The situation
A global fashion retailer ran one of the larger AI programmes in its industry: demand forecasting, inventory optimisation, pricing and markdown, recommendations, generative AI. Activity was high and adoption was real. What the programme could not do was answer the board’s question: what is all of this worth? Initiatives entered through per-function wish lists, were funded per use case, and were rarely stopped once started. The portfolio grew; the value story did not.
What was done
The portfolio was rebuilt around value. Every initiative — running or proposed — was valued upfront against the P&L line it was meant to move, before build, not after. Intake moved to one cross-domain front door, so that a forecasting improvement and a markdown model competed for the same capacity on the same terms instead of living in separate functions’ budgets. And the discipline that most programmes lack was enforced: initiatives that did not pay back were stopped, and their capacity moved to the ones that did.
Around the portfolio, the operating model was made explicit: who owns a capability once it is live, how it is funded, how it is governed, and when to build versus buy — including the shift from custom builds to partnering on prebuilt models where building added no advantage.
The result
A portfolio the organisation could steer: ranked by estimated value, read across domains, and pruned on evidence. The capabilities that mattered — forecasting, inventory, pricing, markdown, recommendations — reached production with an owner, a funding line and a value estimate that could be checked.
Why it is here
This case is the origin of the Ardinois assessment. Classifying a portfolio by what each initiative is for and what economic effect it targets, naming value before build, and having the standing to stop things: that is the discipline the Information Audit brings in from outside.