How to Build an AI Business Case

A rigorous AI business case has four pillars. Each pillar requires a number — not a feeling. The ROI Engine in the next lesson will help you calculate these for your specific situation, but understanding the structure first will help you interpret and use the output.

Pillar 1: Baseline Cost

What does the current-state process cost? This includes direct costs (staff time, error correction, rework, downtime) and indirect costs (opportunity cost of slow processes, customer attrition from poor service, compliance risk from manual errors). The test: Could you defend this number to a CFO?

Pillar 2: Automation Potential

What percentage of the baseline cost can AI realistically eliminate or reduce? This is not the vendor's claimed automation rate — it's the rate achievable in your specific operational context, accounting for your data maturity, your process complexity, and the learning curve of deployment. Conservative rule of thumb: Apply a 50–60% realisation factor to any vendor-quoted automation rate for a first deployment.

Pillar 3: Full Implementation Cost

What does it actually cost to build, deploy, and maintain this AI? This includes: vendor or development cost, data preparation and labelling, integration with existing systems, staff training and change management, ongoing model maintenance and retraining, and the hidden cost of the internal team time required. The mistake: Most business cases include the vendor cost but omit data preparation and internal time. These can double the actual cost.

Pillar 4: Time-to-Value

When do the savings actually start? AI projects typically follow a curve: months 1–2 are data preparation and integration. Months 3–4 are pilot deployment with limited scope. Months 5–6 are full deployment with performance below steady state. Month 7+ is steady-state performance. A business case that shows full savings from month 1 is not credible. A business case that models the ramp-up curve is.