AI Readiness & Value Map
Most AI portfolios go wrong at selection, not engineering: use cases are picked for visibility, with no baseline, no run-cost estimate and no condition for stopping. A value map is worth funding only if every item on it has an owner, a measured starting point and a reason it could be cut.
- Respondents whose organization uses AI in at least one business function, 2025 (McKinsey survey, reported in the Stanford AI Index 2026)
- 88%
- Respondents whose organization uses AI in at least one business function, 2025 (McKinsey survey, reported in the Stanford AI Index 2026)
- Stanford Institute for Human-Centered Artificial Intelligence (HAI), 2026(opens in a new tab)
- Annual labor productivity growth executives attribute to AI over the past three years (nearly 9,000 firms in seven countries, surveyed 2026)
- 0.2%
- Annual labor productivity growth executives attribute to AI over the past three years (nearly 9,000 firms in seven countries, surveyed 2026)
- National Bureau of Economic Research, 2026(opens in a new tab)
- Leaders in high-maturity vs low-maturity organizations whose AI projects stay in production three years or more (Gartner survey, Q4 2024)
- 45% vs 20%
- Leaders in high-maturity vs low-maturity organizations whose AI projects stay in production three years or more (Gartner survey, Q4 2024)
- Gartner, 2025(opens in a new tab)
The problem
AI use is now wide, but the returns are thin. In McKinsey's 2025 survey, charted in the Stanford AI Index, 88% of respondents said their organization uses AI in at least one business function, up from 78% in 20241. Yet in a 2026 survey of nearly 9,000 executives in seven countries, firms reported that AI had raised annual labor productivity growth by only 0.2% over the previous three years2. Measured more narrowly, 19.8% of US businesses said they had used AI in the past two weeks as of May 2026, and 33.9% in finance and insurance3.
The losses happen early. Gartner predicted that at least 30% of generative AI projects would be abandoned after proof of concept by the end of 2025, due to poor data quality, inadequate risk controls, escalating costs or unclear business value4. In a separate Gartner survey of 432 organizations, 45% of leaders in high-maturity organizations said their AI projects stay in production for three years or more, against 20% in low-maturity ones. Gartner tied the difference to choosing projects on business value and technical feasibility. In low-maturity organizations, 37% of leaders named finding the right use case as a top barrier5.
Good looks like the Map function of the NIST AI Risk Management Framework applied to a portfolio: each use case has a documented purpose, expected benefits and the potential cost of its errors before anyone builds it, and a point at which someone decides whether development should proceed6. High-maturity organizations in Gartner's survey did the arithmetic: 63% of their leaders run financial analysis on risk factors, conduct ROI analysis and measure customer impact5. The output is a short list you can defend in a budget meeting, not a long list of ideas.
What you get
- 01
Use case inventory with baselines
Candidate use cases gathered from your business units, each with the process it changes, today's cost, volume or cycle time, and the person who owns that number.
- 02
Value and feasibility scoring
Every candidate scored on the same criteria: value, data readiness, integration effort, regulatory exposure and run cost. One method across business units means the ranking holds up when budgets are argued.
- 03
Data and platform check for the top candidates
For the use cases that rank highest, a check of whether the data exists, can be reached and is fit for the purpose, and what has to be fixed before work starts.
- 04
Risk classification per use case
Each use case mapped to the rules that apply to it, such as model risk guidance or the EU AI Act high-risk categories, so compliance effort is in the plan from day one.
- 05
A prioritized plan you can fund
The first use cases in sequence, with stop-or-scale criteria for each, the team and budget they need, and the decisions required from leadership.
Where it goes wrong
The failure modes we design against from the first week, and the control for each.
Use cases chosen for visibility
Demos and executive enthusiasm pull the portfolio toward showcase projects with no baseline and no owner. Gartner found that high-maturity organizations choose projects on business value and technical feasibility5. The control: score every candidate on the same criteria and require a measured baseline before funding.
Usage counted as value
Adoption numbers rise long before results do: 78% of firms in a 2026 multi-country survey used AI, while executives put its effect on annual productivity growth at 0.2%2. The control: agree one financial or operational metric per use case with finance, and report it next to usage.
Run cost left out of the business case
Pilots hide the cost of inference, integration, monitoring and support. Gartner listed escalating costs among the main reasons generative AI projects are abandoned after proof of concept4. The control: estimate run cost per transaction or per user for each use case and recheck it at every stage gate.
Regulatory exposure found late
A use case in a high-risk category carries documentation, data governance and oversight duties that change its cost and timeline. Under the EU AI Act as amended in 2026, obligations for stand-alone high-risk systems apply from 2 December 20278. The control: classify every use case against the rules that apply before it is ranked.
Standards and rules
NIST AI Risk Management Framework (AI RMF 1.0)
Primary text on nist.gov(opens in a new tab)Its Map and Manage functions ask you to document each AI system's purpose, benefits and potential costs and to decide whether development should proceed; it is voluntary and is being revised under the White House AI Action Plan.
ISO/IEC 42001:2023
Primary text on iso.org(opens in a new tab)Sets requirements for an AI management system, and the value map supplies the objectives, inventory and risk treatment decisions that system needs.
EU AI Act, as amended by the Digital Omnibus on AI (Regulation (EU) 2026/1744)
Primary text on europarl.europa.eu(opens in a new tab)High-risk obligations apply from 2 December 2027 for stand-alone systems and 2 August 2028 for systems embedded in products, so the map should flag high-risk use cases before they are funded.
How to measure it
- Share of funded use cases with a measured baseline and a named business owner
- Realized value against the business case for each use case, reviewed with finance every quarter
- Time from funding approval to first production release
- Run cost per transaction or per user against the original estimate
- Share of use cases stopped or redirected at a stage gate
Questions buyers ask
What is an AI readiness assessment?
It is a structured look at where AI can create measurable value in your business and what stands in the way. A useful one covers use cases, data, platforms, skills and regulatory exposure together, and ends with a ranked list of use cases, the gaps each one depends on, and a plan you can put a budget against. If it ends with a maturity score and no ranked list, it has not done its job.
How should we prioritize AI use cases?
Score every candidate on the same few criteria: the value at stake against a measured baseline, whether the data exists and is fit for purpose, the integration effort, the regulatory exposure and the cost to run it. Weight them openly, rank, and fund the top few with clear stop-or-scale criteria. Re-score as you learn, because data and cost estimates are usually wrong at the start.
How do we measure ROI on AI?
Measure the business metric the use case was meant to move, such as cost per case, cycle time, error rate or revenue per customer, against a baseline taken before launch. Count the full cost, including inference, integration, monitoring and the people who review outputs. Usage and satisfaction scores are useful signals but they are not returns. Agree the method with finance up front so the result is accepted.
Sources
Every figure on this page comes from the sources below. We checked each link on .
- 1.Artificial Intelligence Index Report 2026, Chapter 4: Economy(opens in a new tab)Stanford Institute for Human-Centered Artificial Intelligence (HAI), 2026 · hai.stanford.edu
- 2.Firm Data on AI (NBER Working Paper 34836)(opens in a new tab)National Bureau of Economic Research, October 2026 · nber.org
- 3.Large Firms With at Least 20 Employees Biggest AI Users(opens in a new tab)U.S. Census Bureau, May 26, 2026 · census.gov
- 4.Gartner Predicts 30% of Generative AI Projects Will Be Abandoned After Proof of Concept By End of 2025(opens in a new tab)Gartner, July 29, 2024 · gartner.com
- 5.Gartner Survey Finds 45% of Organizations With High AI Maturity Keep AI Projects Operational for at Least Three Years(opens in a new tab)Gartner, June 30, 2025 · gartner.com
- 6.AI Risk Management Framework (AI RMF 1.0)(opens in a new tab)National Institute of Standards and Technology, January 26, 2023 · nist.gov
- 7.ISO/IEC 42001:2023 Information technology: Artificial intelligence: Management system(opens in a new tab)International Organization for Standardization (ISO) and IEC, 2023 · iso.org
- 8.Digital Omnibus on AI (Legislative Train Schedule)(opens in a new tab)European Parliament, July 24, 2026 · europarl.europa.eu