Pilot programme design
Three AI pilots aimed at one line of the balance sheet
A building-products manufacturer had capital trapped in inventory and a digital team already delivering wins. I proposed three pilots tied to working capital rather than to a technology roadmap.
- Ranked by balance-sheet impact, not by technical interest
- 3 pilotsRanked by balance-sheet impact, not by technical interest
- Each stage priced separately and separately cancellable
- PoC → MVP → scaleEach stage priced separately and separately cancellable
- The maintenance pilot targets equipment they already own
- No new capexThe maintenance pilot targets equipment they already own
The client
A North American building-products manufacturer with dozens of plants across two continents, a competent in-house digital twin team, and roughly three quarters of its inventory sitting in raw materials exposed to price and lead-time volatility.
The engagement
Three staged pilots, each priced as proof of concept, then MVP, then scale — so each stage buys the right to fund the next.
The problem
The client's pain was working capital — inventory days rising year on year, each additional day tying up serious money, and delivery reliability slipping below target with a revenue headwind attached. Their digital team had already delivered a throughput win on their own, which raises the bar rather than lowering it: any proposal that reads as a generic AI capability tour gets dismissed by people who have already done the real thing.
What I did
I ranked the pilots by proximity to the balance sheet rather than by technical elegance. Forecasting came first because inventory days is the metric under pressure, and I specified hierarchical Bayesian models rather than a standard forecasting stack for a stated reason — sparse product lines and tariff shocks are exactly where conventional approaches break, and the output is a probability band a planner can act on rather than a single number nobody believes. The second pilot was deliberately chosen to need no capital replacement, because a proposal that requires new machinery competes with the machinery budget. The third rode on a migration the client was already funding, so it borrowed an existing budget line instead of asking for a new one. Each pilot was priced in three stages with a team shape per stage, so the client could stop after any one of them without stranding the work.
What was built
Three pilots in priority order: hierarchical Bayesian forecasting of demand and lead time, chosen specifically because it handles sparse product lines and tariff shocks where standard forecasting collapses; predictive maintenance on already-depreciated equipment, so throughput improves without capital replacement; and Monte Carlo scenario planning over the enterprise system for disruption and tariff what-ifs.
On the table at the end
- Deep-dive case document covering all three pilots — architecture, timeline and financials
- Executive deck framing the pilots against working capital
- Staged pricing envelope with team shape per stage
What it changed
Anchored an AI conversation to inventory days and on-time-in-full delivery — the two numbers their board already watched — so the programme was evaluated as a working-capital initiative rather than as an innovation budget.
How it ran
- 01
Open on their numbers
Inventory days, cash conversion cycle and delivery reliability — sourced and stated first, so the pilots arrive as answers to a problem they already own.
- 02
Rank by balance-sheet distance
Forecasting first because it moves the metric under pressure; the elegant pilots ranked below the useful one.
- 03
Justify the method, not just the outcome
Hierarchical Bayesian modelling specified with the reason it beats the default: sparse lines, volatile lead times, probability bands over point estimates.
- 04
Ride existing budgets
One pilot on equipment already owned, one on a migration already funded — reducing the number of new budget conversations from three to one.
- 05
Stage the money
Proof of concept, then MVP, then scale, each priced separately with its own team shape and its own exit.
Other work
All case studies →- Capital markets
An agentic core with a human gate on every state change
A thirty-year financial house wanted autonomous AI in a regulated business. I designed the agents to propose and people to approve, then priced it against a payback date instead of a benefit narrative.
- Investment holding
Presenting the build-versus-buy fork before anyone fell in love with building
A multi-entity holding wanted document AI across its finance operations. I designed the pipeline and then showed them the packaged alternative honestly, including where it would beat us.
- Insurance
Sequencing the data foundation before anyone bought an AI agent
A financial group wanted AI scoring across five affiliates. I proposed one affiliate, eighteen weeks, and a data readiness report before a single model — because the alternative is an agent trained on data nobody has reconciled.
Something similar on your plate?
Thirty minutes, no deck. I will tell you whether it is worth doing at all.