AI ROI: a board-ready framework for funding AI that pays back
Enterprises spent $30–40B on GenAI pilots and 95% saw no P&L return. Here's how to fund the 5% that pay back.
AMDIM · Strategy
June 10, 2026
Companies spent over $30 billion on generative AI pilots last year. MIT found 95% delivered no measurable P&L return. This framework covers how the 5% that pay back actually approach it.
The demo dazzles, the pilot ships, and months later finance asks what it actually saved — and the room goes quiet. This isn't a piece about whether to fund AI. It's about how to fund it the way the 5% do — with baselines, real costs, and a board-ready payback case.
The returns are real — they're just not evenly shared
Near-universal AI adoption has coincided with almost no bottom-line impact for the majority — and a tiny elite pulling dramatically ahead. A 2.8× gap between AI leaders and average adopters does not exist because leaders are using different models. It exists because they are taking a fundamentally different approach to measurement, sequencing and governance.
/ MIT's finding
The divide between winners and everyone else is not driven by model quality — it's driven by approach.
78% of organisations now use AI. Only 6% can trace more than 5% of EBIT to it. 4% have built genuinely advanced AI capability. The tools work. The money is being spent. The returns are missing for almost all of them.
It's not a technology problem — which is good news
Four Habits That Keep 95% of Companies From Seeing ROI
- 01
Never setting a baseline
Nobody wrote down what the process cost before the AI — so when the model improves things, there's no before-state to measure against, and the win stays invisible. Set the baseline before you write a single line of code.
- 02
Banking on soft benefits
Productivity and experience are real and valuable — but both are almost impossible to put in front of a board. Boards fund hard, cash-traceable savings tied to a named KPI with a named owner.
- 03
Leaving the workflow untouched
McKinsey found roughly 80% of organisations bolt AI onto existing processes — and workflow redesign turned out to be the single strongest predictor of whether EBIT actually moved. A faster version of a broken process is still a broken process.
- 04
Underestimating the full cost
The licence is the part everyone sees. Data preparation, integration, evaluation, governance, monitoring, and oversight are the part that quietly turns a six-month payback into eighteen. Price the whole iceberg.
Workflow redesign is the strongest predictor of impact
A faster version of a broken process is still a broken process. McKinsey found roughly 80% of organisations bolt AI onto their existing processes instead of redesigning them — and workflow redesign turned out to be the single strongest predictor of whether EBIT actually moved.
How do you actually measure AI ROI?
The Four Moves — In Order
Baseline the process
Measure what the target process costs today before a line of code: volume, hours, error rate, cycle time, cost per unit. 'We process 40,000 invoices a month, nine minutes each, with a 6% exception rate' is a baseline. Without it, every ROI number you produce is fiction.
Map to a value driver
Take your P&L apart into the levers that actually move it and tie the AI to one of them. Suddenly 'we built a chatbot' becomes 'we cut cost-to-serve by taking three minutes off average handle time.' One lever, one metric, one owner.
Price the real cost
The full iceberg — not just the API bill. Data preparation and integration, evaluation and testing, governance and compliance, production monitoring, human-in-the-loop oversight, and change management. The true cost is typically 2–4× the visible licence cost.
Express as a payback period
Most boards trust 'recoups in seven months, then £1.4m a year' far more than '320% ROI.' Payback in months is harder to game and easier to govern. Present a conservative case and an expected case — boards fund things they can defend.
Types of Value: How to Present Each
| Feature | Lead with (fundable) | Treat as upside |
|---|---|---|
| Hard cost savings — cash-traceable | ||
| Throughput gains at flat headcount | ||
| Error rate reduction / compliance | ||
| Productivity / time savings | ||
| Revenue attribution (clean only) | ||
| Experience improvements |
If you can't answer the value question, don't launch yet
Which number moves, by how much, by when. That single rule is most of the difference between a portfolio that compounds and one that stalls.
The cost you're not counting
The reason so many 'profitable' projects disappoint is that the business case only counted the visible costs. The licence is visible. What comes below the waterline is not — and it's larger.
The Full AI Cost of Ownership — Below the Waterline
| Cost layer | Typically counted? | Notes |
|---|---|---|
| Licence / API fees | Yes | The tip of the iceberg |
| Data preparation & integration | Rarely | Often 40–60% of true project cost |
| Evaluation & testing | Rarely | Skipping causes silent degradation |
| Governance, security & compliance | Sometimes | Mandatory; scales with risk tier |
| Monitoring & retraining | Rarely | Models decay; software doesn't |
| Human-in-the-loop oversight | Rarely | Required for high-stakes decisions |
| Change management & adoption | Almost never | Drives 88% vs 13% success delta |
Frequently asked questions
IDC and Microsoft's 2025 research puts the average at $3.70 returned per $1 invested — AI leaders achieve $10.30 per $1. A reasonable target for a well-scoped back-office automation project is payback within 12 months and 2–4× return over three years. Customer-facing and complex decision-support projects take longer to pay back but carry higher long-term ceilings.
MIT's research points to four recurring causes: no baseline measurement (so impact cannot be proven), reliance on soft benefits that don't translate to P&L, leaving the underlying workflow unchanged while overlaying AI on top, and underestimating the true total cost. The technology itself is rarely the problem — it's the approach.
Start with the visible costs: API and compute, licences, and implementation time. Then add the iceberg: data preparation and integration, evaluation and testing infrastructure, governance and compliance, production monitoring, human-in-the-loop oversight, and change management. The true cost is typically 2–4× the visible line items.
Cost savings are almost always easier to measure and faster to prove — the baseline is clear, the change is attributable, and the payback period is shorter. Revenue uplift attribution is genuinely hard: customer behaviour is multi-causal, and isolating the AI's contribution requires rigorous A/B testing that many organisations cannot run cleanly. Start with cost reduction; move to revenue uplift once you have the measurement infrastructure to do it honestly.
Which use cases will pay back first for your organisation?
The AI Maturity Scorecard takes ten minutes and shows you exactly where you stand — which use cases have the highest ROI, and what's quietly blocking the rest.
Take the AI Maturity Scorecard/ go deeper
Put this to work on your actual numbers.
A ten-minute assessment maps exactly where you are today — and what to do first.
/ keep reading
Agentic AI · 14 min
The Agentic AI Playbook: Architecture, RAG, Multi-Agent Systems, and What Actually Works in 2026
A practitioner's guide to the full agentic AI stack — from the agent reasoning loop to multi-agent orchestration, RAG pipelines, and the evaluation frameworks that keep it all honest.
Article · 7 min
Cloud-ready isn't AI-native — and that gap is where pilots die
Most AI pilots stall on infrastructure, not models. Here is the real gap between cloud-capable and AI-native, and how to close it without a rebuild.
Engineering · 10 min
Pilot to production: why most AI stalls — and the operating model that ships it
Only ~48% of AI projects reach production. The blocker isn't the model — it's the operating model. Here's the playbook that ships.