When a workflow hurts, the instinct is to reach for the most impressive tool in the room. Right now that's AI. But "add AI" is an answer looking for a question — and it's often the wrong fix for the problem in front of you.
There are really only three ways to fix a broken workflow, and they cost wildly different amounts. Knowing which one you need is most of the decision.
The three fixes
Buy — someone already sells software that does this. Invoicing, scheduling, CRM, help-desk, inventory: these are solved problems with mature products. You're paying to not build.
Automate — the steps exist across tools you already own, and the work is moving data between them by hand. No new product needed, just a connection: when this happens there, do that here.
Build — the task needs judgment, or your process is genuinely unusual, and nothing off-the-shelf fits. This is where custom software and AI agents earn their cost — and where they're wasted if one of the cheaper fixes would have worked.
The most expensive mistake isn't picking the wrong tool. It's building something custom when a $40/month product would have done it.
How to tell which one you need
Ask three questions about the workflow, in order.
1. Does a product already do this well? If the task is common across businesses — bookkeeping, appointments, email marketing — the answer is almost always yes. Buying means someone else maintains it, fixes the bugs, and ships the improvements. Reach for build here and you've signed up to reinvent a wheel that's already round. This is the software-implementation path: the hard part isn't the tool, it's choosing, migrating, and getting the team to actually use it.
2. Is the pain just moving data between systems? If your people spend their day copying a number from one screen into another, you don't have a thinking problem — you have a plumbing problem. That's an automation, not a build, and it's usually the highest-ROI fix there is: cheap, fast, and it removes an entire category of human error. (More on that in The copy-paste tax.)
3. Does the step actually require judgment? Reading an unstructured email and deciding what it's asking. Triaging a case. Drafting a reply that fits context. This is where AI agents belong — tasks a rules-based automation can't handle because they need interpretation. If a workflow clears the first two questions and still hurts, you've found a real candidate for a build.
Why the order matters
Run the questions in order and each one is a filter. Most workflows get resolved at question one or two — bought or automated — long before you reach the expensive answer. The ones that reach question three are the few genuinely worth a custom build, which means your build budget goes toward problems that actually need it.
Skip the order and you get the two classic failures: buying a heavyweight platform to solve a plumbing problem, or building a bespoke system to do what a standard product already does.
Where the audit comes in
This is exactly the sorting an AI-readiness audit does — not "where can we bolt on AI," but "for each painful workflow, what's the cheapest fix that actually solves it." Sometimes that's an agent. Often it's a product you could buy this afternoon, or an automation that pays for itself in a month.
The goal was never to use AI. It was to fix the workflow. AI is just one of the three tools — and knowing when not to reach for it is what keeps a project honest.
Want that sorting done for your business? Book a free audit — we'll tell you, workflow by workflow, whether it's a buy, an automate, or a build.