Before you pile on more AI, fix the boring stuff

Before you pile on more AI, get the boring stuff right. I'm talking about insurance certs, licenses, who got paid, the kind of paperwork that still eats jobsites alive.

I keep watching companies buy another round of AI seats while the same admin mess sits in the middle of the week. Someone is chasing a certificate that should have been on file, someone else is retyping the same numbers into three systems, and the superintendent is waiting on an answer that lives in a folder nobody named correctly. Then leadership asks why AI isn't "paying off," and I think they are looking at the wrong layer. The messy work under the tools is usually what broke down, not the model itself.

What I mean by boring

Boring is the stuff nobody wants to put on a slide: who is allowed on site, who got paid, which version of the submittal is current, which change order is still stuck in email, and which spreadsheet is the real one when two people swear theirs is.

In construction that boring layer is where margin leaks, and it is not always out in the field. A lot of the time it is in the office, in the handoffs, in the "we'll catch it later" pile. If you serve architecture, engineering, or construction firms across the South and the Midwest, you already know this movie. We wrote about how we think about that industry on our construction page, and the pattern is the same whether the trailer is in Knoxville or Columbus.

AI does not clean that up on its own, it just copies the mess at speed.

A scene I see over and over

I can picture the whiteboard with boxes and arrows all over it while a team maps how a contract, a cert package, or a pay app actually moves. Sales thinks it ends at signature, ops sees it late, finance sees it later, and the field finds out when something doesn't match. These are smart people, and the order of the work is what's off.

If you automate that order, you get faster rework, a nicer looking wrong answer, and junior work that looks senior on the page until someone has to stand behind it with a customer, an insurer, or an owner. You can hand off the typing, but you can't hand off the responsibility.

So before we talk tools, I ask a simpler question: which boring process is eating the most hours, and who owns it end to end?

Three checks before you buy more AI

I use a short filter with owners. No fancy scorecard, just three checks.

1. Can a new person follow it without a hallway conversation?

If the process only lives in one person's head, you don't have a process, you have a person everyone has to go find. Write down the steps as they really happen, not as the policy says they happen. Ugly truth beats a clean story.

2. Did you cut a dumb step before you automated it?

A lot of the "AI projects" I see are really process projects in disguise. Move a review earlier, kill a second form, stop copying the same data into another system. Fix the order, then decide what a tool should do.

3. Who answers when the output is wrong?

If nobody owns the answer, the tool owns your name with the customer. Name a person, give them the right to say no, and put the rules where the work happens, not in a chat that disappears.

When those three are true, AI has somewhere useful to land. That is the heart of AI enablement for us: pick real workflows, set simple rules, get the data ready, and train people on their own work. And when the boring handoffs are what slow you down, workflow automation beats another generic chatbot every time.

Better tools should mean more work gets done

I don't buy the story that better tools have to mean fewer people doing meaningful work. Better tools should mean more projects get built, more proposals leave on time, more pay apps go out clean, and more people spend time on judgment, relationships, and the hard problems that still need a person in the room.

AI is a lever, not a plan. The plan is still the same one good operators have used forever: slow down long enough to see the line, pick the line, then let the tool do what it's good at.

Where to start this week

Don't start by shopping platforms. Start with one boring process your team already complains about, map it, cut one dumb step, and name an owner. Then ask where AI or automation would take away typing, chasing, or entering the same data twice without taking away who owns the work.

If you can't name the boring process, you aren't ready for more AI, you're ready for a whiteboard 🤷‍♂️ If I were sitting with you I'd ask what process your team complains about most, and that's where I'd start.