AI use case prioritisation: settle the owner and the measure
Before a leadership team ranks AI use cases, it should decide who owns each chosen business problem and which measure will show the effect.
For each business problem chosen for investment, the session needs to settle who can act on the result, how it will be judged and what evidence is available now.
What should the leadership team rank?
Rank business problems, stated in terms of the result the company needs to improve.
Microsoft's plan for AI adoption opens with "Start with business problems" and asks for each problem to be turned into "a short statement that names the activity and the expected result", with the example "help support agents answer from internal documents so resolution time drops."
Next to each problem, note how AI is expected to improve the work, what it depends on and who owns it, and rank the list into an opportunity map for the leadership team. Before the session, test that ranking with someone outside the business, who will ask who owns each problem and how its effect will be measured.
What does your framework ask for on the owner and the measure?
They differ in who is accountable, what is measured and when it is agreed.
| Framework | Written for | Ownership and accountability | Business outcomes and measurement |
|---|---|---|---|
| Microsoft, Cloud Adoption Framework: AI adoption and Plan for AI adoption | Written for Organisations adopting AI | Ownership and accountability Leaders "provide visible sponsorship" for the priorities, and a strategy or charter documents "decision rights". Governance policies cover roles and responsibilities | Business outcomes and measurement "define specific success metrics, timelines, and resource requirements for each project", in the step before the proof of concept |
| Microsoft, Business Envisioning | Written for Software vendors choosing an AI use case to build | Ownership and accountability A question to consider: "Who is the executive sponsor for the solution? Who is accountable for the Objectives and Key Results (OKRs)?" | Business outcomes and measurement A question to consider: "How is success measured?" Key results then count toward the use case's strategic fit score |
| AWS Cloud Adoption Framework for AI, ML and Generative AI | Written for Organisations adopting AI | Ownership and accountability In the first phase, initiatives are tied to "key stakeholders (that is, senior individuals capable of influencing and driving change)" | Business outcomes and measurement In the same first phase, initiatives are tied to "measurable business outcomes" |
| Google Cloud's AI Adoption Framework | Written for Organisations building AI capability | Ownership and accountability For the first successful use case: identify the right business problem and "find the senior executive to own it" | Business outcomes and measurement At its most mature stage, the Automate section describes an automated A/B testing system "to evaluate the effectiveness of a newly released model service" |
| NIST, AI Risk Management Framework | Written for Any organisation that designs, develops, deploys or uses AI, on a voluntary basis | Ownership and accountability Roles and responsibilities for managing AI risks "are documented and are clear to individuals and teams throughout the organization" | Business outcomes and measurement "The business value or context of business use has been clearly defined", as part of the mapping that informs the first decision to proceed |
| Singapore, Model AI Governance Framework | Written for Organisations deploying AI solutions at scale, on a voluntary basis | Ownership and accountability "It is ideal for the models deployed in an intelligent system to have an internal departmental owner" | Business outcomes and measurement Organisations can start by deciding the "commercial objectives of using AI", then weigh them against the risks |
| UK government, AI Playbook | Written for Government and public sector organisations in the UK | Ownership and accountability "Nominate a senior responsible owner (SRO) who will be accountable for the use of AI in a specific project." | Business outcomes and measurement Early user research "will help establish a baseline of metrics to measure your project outcomes against." |
| US Government Accountability Office, AI Accountability Framework | Written for US federal agencies and other entities | Ownership and accountability "Define clear roles, responsibilities, and delegation of authority for the AI system" | Business outcomes and measurement Defining clear goals and objectives is "a key first step in the life cycle of an AI system", and they should be "specific and measurable" |
| US Office of Management and Budget, Memorandum M-25-21 | Written for US federal agencies | Ownership and accountability Officials responsible for AI risk decisions "at the lowest appropriate level"; a footnote encourages choosing officials "accountable for the mission outcome of the AI use case" | Business outcomes and measurement Before a high-impact use is deployed, an impact assessment states its "expected benefit, supported by specific metrics or qualitative analysis" |
Who should own a chosen use?
One person who answers for the result, named in the session before the work starts.
In a company, start with the manager responsible for the result in the problem statement. For customer-support resolution time, ask whether that manager can change the work, and which decisions still need another person's agreement. Record the AI team's delivery responsibilities separately from responsibility for the service result.
When should the measure be fixed?
Before the build, the pilot or the purchase begins, with a starting value to compare against.
The measure belongs to the business result in the problem statement, such as resolution time in the Microsoft example. A model's accuracy score does not, by itself, show whether customer-support resolution time fell.
Write down the measure, today's value and who reads it after the work. If today's value is unknown, agree who will establish it before the team commits to the proposed build, pilot or purchase. In the business case, keep the numbers you have measured apart from the ones you assume.
What about the ideas that are not on the list?
Give them room to be tried within clear limits, and keep them apart from the ranked investments.
A short list of centrally selected AI use cases is likely to miss a substantial part of the opportunity, because people closest to a workflow can see possibilities that are difficult to specify in advance.
Set the limits before anyone starts: which data a team can work with, which systems its tools can access, what it is allowed to change, and who is responsible if something goes wrong. When an experiment becomes something the business relies on, agree who owns it, how its effect is judged and how it will be supported. Put further investment through the same prioritisation process.
What should the session end with?
A short ranked list of business problems, each with one owner, an agreed measure and its starting value.
1. The ranked business problems, each with how AI is expected to improve the work, what it depends on and who owns it.
2. For each problem chosen for investment, the owner's name, the measure of effect and its value today, agreed before any build, pilot or purchase.
3. The evidence needed first, and who will bring it.
4. The limits for experiments outside the list: data, systems, what may change and who is responsible.
Sources
- AI adoption. Microsoft Learn, Cloud Adoption Framework, 15 October 2026.
- Plan for AI adoption. Microsoft Learn, Cloud Adoption Framework, 15 October 2026.
- Evaluate and Prioritize an AI Use Case with Business Envisioning. Microsoft Learn, 20 September 2024.
- AWS Cloud Adoption Framework for Artificial Intelligence, Machine Learning, and Generative AI. Amazon Web Services, 13 February 2024.
- Google Cloud's AI Adoption Framework. Google Cloud, undated.
- Artificial Intelligence Risk Management Framework (AI RMF 1.0). National Institute of Standards and Technology, January 2023.
- Model Artificial Intelligence Governance Framework, Second Edition. Infocomm Media Development Authority and Personal Data Protection Commission, Singapore, 21 January 2020.
- Artificial Intelligence Playbook for the UK Government. Government Digital Service, GOV.UK, 10 February 2025.
- Artificial Intelligence: An Accountability Framework for Federal Agencies and Other Entities. US Government Accountability Office, June 2021.
- M-25-21, Accelerating Federal Use of AI through Innovation, Governance, and Public Trust. Office of Management and Budget, 3 April 2025.
Facts last checked on 28 October 2026.