Most firms aren't measuring real financial returns on AI
A new industry report from Cynozure shows a hard truth: only 15% of organisations can point to pounds or dollars when asked what AI has delivered. Meanwhile, adoption is now mainstream across both traditional and generative AI, and interest in agentic AI is climbing, with 52% using it or planning to.
This gap between activity and proof is exactly where boards are pushing. Without consistent measurement, AI spend becomes an article of faith rather than a line on the P&L.
What the report found
- Sample: 60 senior leaders across retail, financial services, consumer goods, tech and non-profits; nearly half from organisations above £1B in revenue.
- Measurement: 30% do not measure value consistently; only 15% quantify financial impact in currency.
- Adoption: Traditional and generative AI in wide use; agentic AI interest rising (52% using or planning).
- Impact focus: Effort concentrates on automation and productivity more than growth.
- Ownership: 80% give data strategy to a CDO/Head of Data, but only 28% give AI strategy to the same role; 40% split AI ownership across multiple execs, 17% have no clear owner.
- Priorities and blockers: Data culture and literacy top the 2026 priority list (43%). Budget/resource limits lead as blockers (25%), with legacy tech (20%) and lack of buy-in (17%) holding larger firms back.
"AI has propelled data into the boardroom. The challenge now is not whether organisations can use data and AI, but whether it is making a meaningful difference to the P&L." - Jason Foster, Founder and CEO, Cynozure
Why this matters for your P&L
Boards don't fund demos; they fund returns. If AI is everywhere but ROI is nowhere, the issue isn't technology. It's ownership, measurement, and commercial discipline.
Teams have invested in platforms and pilots. The next move is to treat AI as a product portfolio, tie each product to a specific outcome, and report results in financial terms. Anything less will struggle to get continued funding in 2026.
A practical playbook for executives
- 1) Assign one accountable owner for AI
Pick a single executive (CDO/CAIO/CTO) with end-to-end accountability for AI outcomes, not activity. Stand up a cross-functional AI steering group that includes Finance, Risk, Legal, and product lines, with clear RACI and decision rights. - 2) Run AI as a product portfolio
Group initiatives into Run (efficiency), Grow (revenue), and Control (risk/compliance). Fund with stage gates: problem framing → pilot with baseline → scale with target ROI → operate with quarterly benefit audits. Stop what isn't clearing the hurdle rate. - 3) Put a price on outcomes (RODI)
Define Return on Data Investment: net benefits (revenue uplift, cost out, cost avoidance, risk reduction) minus total cost of ownership (people, data, models, platforms, change). Baseline processes pre-AI, instrument for uplift, and agree with Finance on how to count avoided cost and reduced risk. - 4) Balance efficiency with growth
Don't stall at automation. Build a queue of growth bets: pricing optimisation, cross-sell propensity, churn prevention, lead scoring, real-time next-best-action, and new product features powered by AI. Tie each to a revenue model and sales motion. - 5) Standardise measurement
Report monthly at the portfolio and product level: benefits realised vs. plan, run-rate savings, revenue impact, payback period, IRR, and risk posture. Make one dashboard the single source of truth shared with Finance. - 6) Fix ownership gaps in delivery
Make product owners accountable for P&L-linked metrics. Pair them with data scientists/engineers and process owners. Incentives should include realised benefits, not just delivery milestones. - 7) Build decision products, not just data products
Wrap data and models around a specific decision (e.g., approve/decline, allocate/route, set price). Define the decision window, thresholds, overrides, and feedback loops. This is where value compounds. - 8) Fund change, not just models
Budget for process redesign, training, and sales enablement. Most ROI is lost in last-mile adoption. Treat change management as non-negotiable scope. - 9) Tackle legacy constraints
Sequence platform upgrades where they unlock the biggest benefits path. Don't wait for a perfect data estate-ship value in bounded domains with clear data contracts and MLOps basics. - 10) Govern with lightweight, auditable guardrails
Adopt an enterprise standard for model risk, monitoring, and human-in-the-loop. If you need a reference, see the NIST AI Risk Management Framework.
What leaders are saying
"Data products, and increasingly decision products, are how leaders are turning strategy into reality... and track Return on Data Investment (RODI) in a way that resonates with boards and investors." - Tim Connold, Chief Client Officer, Cynozure
Your executive scorecard (use this)
- Portfolio mix: % Run / Grow / Control by spend and value
- Realised value YTD: £ savings, £ revenue, £ risk reduction
- Run-rate value next 12 months
- Payback and IRR by product
- Adoption: % assisted/automated decisions, override rate
- Data and model quality: drift alerts, incident count, SLA hits
- Unit economics: cost per decision/insight
- Time-to-value: idea → pilot → production
- Compliance: documented use cases, risk ratings, reviews on time
- Team: % of staff certified or trained for their role
Priorities and blockers: how to unblock fast
- Data culture and literacy (43% priority): Stand up role-based training for leaders, product owners, and frontline teams. Tie training to live use cases and the dashboard they'll use weekly. If you need curated options, see executive and team AI courses.
- Budget and resources (25% blocker): Reallocate from low-yield pilots to products with verified baselines. Introduce a "prove-and-scale" fund with strict stage gates and sunset rules.
- Legacy tech (20%): Isolate domains, use data products with clear contracts, and modernise incrementally where ROI is highest.
- Lack of buy-in (17%): Present value stories in currency with pre/post metrics. Start with one decision product that pays back in a quarter and socialise the win.
Bottom line for 2026
AI activity is no longer the signal. Financial proof is. Move ownership to one accountable leader, frame AI as decision products, and measure RODI with the same discipline you use for any commercial bet.
Do that, and AI funding gets easier, not harder. Don't, and it stays stuck in experimentation while competitors take the margin.
Survey details: Findings are from Cynozure's 2026 State of the Industry Report, "The Next Horizon: Data, AI and Impact," based on responses collected between late October and late November 2025 from CDOs, CDAOs, CDAIOs, VPs/directors of data, heads of data, and other senior executives.
Your membership also unlocks:
