Adoption is finished and the return never arrived: almost every large company now uses AI somewhere, yet the money shows up in no accounts anybody can point at. The leak is structural, it happens before the first line of code, and that is the only point at which it is cheap to fix.
The trap most companies walked into run pilots everywhere, count the ones that shipped a demo, and mistake activity for return, while the business case quietly rests on hours saved that never leave the cost base and a baseline nobody thought to capture before go-live.
What the minority doing better does instead: start from a costed problem with a named budget holder, capture the baseline before deployment changes anything, then concentrate engineering on three survivors instead of spreading it across twenty pilots that each miss the minimum bar for production
In the programmes that produce numbers, the accountable owner is the person whose cost or revenue line is supposed to move. The AI team builds and the operations director owns the outcome. A budget holder will not fund what cannot be measured, and will stop what has stopped improving.
Volume, cycle time, error rate, rework and fully loaded cost, recorded before a single component is deployed and stored where finance can see it. Two weeks of work. Once the process has changed, the pre-AI comparison cannot be reconstructed and every claim becomes contestable.
Retrieval, evaluation, access control, cost governance and observability built once and reused. The first use case looks expensive. The fifth is cheap and fast, which is the only shape a real return curve takes. Forty pilots with no platform is further from value than one deployed system.
List every initiative with its owner, annual run cost, stage and claimed benefit. The list is almost always longer than leadership believes, and several entries will have no named owner at all.
List every initiative with its owner, annual run cost, stage and claimed benefit. The list is almost always longer than leadership believes, and several entries will have no named owner at all.
Give each one to the person whose numbers should change, with authority over process and headcount in their own area. Capture the baseline within two weeks, signed off by finance before deployment touches the process.
Agree the metric movement required to keep funding at 30, 60 and 90 days, and write down who decides. Then report one page per initiative each month: baseline, current, delta, spend, decision.
A demo and a production system are different builds. Deployment needs evaluation pipelines, access control, audit logging, cost ceilings and fallback
behaviour. Teams that defer those are not moving faster, they are moving the whole cost to the day somebody asks to ship.
Partly. The pre-AI comparison cannot be reconstructed once the process has changed. You can still stage the remaining rollout by region, team or queue,
which gives you a real comparison group from here forward at no extra cost.
The opposite. Engineering capacity is fixed and production readiness has a minimum cost per use case. Spread across twenty pilots, nothing clears that minimum. Concentrated on three, all three can ship, and shipping is the only state that returns anything.
It measures the tool, not the business. Queries answered and satisfaction scores tell you the system is being used. Your controller needs a ledger comparison or a metric from the operational system of record, which is a different dataset entirely.
Saved hours are not money until something changes in the cost base or the revenue line, and in most organisations nothing does. The time gets absorbed, headcount stays flat, and finance correctly declines to book the saving. Ask which budget line will be smaller next quarter, and by how much.
CEOs, CTOs and the finance leaders being asked what the AI spend returned. It assumes you have already run pilots, already heard the vendor pitch, and are now the person who has to answer the return question at a board meeting with numbers rather than activity.
Drop your details and we'll send Why 56% Of CEOs See No Revenue Gain From AI straight to your inbox - no spam, unsubscribe anytime.
Bring your current AI project list to a working session with our engineering leads and we will tell you which three are worth finishing.
Download White Paper