Fix the Process Before You Automate It

Every few weeks a client asks some version of the same question: should we be using AI for this?

It’s a fair question, and the tools are genuinely fun to pilot. But before answering, it helps to ask a different question first: what does the process look like today, step by step, without the tool?

Most of the time, the honest answer is “it depends who does it” or “we mostly wing it.” That answer tells you most of what you need to know about whether AI will help or hurt.

What the research actually shows

A widely cited study tracked corporate generative AI pilots through 2025 and found that 95% of them failed to produce a measurable return. That number gets quoted often, usually to argue AI doesn’t work. That isn’t quite what the research found. The pilots that succeeded shared one trait: they were built around a workflow the team already understood well enough to hand off cleanly. The pilots that failed were usually bolted onto a process nobody had actually written down.

Small businesses run into the identical trap at a smaller scale. If month-end close today is a mix of memory, three spreadsheets, and a sticky note on the monitor, dropping an AI tool into that mix doesn’t fix it. It automates the chaos faster, and now there’s a subscription fee for the privilege.

Three questions to answer before you shop for a tool

1.    What are the exact steps, in order, that happen today? Write it down, even if it feels tedious. If you can’t describe the process without saying “it depends,” that’s the first problem to solve, not the tool.

2.    Which steps actually need to happen, and which exist because “that’s how we’ve always done it”? Half the time the biggest time savings don’t come from a new tool. They come from cutting a step nobody remembers the reason for.

3.    What does “done well” look like for this process, specifically? If you can’t describe success in a sentence, an AI tool can’t aim for it either.

Where this shows up for our clients

We see this most often in reporting and reconciliation work. A business owner wants a faster monthly close, so the instinct is to reach for automation. But if the underlying chart of accounts is inconsistent, or transactions get coded three different ways depending on who’s working that week, no tool fixes that on its own. The fix is defining the process first: consistent categories, a repeatable close checklist, a clear owner for each step. Once that exists, the right tool, AI or otherwise, makes it faster. Before that, it just makes the mess move faster.

The takeaway

AI is a real productivity gain when it’s layered onto a process someone has already thought through. It’s an expensive way to add confusion when it’s not. Before your next tool purchase, spend an hour writing down what actually happens today. That hour will tell you more about where AI will help than any demo will.

Have a process you’re not sure is “automation ready”? That’s exactly the kind of conversation we have with clients before recommending any tool.