Ask someone whether an AI project should bolt onto a process or redesign it, and most people answer based on how big the task feels. That’s the wrong test. Bolt on when you actually needed a redesign, and the gains cap out fast. Redesign something that only needed a bolt-on, and you’ve spent months on change management for a task that didn’t require it.
The right test is easier to apply than it sounds. Two real examples make it clear: drafting a client memo, and untangling an eight-step AP/AR process.
Bolting on means the process stays the same. Same tasks, same order, same handoffs. AI just gets handed to the person doing the work. Redesigning means the process itself changes. Some tasks disappear, some get added, and the checks get built in on purpose instead of relying on whatever review already happened to be in place.
Example one: drafting a management letter (bolt it on)
Drafting a first-pass management letter or client memo is a textbook bolt-on. It’s bounded. You already know what a good one looks like, because you’ve written plenty of them before. AI drafts it in a fraction of the time it would take you starting from a blank page. You edit it, you own the result, and the task gets done faster.
Nothing about the process changes. You’re still responsible for the final product, the same way you always were. What changed is that the blank-page problem went away. That’s a real win. It’s just not a redesign. You raise the floor on how fast a first draft happens, but you don’t change who’s involved or where the checks sit. The task doesn’t ask for that.
Example two: an eight-step AP/AR process (redesign it)
Now take an eight-step accounts payable and receivable process that moves between clerks and software, in and out of different systems, in and out of spreadsheets. Bolting AI onto step three of that chain doesn’t fix anything. It just makes one link faster while the other seven keep grinding along the same way they always have.
The move here is to collapse the chain, not speed up each handoff. If you redesign the process instead of patching it, AI can run the steps end to end and hand off to a person only at the exceptions: the invoices that don’t match, the numbers that don’t reconcile. That’s a different task list, with different owners and different checks, built around what AI can actually do instead of built around a person doing every step by hand.
That last part matters more than it sounds like it should. Bolt AI onto a process like this instead of redesigning it, and the old checks usually stay exactly where they were. That sounds safe, but a person who no longer has to do the manual work also has less friction stopping them from just passing something along. A redesign is where you get to decide, on purpose, where a human actually needs to look.
The two-question test
Put those two examples side by side and the test is obvious. Is the task bounded enough that you’d catch a wrong answer just by reading it? Bolt it on. Does the work pass through multiple people or systems before it’s actually finished? That’s your redesign candidate. Bolting AI onto one link in a long chain won’t move the process. It’ll just move the bottleneck.
Key takeaways
- A bolt-on works when the task is bounded and you’d know immediately if the output were wrong. Drafting, first passes, anything you already know how to check.
- A redesign is for work that spans multiple people or systems. Speeding up one handoff in an eight-step chain doesn’t fix the chain.
- The process only actually changes in a redesign. Bolting AI on makes the same process faster, not different. That includes the checks, which is why redesigns are where you rebuild controls on purpose.
- Decide which one you’re doing before you start. A bolt-on with redesign-level ambitions, and a redesign executed like a bolt-on, both underperform.
Want the CPE credit? Take the full lesson on EverydayCPE, which walks through the research behind both examples, and earn 0.2 CPE credits: https://everydaycpe.com/courses/technical-management-services/when-to-bolt-ai-on-vs-redesign/
Leave a Reply