BCMC 2026 · Columbus, OH

Launch Your First
AI Agent.

A roadmap for getting agents running inside your own systems. Fill in the worksheets, then we go through them together.

Education Session · BCMC 2026, Columbus OH Build AI Agents to Improve Margins Across the Truss Sales Cycle

An agent is one decision, running on its own.

Not a chatbot. Not another dashboard.

Pick one decision your plant makes from memory today. Write down how you make it. Connect it to the data it needs and the guardrails it has to respect. Then it runs without you.

The worksheets take you through that for your own plant. The call turns it into a build order.

Three steps to agents running in your system

1

Fill in the worksheets

Nine pages. Pick the decision, map it, write the prompt, set the guardrails. Do it with a pen, on your own plant.

2

Book a call

Thirty minutes. We work through what you wrote, page by page, and find where your first agent actually belongs.

3

Leave with a roadmap

What to build first, what it plugs into, what order the rest comes in, and how long each piece really takes.

Step one

The worksheets

Nine pages you write in, the same forms we worked through in the room.

  • The seven tests for picking your first decision
  • The decision blueprint, blank and ready to fill
  • The six-part system prompt template, with a worked example
  • The agent loop, all six stages, for your own decision
  • The guardrail sort, so it only does what you allow
  • The First Move card to bring to the call

Send it to me

The download opens the moment you submit. I read every reply myself. No drip campaign.

What you fill in

Nine pages, one decision

Bring a pen and you finish with a decision worth building, a written system prompt, and a way to tell whether it worked.

Pages 1 to 2

Pick the decision

List the calls your plant makes from memory, then run your best candidate through the seven tests: frequent, valuable, repetitive, context-rich, measurable, low risk, and fails loud.

Pages 3 to 4

Map it

The decision blueprint. Stage, decision, who owns it, what it needs to know, and how you would know it worked. Plus the three things only you can give it.

Pages 5 to 6

Write the system prompt

Six parts, one box each: role, context, inputs, output, decision logic, rules. A worked example to model, then blanks for yours.

Page 7

Run the loop

Observe, understand, decide, act, measure, learn. Fill in all six for your own decision.

Page 8

Set the guardrails

Sort every action into never, needs approval, or act on its own. Start further left than feels necessary.

Page 9

Your first move

One failure point, one tool, one action, one date. This is the page we open first on the call.

Where the agent plugs in.

Five layers, built in this order. An agent runs on top of the first three, so your roadmap starts wherever yours stops.

Database

One place every job lives and every department connects to.

API

The permit. It is what lets your tools talk. Ask every vendor before you sign.

Integrators

The plumbing. One tool tells the next, with nobody re-entering anything.

AI

The intelligence. It comes fourth because smart AI on bad data gives confident wrong answers.

Interface

The dashboards and the chat window. Last, because it can only show what the four layers underneath already know.

Most plants already have layer one. Bring yours to the call and we start where you are.

Step two

Bring the worksheets to a call

Thirty minutes, no pitch. We open page nine first, work backwards through what you wrote, and put the build in order.

  • Whether your decision is an afternoon automation or a real agent
  • Which layer your plant is standing on today
  • What to connect first, and what can wait until renewal
  • A written roadmap from where you are to agents running on their own
Craig Mooney

Craig Mooney

AutoBuild Solutions · Bend, Oregon

I sell trusses for a living and I build these systems inside plants. Every demo at BCMC ran on real truss data.

Asked on stage

Four questions everyone has

"Our plate software has no API."

Build the database first, connect what you can, revisit the plate question at renewal. No API limits what connects automatically. It does not stop you starting.

"How much does this cost?"

Low hundreds a month, and most of what I showed runs on free plans. You spend setup time, not budget.

"Where do I start?"

Layer one. Pick a database, set up a project record, connect one workflow. Everything else sits on that.

"How long does it take?"

An automation, an afternoon. An agent, a few weeks, because clean data comes first. Your first move is probably the afternoon.