I have read a lot of AI project proposals. The ones that do not get approved almost always fail for the same reason: they describe a technology solution when what the approver needs is a business case.
A CFO or CEO approving an AI investment is not asking ‘is this technically feasible?’ They are asking: what problem does this solve, how much will it cost, how will we know it worked, and what happens if it does not. An AI brief that leads with architecture diagrams and model selection has answered the wrong questions.
The data on what happens when briefs do not get this right is instructive. Deloitte’s research found that 56% of AI projects lose C-suite sponsorship within six months – almost always because the business case was not sufficiently clear at approval for sponsors to defend it when delivery became complicated. Projects with sustained executive sponsorship achieve a 68% success rate; those that lose it drop to 11%.
The Six Questions Every AI Project Brief Must Answer
1. What specific problem are we solving?
Not ‘we want to use AI for customer service’ but ‘we handle 4,200 inbound customer queries per month, 60% of which are routine status requests that take an average of 8 minutes each to process manually. We want to automate the handling of those 2,520 monthly queries.’ Specificity at this stage is what makes the rest of the brief credible.
2. What does success look like and how will we measure it?
Define the metric before the project starts. If the AI is handling customer queries, the metrics might be: query resolution rate above 85%, average handling time under 90 seconds, customer satisfaction score maintained or improved. These numbers need to be agreed and baselined before any development work begins. Briefs that leave success measurement to ‘after deployment’ almost never produce it.
3. What data do we need and do we have it?
Gartner’s February 2025 research found that 85% of AI projects fail due to poor data quality or missing relevant data. The data question is not an implementation detail – it is a viability question. A brief that cannot specifically name the data sources the AI will use, confirm access to them, and describe their quality status has not yet established that the project is feasible.
4. What is the implementation approach and timeline?
Not a full technical specification, but a clear description of the phases: data preparation, model development, integration, testing, pilot deployment, and full rollout. Each phase should have a realistic duration and a defined exit criterion – what needs to be true before the project moves to the next phase. This is what allows a sponsor to understand what they are approving and what governance checkpoints exist.
5. What are the risks and how will we manage them?
Every AI project has at least three risk categories: data risk (the data is worse than expected), model risk (the AI does not perform at the required accuracy), and adoption risk (the intended users do not use it). A strong brief names these risks honestly and describes the mitigation approach for each. Briefs that present no risk are not credible and do not survive scrutiny from a board or a CFO.
6. Who owns the outcome?
AI projects without a named business owner – not a technical lead, but a business leader who is accountable for the outcome – lose sponsorship. The brief should name the business owner, define their accountability, and confirm their active involvement in the project. This person is the one who will be defending the investment when delivery gets complicated.
What a Good Brief Does Not Include
Model architecture details, vendor comparisons, or technical infrastructure specifications do not belong in an approval brief. They belong in a technical specification that follows approval. Mixing them into the brief obscures the business case and gives approvers the opportunity to stall on technical questions they are not qualified to evaluate. Keep the brief focused on business logic. Get it approved. Then go technical.
Where We Come In
DoSystems helps businesses scope and brief AI projects before any development begins. The brief is where the project either gets real or stays conceptual – and getting it right at this stage costs a fraction of what it costs to correct a poorly scoped project mid-build. DoSystemsInc.com
Frequently Asked Questions
What should an AI project brief include?
An AI project brief should answer six questions: what specific problem is being solved, how success will be measured, what data is needed and whether it is available, what the implementation phases and timeline are, what the key risks are and how they will be managed, and who the named business owner is.
Why do AI projects lose executive sponsorship?
Deloitte’s research found 56% of AI projects lose C-suite sponsorship within six months – most often because the business case at approval was not specific enough for sponsors to defend when delivery became complicated. Clear, measurable outcomes defined upfront are the most effective way to maintain sponsorship.
What is the most common AI project brief mistake?
Describing the technology solution instead of the business case. Approvers need to understand what problem is being solved and how they will know it worked – not which model was selected or what the architecture looks like.
How long should an AI project brief be?
A well-structured AI project brief for an SMB project can be 3–5 pages. Longer briefs do not improve approval rates. Clarity on the six core questions is what drives approval, not document length.
#AIStrategy #AIProjects #AIImplementation #DoSystems #AIConsulting #BusinessAI #AIPlanning
Comments are closed