Early August, quarterly review at Norland Valves Ltd, a mid-sized maker of industrial valves — that is, the valves and control equipment used in plant engineering and by utilities. Millie Lawton, Head of Order Processing, presents her initiative, Project Orderflow: using AI to automate order entry. George Farrell, Managing Director and third-generation owner, listens. Then he delivers the line that has greeted every investment proposal in the company for years: "We spent £300,000 on the CRM, and to this day nobody can say what it got us."
Claire Bratton, the Finance Director, puts it more gently, but no less firmly: "We're not saying no. What we're saying is this: this time we want to agree up front how success will be measured, and then see afterwards whether it happened." Millie gets her assignment: a business case, to be submitted by September, ahead of the budget round.
By the end of this lesson you will be able to:
- explain why "we delivered" no longer counts as proof of success
- name the three patterns by which initiatives squander their benefit before it is ever measured
- put the pressure from finance into context — and use it as an opportunity rather than treating it as an attack
Delivered is not the same as effective
For decades, projects ran on a simple definition of success: on time, on budget, to the agreed scope. If you could tick all three boxes, you had won. The problem: none of those three ticks says anything about whether the organisation is better off after the project than before it.
A project can deliver a system on time and on budget that nobody uses. It can hit the agreed scope exactly, and the scope itself was wrong. Go-live is the moment when project costs stop growing. The running costs are only just beginning, and the benefit could start. Whether it does is decided afterwards, and under the old definition of success, nobody is looking by then.
That is exactly what happened at Norland with the CRM. The system runs. It probably even helps. But there was never a benefit calculation, never a baseline, never a measurement. The benefit is not zero — it is unproven. For the management team, the practical result is the same: £300,000 spent, and nobody can show what it achieved. Keep this distinction in mind, because it runs through the whole course: not proven is not the same as not there. But if you cannot produce the proof, you lose the argument, and by the next initiative you have lost their trust.
Three patterns that squander the benefit
When initiatives cannot demonstrate their impact, it almost always comes down to at least one of three patterns, and usually all three at once:
Benefit claimed and never measured. The proposal says "saves time and reduces errors". It says so without a number, without a baseline, without a measurement plan. After go-live, the claim can be neither confirmed nor refuted — it was never built to be testable.
Costs underestimated. The proposal contains the project costs. What is missing: operations, licences, maintenance, training — the running half. Even if the benefit arrives, the costs nobody quantified eat it up, and in hindsight the initiative looks more expensive than promised.
Success defined as a date. What gets celebrated is go-live, not impact. The project team is disbanded, the project lead moves on to the next initiative — and from that moment on, nobody owns the question of whether the promised benefit actually arrives.
These patterns are not down to careless individuals. They are the logical consequence of a system that measures projects by delivery and then stops looking.
The pressure is real — and measurable
The scene from Norland's quarterly review is playing out in a great many organisations right now. Money has been pouring into AI initiatives for several years, and finance functions are discovering that they cannot pin down the return.
Two numbers are enough to make the point. In a KPMG survey of more than 2,100 senior executives in 20 markets (Q2 2026), only 7 per cent say they have an established ROI measure for their AI investments; 42 per cent do not even have full visibility of their AI spending. And in a Deloitte survey of 1,854 executives across 14 countries — 249 of them in Germany, with fieldwork from mid-August to early September 2025 — more than two thirds of organisations report that they only achieve their typical AI ROI after two or more years.
A word on how to read these numbers: they point to an organisational problem. Companies invest without being able to prove impact. They are not evidence that the initiatives are useless. But they explain why boards and finance teams are asking harder questions than they used to: they have learned that "it's bound to pay off" is not an answer. Further vetted studies are collected in the handout at the end of this module, deliberately there and not here, because numbers like these age quickly.
Why this is good news
If you lead projects, this pressure feels at first like the screws tightening: more justification, more queries, higher hurdles. In fact it is an opportunity, for two reasons.
First, the question "what does this get us?" is coming anyway. The only thing you choose is when. If it comes before the decision, you answer it with a business case you built yourself — with your assumptions, your scenarios, your yardstick. If it comes after go-live and you have nothing prepared, someone else answers it for you, with the CRM line.
Second, anyone who can show what their initiatives deliver gets the next one approved more easily. Trust with the management team and with finance is cumulative: every proof you produce lowers the hurdle for the next proposal. Millie Lawton has understood this: she treats Farrell's CRM line not as an accusation but as her brief.
Wrap-up
This is the starting point of the course: a successful go-live is not success. It is the precondition that makes success possible. Whether it materialises, and whether you can show it, comes down to tools you will build piece by piece over the next modules. You will meet the first of them in the next lesson: the benefit chain, the golden thread of the whole course.
Key Learnings
- On time, on budget, delivered — none of these ticks proves the organisation is better off after the project. Go-live is the starting point of the benefit, not its proof.
- Not proven is not the same as not there — but if you cannot produce the proof, you lose the argument. The CRM at Norland is not ineffective, it is unproven.
- Three patterns squander the benefit: claimed instead of measured, project costs instead of total costs, a date instead of impact as the definition of success.
- The pressure from finance is real (KPMG: only 7 per cent with an established ROI measure for AI; Deloitte: more than two thirds of organisations report that they only achieve their typical AI ROI after two or more years) — and it is an opportunity: the question is coming anyway; your only choice is whether you are ready with an answer.
Sources
- KPMG: "Global AI Pulse Q2 2026" (https://kpmg.com/xx/en/media/press-releases/2026/06/growing-adoption-signals-progress-as-cost-visibility-and-accountability-drive-ai-value.html) — survey of 2,145 executives from managing-director level upwards (over 40 per cent C-suite) in 20 markets, fieldwork 28 April to 25 May 2026; source of the 7 per cent and 42 per cent findings
- Deloitte Germany: “Deutschland im KI-Paradox”, March 2026 (fieldwork 15 August to 5 September 2025) — 1,854 executives across 14 countries, 249 of them in Germany. Only organisations that use AI in daily operations or run generative-AI pilots were surveyed. The underlying study is “The ROI of AI: The paradox of rising investment and elusive returns – German Cut”. The Deloitte page does not state the sample base for the two-thirds finding. Primary source: https://www.deloitte.com/de/de/issues/innovation-ai/deutschland-im-ki-paradox.html