Automate first
High volume, same steps every time. Software can own the whole flow with a person checking exceptions. This is where a build pays back fastest, and it's what would sit at the top of a real audit.
What should you systematise first?
Most AI projects fail on the choice, not the technology. Pick the processes your business actually runs, rate them on volume and how much judgement they need, and get a ranked shortlist: what to hand over first, what needs a person kept in the loop, and what to leave well alone.
What each result means, and what we'd do about it
Assist, keep a person
Real volume, but a decision gets made each time. AI gathers the context and drafts the work, a person makes the call. Automating these end to end is how businesses end up with confident wrong answers at scale.
Worth doing, not first
Genuinely automatable, but the volume doesn't justify going first. These get bundled into a later phase, usually cheaply, because the plumbing is already there once the big flows are running.
Leave it alone
Low volume, or it needs a person thinking every time. Automating it would cost more than it returns. We'd tell you the same on a call, and an audit that comes back with several of these has still saved you the build.
Questions about the ranking
How does the ranking actually work?
Two things drive it. Volume, which is how often a process runs multiplied by how long it takes. And feasibility, which is how rules-based the work is. A process that runs 20 times a week and follows the same steps every time scores far higher than one that runs twice a week and needs someone thinking. We multiply the two, because high volume work that needs judgement every time is a bad automation candidate however expensive it is.
Why does something high-volume get told to leave it alone?
Because volume on its own isn't the case for automating. If a process needs a person making a real decision each time, software that tries to replace that decision produces confident wrong answers at scale, which is worse than the slow manual version. Those flows still benefit from AI, just in a different shape: gather the context, draft the work, hand it to a person. That's what the assist verdict means.
What's the difference between automate first and assist?
Automate first means software can own the whole flow end to end, with a person only checking exceptions. Assist means AI does the gathering and the first draft and a person still makes the call. Both are real builds. The difference matters because a business that automates a judgement-heavy flow end to end usually ends up unpicking it six months later.
My process isn't on the list. Does that matter?
Not much. The twelve here cover the shapes we see in most audits, and an unlisted process almost always behaves like one of them. Pick the closest match and adjust the numbers. If your business runs something genuinely unusual and high volume, that's worth a conversation rather than a scorecard.
How accurate are the default numbers?
They're starting points from processes we've mapped, not claims about your business. Change them. The ranking is much more sensitive to your real frequency and duration than to anything we've pre-filled, and the whole point is to get your numbers in front of you rather than ours.
What happens after I've got my shortlist?
Sanity-check it with whoever runs each process, because the volumes usually come back higher than the person managing them expects. Then decide whether the recoverable hours justify doing something. If they do, our AI audit maps these properly, costs them in hours and pounds, ranks them, and delivers a working proof of concept on the top item so you can see it before committing to a build.
Put a pound figure on it
You know the order now. The manual work cost calculator turns those hours into the annual figure, which is usually the number that gets a decision made. Our guide to which processes to automate first explains the ranking this scorecard runs and the traps that sink a first project, what repetitive admin really costs shows the arithmetic behind the pound figure, and our guide to AI for small business covers what to automate, what to leave alone, and what it costs to run.