How UK CIOs Should Measure AI ROI Before Budgets Are Cut
UK CIOs should measure AI ROI with a baseline-to-KPI framework and named business sponsors, not experimentation metrics, says Foundry’s 2026 State of the CIO survey.
Enterprise AI spending has moved past the pilot stage, but proof of value has not kept pace. According to Foundry’s 2026 State of the CIO survey, just 19% of respondents say their AI initiatives have met or exceeded business goals, and 18% report that fewer than a third of their AI use cases are meeting defined expectations. For CIOs facing a tighter 2027 budget round, closing that gap is no longer optional — it is the difference between AI programmes that scale and AI programmes that get quietly defunded.
Why Is AI ROI So Hard to Prove in 2026?
Most organisations still treat AI value the way they treated early cloud migration: as a technology deployment rather than a funded business case. That approach breaks down once boards start asking for numbers. According to the same Foundry survey, 32% of CIOs cite ill-defined ROI metrics as a hurdle to scaling AI, close behind murky corporate AI strategy (31%) and a shortage of in-house expertise (40%).
The root problem is sequencing. Teams build models first and ask what they’re worth afterwards, which makes ROI a retrospective argument rather than a designed outcome. A credible framework reverses that order: define the business metric, set a pre-AI baseline, then only fund use cases with a plausible path to moving that metric within a fixed window.
Executive Action:
- Audit current AI use cases against a single question: what pre-AI baseline was recorded before build began?
- Flag any use case with no named business metric and pause further investment until one is agreed.
- Use INFORMD’s AI governance self-assessment to benchmark where your organisation sits against peers.
What Should a CIO’s AI ROI Framework Actually Measure?
A workable framework measures three things: the size of the benefit, the speed at which it appears, and the confidence attached to the number. Operational efficiency and process improvement remain the most commonly tracked outcome, cited by 40% of CIOs in Foundry’s research, followed by employee productivity (34%) and cost reduction (30%). Revenue or growth impact trails at 27%, largely because it takes longer to attribute and is easier for a finance director to challenge.
CIOs building this out for the first time should avoid over-engineering it. A single-page scorecard per use case — baseline, target metric, 90-day and 180-day checkpoints, and a named business and technical sponsor — is more useful to a board than a dashboard with forty KPIs nobody reviews. Stage-gated funding tied to those checkpoints, rather than to delivery milestones like “model built,” keeps spend honest and gives the CIO grounds to kill projects that stall.
Executive Action:
- Standardise a one-page ROI scorecard covering baseline, target metric, and 90/180-day checkpoints for every live AI use case.
- Require a named business sponsor and technical sponsor who jointly own the outcome, not just the build.
- Tie funding release to checkpoint evidence, not to development milestones.
How Should CIOs Structure Governance to Protect ROI?
Governance structures are ahead of measurement in most enterprises, which is itself a warning sign — organisations are better at controlling AI than at proving it pays. Foundry’s data shows 83% of IT leaders already have, or are building, a cross-functional AI steering committee, and 53% have a formal project approval process in place. Without a matching ROI discipline, those committees end up approving projects on strategic merit alone, with value tracking bolted on later or never.
The CIOs getting the best results are replacing centralised AI centres of excellence — which tend to become clearinghouses nobody owns — with embedded delivery squads inside the business units accountable for the metric being moved. That structure forces the same team that requested the use case to defend its return, which sharpens both the initial business case and the willingness to shut down projects that miss two consecutive checkpoints.
Executive Action:
- Review whether your AI steering committee has an ROI mandate, or only an approval and risk mandate.
- Consider embedding AI delivery capability inside business units rather than a single central CoE.
- Set a rule to formally review — and where needed kill — any use case that misses two consecutive checkpoints.
What Should CIOs Report to the Board?
Boards do not need every KPI a CIO tracks; they need a small number of numbers that answer one question — is AI spend converting into business value, and how fast? A quarterly board pack built around portfolio-level metrics (percentage of use cases meeting checkpoint targets, aggregate value delivered against forecast, and use cases killed versus scaled) gives non-executives a governance handle on AI spend without requiring technical fluency. This complements the reporting discipline INFORMD has set out previously in its guidance on AI risk reporting for the board and on building digital transformation ROI governance.
CIOs should also be candid about what the data currently shows. With only 47% of organisations having established formal AI metrics at all, a board that receives a polished ROI report should ask how that figure was produced, and whether it would survive scrutiny from the audit committee. Overstating maturity here creates a credibility problem that is harder to fix than a disappointing number honestly reported.
Executive Action:
- Build a quarterly, portfolio-level AI value pack for the board rather than project-by-project updates.
- Disclose methodology alongside any ROI figure so the audit committee can test it.
- Benchmark your reporting maturity using INFORMD’s technology strategy review template.
INFORMD provides intelligence briefings, tools and frameworks for senior business leaders across technology, finance, strategy and compliance. Based in Milton Keynes, UK, we help executives stay informed and act with confidence. Explore our full briefing library (/resources/) or access our free assessment tools (/tools-assessments/). Watch for upcoming video briefings on this topic at /videos/, or get in touch via /contact/.
Stay ahead. Subscribe to INFORMD’s weekly executive briefing at informd.co.uk (/resources/).
There is no single benchmark, but Foundry’s 2026 State of the CIO survey found only 19% of organisations report AI meeting or exceeding business goals. Treat any use case with a documented baseline, a moved metric, and a positive stage-gate review at 90 and 180 days as ahead of the field.
Most projects are funded before a business metric or baseline is agreed, so ROI becomes a retrospective justification rather than a designed target. Foundry’s research found 32% of CIOs cite ill-defined ROI metrics as their top barrier to scaling AI.
Ownership should sit jointly with a named business sponsor and technical sponsor for each use case, rolling up to the CIO for portfolio-level reporting. Embedding delivery inside business units, rather than a central AI centre of excellence, sharpens accountability for the number.
Quarterly, at portfolio level, is sufficient for most boards: percentage of use cases meeting checkpoints, aggregate value against forecast, and projects killed versus scaled. Project-by-project updates should stay at executive committee level, not board level.
