The person writing an AI business case has usually spent sixty days getting to this point. The readiness assessment is done. Opportunities are identified. One or two pilots are teed up. Now the case has to be written in a way that will survive a room full of people who have heard too many AI pitches and seen too few results.
The board’s scepticism is reasonable. BCG’s research on AI adoption shows a consistent gap between AI activity levels and measurable business impact. A delegate who ignores that gap and argues from enthusiasm will lose the room. One who builds the case around the board’s own concerns, using the board’s frame rather than the vendor’s, has a much better chance of leaving with a budget approval.
What does a credible AI business case actually contain?
A credible AI business case has four components that need to work together. The outcomes must be specific and tied to measurable KPIs. The timeline must be honest about what the technology can deliver and when. The plan must be phased so investment grows with evidence. The ROI frame must separate early indicators from financial results, because they operate on very different timescales.
The outcomes need to name specific problems in specific operations. “Reduce time spent on invoice processing by 40% in the first six months” is a business outcome. “Improve operational efficiency” is a statement of intent that gives a board nothing to evaluate. KPIs should also acknowledge which metrics are lagging indicators and will not move in year one. A board told to expect quarterly returns from AI, and then denied them, becomes a hostile audience by month three.
This is where the MIT research matters. The finding that roughly 95% of AI pilots fail to show measurable profit-and-loss impact is on the agenda of every risk-aware board. A case that acknowledges this failure pattern and explains how the programme is designed around it signals a level of preparation that few proposals reach.
Why does the dual-ROI frame matter more than a single number?
The single most common reason a well-intentioned AI business case fails to secure budget is that it promises financial returns on a timeline the technology cannot reliably meet. A dual-ROI frame separates two things that operate at different speeds. Trending ROI covers the early indicators that show an initiative is working, including process times, error rates, and adoption figures. Realised ROI covers the financial outcomes that typically take twelve to twenty-four months to appear.
Presenting these two frames separately inside the business case changes the board’s relationship with progress. A board given trending indicators it can read every quarter, alongside honest financial timelines, is far easier to keep aligned than one told results would arrive in six months and now asking why they have not.
Propeller’s research on AI ROI measurement suggests that some practitioners put the honest financial horizon closer to two to four years for businesses making their first serious AI investments. That reality is an argument for making the measurement plan credible, not for slowing the programme. A board briefed on realistic horizons will hold its nerve through the difficult middle period of an AI programme. One briefed on optimistic ones will pull the plug.
Including the dual-ROI frame in the written case also signals that the person who produced it has read enough to know where AI programmes typically fail and has designed around the failure point rather than hoping to be an exception.
How do you phase the plan without overpromising?
A phased plan organises investment into three horizons. Foundation runs from roughly month zero to six and focuses on what makes everything else possible. Expansion runs from month six to eighteen and extends what the foundation phase delivered. Deeper change, from month eighteen to thirty-six, is deliberately left more open because what the business needs then depends on what it has learned from the earlier phases.
In the foundation phase, the priority is data readiness, process mapping, and one or two tightly scoped pilots. Research on AI adoption patterns consistently shows that back-office automation, covering document processing, invoice handling, and scheduling, produces stronger early returns than customer-facing or sales-led initiatives. This is counterintuitive if the board’s expectations have been shaped by vendor demonstrations, but it is where the evidence points. Starting with high-volume, rule-based operations gives the programme the early results it needs to hold board confidence into phase two.
The expansion phase uses that evidence to justify a wider rollout. The deeper-change horizon is left deliberately open in the written case because committing to a specific programme three years out, based on what the technology can do today, is not rigour. A board that asks probing questions will recognise it as guesswork. Leaving the horizon open signals that the plan will respond to what is actually learned rather than to what was predicted, which is stronger planning than false precision.
Where do the board’s objections go in the document?
The most effective place to address board scepticism is inside the case itself, not in the question-and-answer session afterwards. A business case that anticipates the standard objections and builds answers into its own structure signals a level of preparation that few proposals reach. It changes the document’s register from advocacy to assessment, which is the register a board that has seen too many AI pitches will respect.
EY’s board governance research and Scaled Agile’s analysis of board AI questioning identify five questions that recur across ownership structures. First, how does the initiative align with the core business model? Second, how will ROI be measured and over what horizon? Third, what risks are being managed and how? Fourth, what internal capabilities are being built that will remain after the initiative? Fifth, what competitive advantage is being created?
Each question has a natural home in the business case structure. Alignment belongs in the opening rationale. ROI measurement belongs in the dual-ROI section. Risk belongs as a named section, not a footnote. Capability building belongs in the phased plan. Competitive context, handled carefully to avoid overstating what AI can deliver, belongs in the strategic opening.
Writing the document this way means the board’s objections are answered before they are raised. That changes the meeting from an interrogation into a conversation about implementation, which is a much better room to be in.
What does this case connect to across the wider mandate?
The business case document does not sit in isolation. It sits downstream of the readiness assessment that identified where genuine opportunities are, and upstream of the board reporting cadence that requires ongoing progress updates once budget is approved. The discipline of writing the case rigorously, with honest timelines and pre-empted objections, also sets the terms on which progress will be judged twelve months from now.
Poor data quality is cited by around three-quarters of organisations as the biggest barrier to responsible AI deployment, according to Gartner research. If that finding surfaced in the readiness assessment, it belongs in the business case as a named constraint with a plan for addressing it. A case that acknowledges its own obstacles is harder to hold against a delegate when timelines slip than one that presented no obstacles at all.
The mandate does not end with budget approval. The case written now, with its KPIs, its dual-ROI frame, and its phased horizons, becomes the accountability document the board will return to in twelve months. Write it for that conversation, not just for the one today.
