How to Build the AI Business Case Your Board Will Actually Approve

Most AI proposals die in the boardroom, and it’s not because of the technology

A McKinsey survey from late 2024 found that while 72% of companies had experimented with AI, only 28% had scaled it beyond pilots. The gap isn’t technical capability. It’s internal buy-in. Specifically, budget approval from the people who control spending.

We’ve sat through dozens of these conversations. A team has found a genuine AI opportunity, one that could save real money or unlock real capacity. They put together a proposal. It includes the technology, the vendor, maybe a demo. And the board says no. Or worse, they say “interesting, let’s revisit next quarter,” which is a polite no.

The problem is almost always the same: the proposal speaks the language of technology, and the board speaks the language of business. They don’t want to hear about large language models or neural networks. They want to know three things: How much does it cost? What does it save us? What happens if it goes wrong?

The companies that get AI budgets approved consistently aren’t better at AI. They’re better at framing. They understand that a board presentation is a financial argument, not a technology pitch. And they structure their case accordingly.

This matters more than most people realise. A rejected proposal doesn’t just delay a project. It poisons the well. Once a board has said no to an AI initiative, getting them to revisit it takes twice the effort. The first impression sticks. Which means the way you present the first case sets the trajectory for every AI investment that follows.

The four-part structure that works

After working through AI business cases with companies across healthcare, logistics, professional services, and manufacturing, we’ve landed on a structure that consistently gets through board review. It’s not complicated. It’s just disciplined about what boards actually need to decide.

Part 1: The cost of the status quo. This is where most proposals fail before they start. They open with the AI solution. Boards don’t care about solutions until they feel the problem. Start with what the current process costs: in euros, in hours, in error rates. Be specific. “Our accounts payable team processes 2,400 invoices per month. Each one takes an average of 14 minutes of manual handling. That’s 560 hours per month (roughly 3.5 FTE equivalents) at a fully loaded cost of €18,200/month.” That’s a problem a board can feel.

Part 2: The opportunity, in numbers. Now introduce what changes. Not the technology - the outcome. “An automated processing system handles 80% of standard invoices without human intervention. That reduces manual handling to 112 hours/month. Net saving: €14,560/month, or €174,720 annually.” Notice there’s no mention of GPT, machine learning, or APIs. Just numbers. Boards approve numbers.

Part 3: The investment and payback. Every board member’s next question is “what does it cost?” Be ready with a clear answer. Total implementation cost, monthly operating cost, and the payback period in months. According to our audit data, the strongest AI business cases have payback periods under six months. Anything over twelve months and you’re fighting for attention against a dozen other capital requests.

Part 4: The risk mitigation. This is where boards lean in. They’re not just evaluating upside - they’re protecting downside. Address it directly. What’s the implementation timeline? What happens if it doesn’t perform? Is there a kill switch? Can you pilot it on one team before rolling out company-wide? A phased approach with clear exit points turns an all-or-nothing bet into a controlled experiment. Boards fund experiments.

Putting the framework into practice

The framework above sounds straightforward. Applying it well requires a few things that most teams skip.

Do the audit work before the board meeting. You cannot build a credible business case from assumptions. You need real numbers: actual time spent, actual error rates, actual costs. This is why we always recommend a structured operational audit before any AI investment proposal. An audit gives you the data that makes the business case bulletproof. Without it, you’re guessing. And boards can smell guesswork.

Benchmark against alternatives. Boards don’t evaluate AI proposals in isolation. They compare them against every other way they could spend that money. A new hire. A process redesign. A different tool entirely. Your business case needs to acknowledge these alternatives and explain why the AI approach wins. Often it comes down to scalability: a new hire handles one workload; an automated system handles ten without additional cost.

Name the internal owner. Boards are deeply sceptical of projects without clear ownership. If your proposal doesn’t name the person who will be accountable for the result (not the technology, the result) it reads as a pet project. The best owners are operational leaders who feel the problem daily. Not the CTO. Not the innovation team. The person whose team wastes 40 hours a week on the process you’re fixing.

• Map the current process and calculate its real cost before proposing any solution
• Present outcomes in financial terms: euros saved, hours recovered, errors reduced
• Include a clear payback period (aim for under 6 months for the strongest case)
• Propose a pilot phase with defined success metrics and exit criteria
• Name an operational owner who lives with the problem and will own the result

One more thing: keep the presentation short. We’ve seen 40-slide AI proposals that should have been 6 slides. Boards don’t have patience for technical architecture diagrams. They want the problem, the solution, the cost, the payback, the risk, and the next step. Everything else goes in an appendix they’ll never read.

The counter-piece, the one that quantifies what the board is approving by refusing, lives in the real cost of waiting on AI; pair the two when the room is sceptical that doing nothing has a price tag. The audit-first engagement shape that produces the operational data Part 1 and Part 2 of the framework demand is documented on the process page. A structured audit gives the board the operational evidence the deck needs, in the form they need it.

When the four-part structure isn’t the right pitch

The four-part structure assumes a traditional board approving a discrete operational investment with a measurable payback. A handful of decision contexts don’t fit that shape, and using this framework in them weakens the case rather than strengthens it.

• The board is approving a strategic capability, not a project. Some asks are about building AI competence as a foundation, with no single-project payback inside twelve months. Forcing a six-month ROI on a capability investment misrepresents the bet. The right frame is portfolio thinking and option value, not status-quo cost minus running cost.
• The buyer is a founder-led private company, not a board. In an owner-operated SME the decision-maker isn’t weighing your AI proposal against ten other capital requests. They’re weighing it against their gut feel about the business. The four-part deck reads as over-engineered. A one-page memo with the same numbers usually wins faster.
• The risk profile is regulatory, not operational. Healthcare diagnostics, automated hiring, credit decisioning, biometric processing. In these domains, “risk mitigation” isn’t a pilot scope and a kill switch. It’s an EU AI Act risk-classification analysis, a documented human-oversight model, and bias testing. The four-part structure underweights regulatory work that should sit at the centre of the case.
• The opportunity is defensive, not financial. Some AI investments exist because the competition has them and not having them costs you deals, not because they generate measurable savings on a known process. The right argument here is competitive positioning, win-rate data, and customer interview evidence. Forcing a payback calculation when the real argument is “we’re losing deals to firms that do this” produces a weak number for a strong case.