I had a conversation with an ops lead at a 50-person agency a few weeks back. She wanted to use AI to speed up their client onboarding. So I asked her to walk me through the current process. She got about four steps in before she said "well, it depends on who's doing it." That's the problem right there.
Gallup's recent data showed that only 10% of businesses feel AI has properly changed their workflows. The other 90% are using AI to tinker around the edges. Drafting emails a bit faster. Summarising documents. Maybe generating a first pass of some copy. All useful in isolation, but none of it changes how the business actually runs.
The documentation problem
You can't fix a process that only lives in one person's head. And that's where most firms are right now, if that makes any sense as a generalisation.
The account manager knows how client onboarding works because they've done it something like forty times. The strategist knows the review process because they built it. The ops lead knows which spreadsheet tracks what because they set the thing up three years ago.
But none of that is written down. So when someone new starts, they shadow someone for a week and pick it up as they go. When something goes wrong, the fix lives in a Slack thread. And when someone asks "could AI help with this?", nobody can point to a document that shows what "this" actually involves step by step.
That's the real blocker. You can't rebuild a workflow with AI if you can't clearly describe the workflow in the first place.
What an SOP actually does for you
A standard operating procedure sounds boring, I know. But writing one forces you to do something surprisingly useful: think clearly about what actually happens at each stage of a process.
When I work with clients, the first thing we do in any workflow rebuild is document the current state. Every step, every handoff, every decision point. Who does what, when, and what triggers the next step.
That exercise alone surfaces problems. You see where the bottlenecks sit. You see which steps get skipped when things get busy. You see where the same information gets entered into two different systems. And you see exactly where AI could take over a chunk of the work.
Without that documentation, you're guessing. And guessing with AI tools is just tinkering with extra steps.
How to actually do this quickly
I demonstrated this on today's show. I described a fictional agency's client onboarding process out loud, in plain English. How the handover works from sales to accounts. Where the brief document lives. What happens when the client is slow sending assets. Where the strategist skips competitor research because they're slammed on other accounts.
Then I fed that into an SOP builder and it turned the whole thing into a structured standard operating procedure with phases, steps, ownership, and quality gates. It flagged the three steps where AI could save the most time and estimated how many hours per week the team could get back.
Something like five minutes, start to finish.
You don't need a fancy system for this. You need ten minutes, a recording tool or a text box, and the willingness to describe how a process actually works rather than how you think it works. Those are often two quite different things.
The connection between SOPs and AI adoption
This loops back to the Gallup data. 10% of businesses feel AI has changed how they work. 46% of employees aren't even using the AI tools available to them.
The missing piece between having AI tools and getting real workflow change is process documentation. Once you know what the steps are, you can spot which ones AI should handle. Once you've written it down, anyone on the team can follow it. And once it's structured, you can measure whether the new way is actually faster.
The firms I work with that get the fastest results all start the same way: documenting one process properly, rebuilding it with AI baked in from the start, and then having the team run it on real work within the first few weeks. That's the AI-plus approach, and the documentation is what makes it possible.
Start with one process
Go write down one process today. Pick the one your team does most often, or the one where things keep going wrong. Describe it step by step. Don't worry about formatting. Just get the knowledge out of someone's head and onto a page.
That's the starting point for real workflow change, and everything else builds from there.
If you want help documenting and rebuilding your team's workflows with AI, that's the core of what I do as a fractional AI lead. Book a discovery call and we'll look at the highest-value process to start with.
