Most AI business cases fail not because the idea is wrong, but because they're written for technologists rather than decision-makers. A board or exec team approving an AI investment needs to understand the business outcome, the cost, the risk, and the timeline — not the technical architecture.
Start with the business problem
The strongest AI business cases start from a specific, measurable business problem rather than from a technology. 'We want to implement AI' is not a business case. 'We spend 40 hours per week on manual scheduling and it causes 15% of our customer service failures' is the start of one. Define the problem in business terms first, then explain why AI is the right solution — and why now.
The elements a board needs to see
A credible AI business case for board approval needs to include: the problem and its current cost to the business; the proposed solution at a level non-technical stakeholders can evaluate; the expected ROI with realistic assumptions; the implementation costs (technology, team, time, and change management); and the risks with mitigations. Boards are increasingly sceptical of AI ROI projections — build yours conservatively and explain your assumptions.
- What problem does this solve and what is its current cost?
- What specifically will the AI do — and what will it not do?
- What is the realistic ROI and over what timeframe?
- What are the implementation costs, including people and change management?
- What are the risks and how will they be managed?
Common mistakes
The most common mistakes are overstating the ROI with optimistic assumptions, underestimating the time and cost of implementation (especially data preparation and change management), and presenting the case as purely positive with no honest discussion of risk. Boards trust business cases more when the presenter has clearly thought through what could go wrong.
Want to talk through any of this?
The best AI business cases are written by people who understand both the technology and the business — and who are honest about the difference between potential value and certain value.
Get in touch →