AI, without the snake oil

You've heard the pitch at every conference and probably on the golf course too. Most of it comes from people who have never closed out a project. Here's what I've seen work on real jobs, and what hasn't.

A man picks up a suit from Levine the tailor. The jacket is too big in the back, so Levine tells him to hunch over. The right sleeve is three inches too long, so lean and hold the arm out. The left leg is short, so pull it up at the hip. He walks out twisted to make it fit. A stranger sees him and says, "What a great tailor. He fit a suit to a man built like that."

That's how construction software has always worked. It's cut for everyone, so your team twists itself until the tool looks like it fits.

AI finally makes it practical to cut the software to your process.

Adapted from Gerald Weinberg, The Psychology of Computer Programming (1971).

Where it helps

  • Turning your foreman's voice notes and photos into the daily report
  • Going through the inbox every morning and flagging RFI answers, change requests and deadlines
  • Reading the bid documents and listing everything you have to price
  • Catching drafting errors, like one window tag used for two different sizes
  • Turning a 70-page contract into the notices and deadlines you have to hit
  • Lining up five sub quotes side by side so the scope gaps show

Where it doesn't

  • Making the judgment call your estimator or super would make
  • Measuring quantities off a PDF drawing. It still misses, and it won't tell you where.
  • Forecasting cost to complete. It doesn't know about the scope gap or the slow crew.
  • Pricing without your own cost history. It makes up numbers that look right.
  • Fixing a broken process. It just runs the broken process faster.

And if someone quotes you an ROI number before they've seen how you work, walk away.

The ChatGPT trap

"My team already uses ChatGPT" usually means one PM has a browser tab open. It helps that one person. When they leave, it leaves with them.

The version worth paying for runs on your own project data, so the whole team can use it and it gets better with every job. ChatGPT has never seen your job costs.

How I roll it out

Your workflow stays put

I build around how your team works. The AI underneath can be swapped for a better one later, the same way Excel updates and your spreadsheets keep working. You're never stuck with one vendor.

One workflow first

I start with the workflow that hurts most and prove it on a live project. If it works, you pick the next one.

A straight no when it's a no

If a better template fixes it for less, that's what I'll tell you. Putting AI where it doesn't belong is the fastest way to lose your team's trust in it.

Got an AI question? Ask me. You'll get an answer from someone who has run the jobs.

Talk to Jared