A tree service contractor asked Bradley Benner to build him an AI operating system. Benner’s first move was not to open a tool. It was to make the contractor walk him through what actually happens when a lead comes in, step by step, and how long each step takes.

That order matters, and most automation projects get it backwards.

“A lot of the times they don’t have their processes defined very well … that’s the only way to build any sort of an AI operating system. You have to clearly define the processes and have them kind of mapped out, and then it can be built.”

Automating a sloppy process makes it worse, not faster

“If they don’t have clear processes in their own business, you can’t automate it, because then you’re just automating a sloppy process … you add another layer of complexity on top of an inefficient process. And that is having AI then also performing an inefficient process. So it compounds the problems.”

His word is compounds. The inefficiency does not stay the same size once you automate it — it runs more often, in more places, with less human attention on it, and then somebody spends their week maintaining the automation instead of doing the work. Benner’s blunt version is that it becomes more work rearranging things than it would have been to just do it manually.

This is the same failure we see across marketing channels when they are bolted together without a shared definition of what a lead is. We wrote about the structural version of it in holistic visibility.

The mapping step is the deliverable, whether or not you automate

Here is his actual workflow, which is unglamorous and repeatable:

• Get the owner on a call and have them narrate the process end to end. What happens when a lead comes in, who touches it, how long it takes.
• Take the recording and the transcript. No notes, no memory, no interpretation at this stage.
• Process the transcript into a documented process and a build plan.
• Take the plan back to the owner before anything is built.

Benner has been writing SOPs for years and says the discipline transfers directly: “it’s really no different with AI, you have to document the process in a very granular way.” The owner usually discovers their process has holes in it. That discovery is worth the engagement on its own, even if nothing gets automated afterwards.

One thing he changed his mind about

He used to hand the model a rigid spec. Now he does not.

“Instead of being so rigid and saying here’s the process, build an automation for this … you send the process to the model and say here’s what it is, this is the desired outcome, this is why we want it. Can you improve this process? Show me what you got.”

His reasoning is that the way a contractor currently does something is not evidence it is the best way — it is evidence it is what they got used to. Asking for improvements surfaces workarounds neither he nor the client would have thought of. He works in plan mode first, and says it often takes a lengthy back-and-forth before any build plan gets agreed.

The rule underneath all of it

Do not accept the first draft. That is the same instruction we give clients about AI-written copy, about channel recommendations, and about automation plans, and it is the cheapest quality control available: ask once more, and make the thing show its reasoning before you commit budget to it.

If you cannot describe your lead process in order, out loud, in under five minutes, automation is not the next thing you need. A whiteboard is. Book an intro meeting if you want help drawing it, or see how we work across SEO, PR and paid together.

Bradley Benner is the founder of Semantic Mastery and Semantic Links. He spoke on the Unscripted SEO podcast about relevance, entity associations and the future of link building.