experimentation from scaling: small, cheap tests are fine; scaling requires a business case and risk review.
- Include conformance and human-behaviour risks in the discussion: staff workarounds, customer trust, regulatory expectations.
- Report a balanced portfolio: remediation, efficiency, growth, and innovation.
The board should model curiosity without suspending scepticism. If the CMO wants personalisation and the CFO is sceptical, the governance answer is a bounded test with agreed metrics and a decision date, not a culture war.