The Ai-pilot was successful and you gained good first results. Also the involved team is convinced for further roll-out. You prepared a slide-deck to present the results to the executive team and they asked for a proposal.

Then the proposal came back: the executive team had more questions, more data needed, and asked to "strengthen the business case." And now six months later, the pilot is still a pilot.

This happens constantly. Because the business case was written for the wrong reader, and not just because the CFO was skeptical about AI.

Most AI business cases are written to persuade people who already want to be persuaded. They document what the technology does, how the pilot performed and what the roadmap for further roll-out looks like. However, these are nice presentations and not investment cases to make decisions upon.

A CFO reading it on a Tuesday afternoon with six other things to review doesn't need to be convinced AI is interesting. They need to know three things: what does this cost, what does it return, and what happens if we don't do it. In numbers and on one page.

Those three questions sound simple. Almost nobody answers them properly.

Where most business cases break down

The failure usually happens at the same point: translating AI output into financial terms.

"Time saved" is the most common metric in an AI business case. It's also the least useful. A CFO reading "this will save 200 hours per month" has to do the work of translating that into something that appears on a P&L. Most won't. They'll ask for clarification, the project will stall, and the team that built the pilot will blame finance for being obstructive.

The translation is the business case. Not the technology. Not the use cases. Not the pilot results. Scaling an AI initiative is not a technology decision. It is an investment decision.

200 hours per month at what cost per hour? If those hours are redeployed to higher-value work, what does that work generate? If the process being automated currently produces errors that cost €X to fix, what's the error reduction worth? If the initiative reduces customer churn by 2%, what does that mean in revenue terms over 18 months?

These aren't difficult calculations. They require the person building the business case to think like the person approving it.

What a CFO-grade AI business case actually contains

You 'only' need four things, in this order.

1. The strategic anchor. What business problem are you solving and why does it matter to the company's targets? Don't describe "we want to be more AI-driven." You need something specific: customer retention is declining, operational costs per order are too high, the current process can't scale to projected volumes. The AI initiative exists in response to something real.
   
   Get inspired, what about the strategic objective of:
   - Reduce operational costs
   - Increase productivity
   - Improve customer service
   - Reduce risk/enhance compliance
   - Decision making insights
   - Sales growth
   
   Tip: try to tie the business problem to multiple strategic anchors to make the case even more compelling and to get a broader buy-in.
   
   Looking for a template? You find one here.
   
   

2. The financial translation. Expected value in euros, not effort in hours. Include:

   1. Cost to deliver (build, run, maintain)

   2. Expected return (revenue impact, cost reduction, cost avoidance)

   3. Timeframe to first return and break-even point.
      If you can't estimate these with reasonable confidence, the business case isn't ready yet.
      
      

      Tip: also add your assumptions, this makes your business case more strong. Don't make executives guess and dare to share what the ROI looks like if it's 30% worse than expected.
      
      

3. The risk of doing nothing. This one is usually missing entirely. If the initiative is addressing a real business problem, what does it cost to leave that problem in place for another year? Make sure to quantify it. This is often more persuasive than the positive business case.
   
   Tip: Flip the order and open with what the status quo is costing the business right now (think about lost revenue, error costs, manual headcount, whatever is measurable). Now you're offering a fix and not only an investment.
   
   

4. The path from pilot to programme. A business case for scaling AI is not the same as a business case for running a pilot. You need to show that the pilot results are representative, that the operating model can absorb the change at scale and that there's a governance structure for what comes after go-live. This is where most proposals stop too early.
   
   

What this looks like in practice

A household appliances manufacturer I worked with was running e-commerce through a patchwork of B2B systems and manual processes. The business case for the transformation wasn't built around technology. It was built around three numbers:
- the revenue being lost to operational errors
- the cost of managing customer complaints caused by poor digital processes
- the revenue growth achievable with a properly structured channel.

That framing got the initiative approved and funded in a single review cycle. The result: 20% revenue uplift and a customer journey consolidated from over ten teams to one central department.

The technology made it possible. The business case is what made it happen. Want to see the case: see it here.

One thing to do differently

Before you write your next AI business case, ask yourself:
What is left after I remove everything about the technology and the pilot results?
Is what remains a financial argument that stands on its own?

If the answer is no, you're not done yet.

Photo by Mika Baumeister on Unsplash

Nick van Zuijlen is a digital transformation consultant based in the Netherlands. He helps e-commerce companies build transformation programmes that reach go-live including AI initiatives. More at vzconsulting.org