AI everywhere, measurable benefit nowhere? Put it where it pays.

Leaders are thinking about AI the wrong way. Not “what tech can we try?”, but “where will AI generate measurable benefit and ROI for my business?” Until that flips, you will just add more tools and more cost, and not achieve meaningful business outcomes.

What you’d see if you are in “what tech can we try” mode

If your organisation is in scattergun AI mode these are four tell-tale signs you will see:

• AI technical pilots everywhere with no clear business purpose or target benefits, and placement decisions living solely in IT.

• Business perspective missing — the why. The business owners aren’t defining the business need based on improving the business operating model.

• Solution fixation. AI discussions revolve around functional and non-functional AI considerations (latency, reliability, safety, privacy, bias, explainability), while the why and where aren’t discussed.

• Spend rising with no measurable return. Lots of AI costs being added with little measurable business return to show for it.

Likely root cause

The root cause is a focus on the solution, not the need. Because AI helps fundamentally reshape the core business model, adoption at scale needs a clear definition of the why and where before there is any focus on the what.

• Skills gap: too few people who know how to define the “why” – strategic direction, mission/vision, £ benefits/ROI logic, and placement in the operating model.

• Method/Tooling gap: no standard way to define mission/vision, turn them into £ benefit formulas with baselines/owners, and map them to a steerable AI rollout.

• Topic capture by tech: it’s natural for a new technical capability to “live” in IT, but when AI becomes all about the tech, the business detaches and the why/where gets no focus. (technical/operational realities matter, but they’re secondary to business placement and prioritization.)

Impact — why it matters

Leading organisations including MIT Sloan Management Review with BCG, Harvard Business Review, and RAND report that only ~10% of companies realise significant financial benefits from AI, and estimates put AI project failure rates at 80–90%, driven primarily by the lack of a clear, quantified why, a shortage of skills and method to steer by ROI, and underestimation of full TCO and operating needs.

AI transformation at scale is extremely expensive — new data plumbing, expensive specialist resources, heavy compute, always-on operations, vendor fees — and you still have to change or remove old work to actually bank the benefits. Getting it wrong is a huge, multi-year, multi-million-pound step backwards - but 80-90% will be in this position.

What to do now

AI is strong at the solution — finding patterns, ranking options, scaling decisions, running the ship once rules exist. It’s weak at the why and where. Only people understand the full business model, the broad competitive landscape, and the trade-offs. Only people can set the business logic that tells AI where to go and what “good” means. In an AI future, that’s where people will live: up front, defining the why, the placement, the business needs, the rules that need to be followed — not buried in solution building land.

6 Key focus areas to ensure AI at scale succeeds include:

• Make the why explicit: be crystal clear on mission, vision, £ benefit targets, and who owns them. Everyone should know what “good” looks like for AI in business terms and by when.

• Focus on the Business Model: use a whole-business view (where money’s made/saved; where constraints sit) to decide where AI goes first — biggest bang for the buck, not the shiniest tech.

• Connect the Why to the What: create one connected view that ties the target business results from AI to the AI execution roadmap.

• Iterate from low-hanging fruit: prove AI on a narrow slice with a high ROI, bank the benefits, then move to the next highest return.

• Specify requirements clearly: capture functional needs and non-functional guardrails — accuracy, latency, privacy, safety, bias, explainability, human-in-the-loop — but don’t let them drown out where AI should be placed and the target business return.

• Track benefits from start to finish: see it through. Don’t just ship the tech — remove the work it replaces and take the cost out. Keep measuring until the £ is actually banked.

Close

Do you see this challenge in your organisation? Can you see how a Benefit-Led Predictive approach would help manage it?

If this topic resonates, Cedus are hosting a series of The ‘Why’ in AI round-tables, where business leaders come together for a facilitated discussion on the topics raised here and share thoughts and ideas. If you would like to participate please drop us a line.

AI pilots everywhere