A business case that prices doing nothing at zero isn't finished. Count the hours, the errors, the cash tied up and the unofficial tools people already use. Then test your assumptions against delay and thinner benefits, and release money in stages as decisions get made, not on a date in a plan.
Every business case has a list of options. The one that’s usually missing, or costed in a single lazy line, is doing nothing.
It’s an easy one to wave through. Nobody has to sign a contract. Nobody has to change how they work. It costs nothing to approve, and it’s often the most expensive option you’ll look at.
Why doing nothing looks cheap
The cost of the current process is spread across budgets and hidden in people’s days. A team spends hours pulling information from different systems before they can make a decision. Some of that information is wrong, so some decisions are wrong. Those lead to rework, complaints, cash that doesn’t come in, and the occasional expensive mistake that ends up in front of a senior leader.
None of that appears on an invoice. So when you put a new tool with a clear price next to “carry on as we are”, the tool looks expensive and the status quo looks free.
The classic structure for a business case, set out by Roberts back in 1997, includes doing nothing as a real option for exactly this reason. It forces you to put a number on the current state.
How to cost it properly
Start with time. How many hours a week go into the work you want to change, and what are they worth? Then errors. How often does the current process get it wrong, and what does each mistake cost to put right? Then the harder things to count: the complaints, the cash tied up for longer than it should be, the good people who leave because the job is mostly searching and copying.
And then the risk that grows while you wait. In most organisations I’ve worked with, people are already solving the problem themselves, with spreadsheets, personal accounts and free tools. That’s a bigger governance risk than anything in the proposal, and it gets worse the longer the official answer is “not yet”.
Treat the numbers as a story, and test it
Net present value is a useful measure, but it’s a story built on assumptions. A good case shows what happens when those assumptions are wrong.
Test four things. What if the benefits arrive a year late? What if they’re 20% lower than you hoped? What if the upfront cost is 25% higher? What if money costs more than you’ve assumed? For AI work, delay tends to do the most damage. The technology is rarely what holds things up. Access to data, approvals and adoption are, and every month they slip is a month of benefit you don’t get back.
When there are two credible figures for something, use the less flattering one and stick with it throughout. Boards spot a case that picks whichever number helps most. They’d much rather see the risks set out plainly than discover them later.
Fund it in stages
The other way to make a case safer is to stop asking for everything at once. Release money when a decision has been made, not on a date in a plan. A first stage might need nothing more than a yes. The next might depend on getting access to the systems you need. The build only gets funded once you know the data is usable.
That way the money follows the permission, rather than running ahead of it. If access never arrives, you’ve spent very little finding that out.
The question to ask
Before your next AI business case goes to the board, look at the do-nothing column. If it says “no cost”, it isn’t finished.
Business case going to the board soon?
Send it over before it goes up. A second pair of eyes on the do-nothing column is usually the cheapest hour you’ll spend.