Topic
Enterprise AI Transformation: Beyond the Pilot
Most enterprise AI stalls in the same place. The tools get deployed, the licences get counted, and the work underneath carries on exactly as it did before — same handoffs, same approvals, same cycle time. Adding AI to an unchanged process is not transformation; it is a faster way to do the wrong shape of work.
This hub is about the other path: finding the workflows where AI changes the economics, redesigning them end to end rather than bolting a copilot onto a step, and being honest about what moved. Written from 18 years inside enterprise systems programmes, where I watched the same gap open between the technology that was bought and the operating model that never changed.
7 articles · ~53 min, in 2 sections — each in reading order
Fundamentals
2 articles · ~15 minWhat AI Transformation Actually Means
Adopting AI tools, automating a task and redesigning the work are three different things. The difference, and a four-question test for which one you're doing.
Why Most AI Pilots Never Become Operating Systems
A pilot proves capability. Running the work needs state, an exception path, decision rights and an owner — which is why good demos never become the system.
Method
5 articles · ~38 minHow to Find Workflows Worth Redesigning With AI
Eight dimensions you can observe rather than debate — frequency, effort, input, judgement, exceptions, error cost, state, control — and the verdict they give.
How to Measure ROI From an AI Workflow
Licences times adoption times a guess measures deployment, not outcome. The ten metrics that move when work is redesigned, and how to baseline them credibly.
Where to Start With Enterprise AI
Not with a pilot. A sequence borrowed from enterprise systems delivery — and the three steps in it that change when the component is probabilistic.
How to Judge an Enterprise AI Vendor's Claims
An accuracy number without a denominator is not a claim. The questions that separate a product from a demo, from someone who runs vendor selections.
AI Pilot Exit Criteria: What a Pilot Must Prove
Not an accuracy percentage. The five gates a pilot has to pass before production — borrowed from cutover, where the questions are older and harder.
Free tools for this topic
Frequently asked questions
What does AI transformation actually mean?
AI transformation means the work itself is redesigned around what AI makes cheap — not that AI has been added to the work as it stands. Three things get called transformation: adopting a tool, so people have an assistant beside the same process; automating a task, so one step gets faster while the process keeps its shape; and redesigning the workflow, so handoffs disappear, decision rights move, and the elapsed time changes rather than the touch time. Only the third one changes how the organisation operates. The first two are useful, but they are inputs, not transformation.
Full article →Why do AI pilots fail to scale beyond the pilot stage?
Because a pilot is allowed to skip the four things that make work run. It picks clean cases and never meets the exceptions, which in enterprise processes are most of the effort. It holds no state, so it can answer a question but cannot carry a case that is open, half-approved or reopened. It has no place in the control environment — no logging, no approval path, nothing an auditor can be shown. And nobody's job description changed, so the old route stays open and gets used the first time volume spikes. None of those are model problems, which is why a better model does not fix them.
Full article →How do you choose which workflow to redesign with AI first?
Assess candidates on eight things you can observe rather than argue about: how often the work runs, how much human effort each case takes, how much of the input arrives as prose or documents, whether experienced people agree on the judgement, how much of the effort sits in exceptions, what a wrong answer costs, what the work has to remember between steps, and how much of the process is a control. Three of those size the prize, two decide whether AI is the right tool at all, and three set the cost and the ceiling. Pick the one or two that survive all eight, not the one that demonstrates best.
Full article →Why do AI ROI calculations overstate the benefit?
Because most of them are built from seats, an assumed number of hours saved per person per week, a loaded hourly rate and an adoption percentage. Only the seat count is observed; everything else is asserted, and the result measures how widely a tool was deployed rather than whether any work got better. The calculation also converts saved minutes straight into money, which only holds if someone actually removes the cost. And because nothing in it was ever measured, nothing in it can be checked afterwards — which is why these numbers are quietly dropped rather than revisited.
Full article →Should the first enterprise AI project be a pilot?
A pilot is a good way to answer a question and a poor way to start a programme, and the difference is whether you wrote the question down first. Run one to find out something specific — whether the input is machine-readable enough, what the real exception rate is, whether reviewers can work at the required pace. Do not run one to demonstrate that the technology works, because that answer is already known and the demonstration teaches you nothing you can build on. A pilot with no written question becomes a slide.
Full article →Work with me → — an opportunity scan, a redesign sprint, or an architecture review before AI meets your real landscape.