The Business Case for AI: Calculating ROI Before You Commit

← Back to AI Insights

Most AI investments fail not because the technology doesn't work — but because the business case was never properly built. Executives approve projects based on vendor demos and analyst hype, only to find that the promised returns evaporate when they encounter the realities of data readiness, integration complexity, and change management.

This guide provides a structured framework for building a credible AI business case: one that stands up to scrutiny from your CFO, satisfies your board, and gives your implementation team a meaningful target to hit.

Why Most AI Business Cases Fail

The typical AI business case makes three critical errors:

• Overestimating benefits. Vendor ROI calculators are built to impress, not to accurately model your specific operational context.
• Underestimating costs. Data preparation, integration, change management, and ongoing model maintenance are routinely omitted.
• Ignoring risk. A business case that doesn't account for the probability of failure, delay, or underperformance is not a business case — it's a wish list.

Step 1: Define the Problem with Precision

Before any financial modelling, you need a precise problem statement. Vague objectives produce vague returns. The best AI business cases are anchored to a specific operational bottleneck with measurable consequences.

Notice how this framing immediately surfaces the data you need to build a financial model: volume, time, cost-per-error, and a directional hypothesis about improvement.

Step 2: Map All Costs — Including the Hidden Ones

A complete AI cost picture includes:

• Data preparation: Cleaning, labelling, structuring, and integrating data sources. In our experience, this accounts for 40–60% of total project cost.
• Model development: Engineering, training, validation, and testing.
• Infrastructure: Cloud compute, storage, API costs, and monitoring tooling.
• Integration: Connecting the AI system to your existing workflows, systems, and UI.
• Change management: Training staff, updating processes, and managing the human side of adoption.
• Ongoing maintenance: Model drift monitoring, retraining, and version management. Budget 15–25% of initial build cost per year.

Step 3: Quantify Benefits Conservatively

Apply a structured discount to all benefit projections. We recommend:

• Take your best-case estimate of efficiency gain or revenue uplift.
• Apply a 30–40% discount for adoption lag (staff take time to use new tools fully).
• Apply a further 20% discount for model underperformance relative to prototype.
• Use this discounted figure as your base case.

If the business case still works at these conservative numbers, you have a robust investment thesis.

Step 4: Calculate Payback Period, Not Just ROI

ROI figures can be misleading without a time dimension. A 300% ROI over five years is very different from a 300% ROI over 18 months. For most AI investments, we recommend targeting a payback period of 12–24 months for tactical projects and 24–36 months for transformational programmes.

Step 5: Stress-Test Your Assumptions

Build a simple sensitivity table showing how your returns change if key assumptions are wrong. What happens if data preparation takes twice as long? What if adoption is 50% lower than projected? If the business case only works under the most optimistic scenario, it needs to be reconsidered.

---

Building a rigorous AI business case takes time — but it's the work that separates successful AI programmes from expensive experiments. If you'd like help building one for your organisation, our team runs a structured Discovery Sprint that includes a detailed financial model tailored to your specific context.