Two and a half hours. No theory dumps. We show you how AI fits into a contractor's day, how the platforms differ, how to prompt them so they actually help, and where it pays off in masonry.
A coaching session, not a slide deck. Get on Wi-Fi and have ChatGPT, Claude, or Gemini open on your phone. Use the controls below as the room raises hands. Click through the platforms. Build a real prompt. Bring questions.
Headshot
Save your photo as ../Assets/Headshot.jpg in the AZ Masonry folder and refresh.
About your guide.
I'm Brennan Gerle. I run POLR AI here in Phoenix. What I do every week is sit down with AEC and commercial real estate leadership teams and figure out where AI actually pays off in their operations.
Not the hype version. The version that shows up in fewer admin hours, faster bid turnarounds, and cleaner handoffs to the field. I built this session specifically for the masonry and block side of construction because the work has a different rhythm than commercial GC or design work, and most AI training out there is not going to land for you. The goal today: leave with practical things you can use Monday morning.
Phoenix
POLR AI is a Phoenix-based operational AI firm working with AEC and manufacturing SMBs
Weekly
Working with AEC and CRE leadership teams every week, watching what gets adopted and what doesn't
2 of 2
Sessions in this AMC series. Today is the foundation. Next is safe research.
Section 01 / 09 · Room scan
Where is the room?
I sent a survey before today. The numbers below are pre-filled from your responses. If you didn't fill out the survey, raise your hand for the level you're at and we'll add it live. Click the plus and minus buttons to adjust as the room shows hands.
The room, as it stands
Total: 19 people
Pre-filled from the pre-session survey. Adjust live.
Never usedHaven't tried any AI tool yet.
Chatbot onlyUse one chatbot occasionally for emails or quick questions.
Daily useWorking it into day-to-day across two or more platforms.
Power userPaid plans, custom workflows, agents, deep integration.
Live · Apples to apples
This room vs. the rest of the country.
This room
Nationally
What this room looks like vs. the country: in this room, 65% of you are at chatbot-only or below. Nationally, only about 25% of contractors are at that level. The good news: that means we're going to spend the next two and a half hours catching you up to where most contractors already are.
Section 02 / 09 · How it works, where it's headed
Not a search engine. Not an expert. A really good guesser.
AI doesn't actually know anything. It learned by reading billions of pages from the internet, and now it predicts what words should come next based on what it's seen before. That's it. That's why how you ask matters so much, and that's why it sometimes makes things up while sounding totally confident. The good news: once you understand that, you can use it really well.
Click any card to flip and see more →
Adoption is moving fast.
Year
What was measured
Adoption
Trend
2023
GCs using or evaluating AI for precon and PM
34%
2024
Construction firms using AI or planning to invest
44%
2025
Construction firms using AI or planning to invest
61%
2025
GCs using or evaluating AI for precon and PM
67%
2026
Project professionals reporting AI used on projects
75%
Sources: AGC, Construction Dive, Construction Owners. Compiled May 2026.
Section 03 / 09 · AI is already in your stack
If you've used these tools, you've already used AI.
This isn't new and scary. It's running quietly in the tools you already pay for. The masonry-specific ones are first. Click any card to expand.
Before you subscribe to a paid tool
You can do a lot of this for free in ChatGPT or Claude.
Specialty tools like Beam AI and Togal are great. But before you spend $300 a month per estimator, try the free version of what they do inside ChatGPT or Claude. Most of you can get 70% of the value with $0 and ten minutes of practice.
Read a spec or scope sheet and pull out quantities
Togal / Beam does this from PDFs.
Free version: open ChatGPT, click the paperclip, upload the spec PDF, type "Read this spec section and list the masonry quantities I need to bid: block, mortar, lintels, flashing, weep screeds." It will read pages 1 to 40+ and pull what's there. Always double-check, but it gets you 80% of the way in 30 seconds.
Identify masonry items on a drawing detail
Beam, Civils.ai, and Revailo do this from drawings.
Free version: take a photo of the drawing detail with your phone, upload it to ChatGPT, type "List every masonry component shown in this detail and what each one is used for." For training your eye and onboarding new hires, this is gold.
Compare two material options or methods
Specialty estimators do this with cost databases.
Free version: "I'm comparing 8 inch CMU vs. precast for an exterior wall on a logistics building in Phoenix. Walk me through cost, schedule, and constructability tradeoffs." You won't get exact numbers, but you'll get the right questions to ask your supplier.
Draft bid clarification questions
Most paid tools don't do this at all.
Free version: use the prompt in Section 7 of this presentation. Paste your scope, get a ranked list of clarification questions. This is one of the highest-leverage uses of AI in estimating.
When to upgrade to a paid tool: when you're doing the same workflow 5+ times a week and the manual paste-and-double-check is the bottleneck. That's the moment Beam or Togal pays for itself. Not before.
Keep it secure
I always recommend the paid plan on whichever platform your team uses most.
$20 a month gets you better privacy defaults, longer document handling, faster models, and (on most platforms) terms that say your conversations don't go into model training. The free version is fine to learn on. The paid version is what you should be on once you're putting real client info in.
Why paid is more secure
Most paid tiers either default to "do not train on your data" or give you stronger controls than the free tier
Enterprise terms include SOC 2, BAA, and other compliance language you don't get on a free account
Longer retention controls and clearer "delete my data" workflows
Project workspaces with custom instructions live on paid plans, which makes it easier to keep your client work organized and contained
How to turn off training, by platform · click to expand
Section 04 / 09 · Platform tour
One wireframe. Five different brains.
All the major AI platforms work roughly the same way. Hover over each numbered area to see what that part of the interface does, and how the five platforms compare on it.
4
Model: GPT-5 ▾
7
⚙
5
Ask anything about your CMU bid, scope, schedule, or write something with me.⏎
6
Platform features
Cowork (Claude)
Files, sheets, slides built into chat.
Code mode
Built-in coding sandboxes.
Voice mode
Talk to it. It talks back.
Image / video
Generate or analyze imagery.
Hover or tap
Click any numbered area on the wireframe.
Each zone explains what that part of the chat interface does, and how the five major platforms compare.
Tip: the interfaces look almost identical across platforms. The differences are in what's available behind each section.
Platform comparison: features and security defaults
Capability
ChatGPT
Claude
Gemini
Perplexity
Copilot
Free tier
Yes
Yes
Yes
Yes
Yes
Paid plan / month
$20
$20
$20
$20
$30
Recommended model
GPT-5
Claude Sonnet 4.6
Gemini 2.5 Pro
Sonar / Auto
GPT-5 (in M365)
Projects / Spaces
Yes (Projects)
Yes (Projects)
NotebookLM (separate)
Yes (Spaces)
Limited
Custom assistants
Custom GPTs
Skills
Gems
Collections
Copilot Studio agents
Built-in coding sandbox
Yes
Yes (Code)
Yes
Limited
Yes (in IDEs)
Files / docs in chat
Yes
Yes (Cowork)
Yes
Yes
Native to M365
Trains on your data by default
Yes, turn off
No
Yes, review
Yes, toggle off
No (Enterprise)
Where to turn it off
Settings → Data Controls → Improve the model
Already off; opt-in only
Activity → Apps Activity
Settings → AI Data Usage
Already off in M365 commercial
Pick one platform to learn deeply. Borrow the others when you need them. The first thing to do on day one of any platform: turn off training on your data if it's on by default.
Section 05 / 09 · Prompt like you mean it
Same question, three different answers.
Pick a scenario. Toggle between a basic prompt, a slightly more detailed one, and a CRIT prompt. Watch how the answer transforms when you give the AI Context, a Role, the chance to Interview you back, and a clear Task.
Live buildThis is a great spot for a live build with the room. Pick the scenario closest to today's project, click through Basic → Better → CRIT, and let the audience call out what they'd add at each step.
You
AI
What changed:
Now build one of your own.
Type into each box for a real situation you're carrying. The fields turn green as they get specific enough. When the meter says ready, copy the prompt and paste it into ChatGPT, Claude, or whatever you use.
CContextEmpty
Who you are, what you're dealing with, what's at stake. Names, numbers, dates make it real.
Type your situation. The more specific, the better.
RRoleEmpty
Who you want the AI to be. Years, specialty, the lens to bring.
Tell the AI exactly who to be in your conference room.
IInterview
Locked
This is the secret ingredient. Locked in for you so you can focus on the rest.
This line tells the AI to ask up to three questions, one at a time, before answering. It's the single most important line in any CRIT prompt. We've locked it so you don't forget.
TTaskEmpty
What the AI should produce after the interview. Be specific about format and deliverables.
Tell the AI exactly what to make for you. List the deliverables.
0% · Start with Context
Your CRIT prompt
Fill in the fields above. Your assembled CRIT prompt will appear here, ready to paste into ChatGPT, Claude, or your platform of choice.
Live response
Apply it · Live exercise
Try CRIT on a real bid leveling example.
You just learned the framework. Now let's apply it to the #1 ask from your survey: bid leveling. We have a fake school addition project with a scope and three real bids in three different formats sitting in a folder ready to go. We'll spend five minutes running CRIT against it together.
Step 01 · 1 min
Set the context
Get a volunteer to call out: "I have three bids on a $1.8M scope. They're in mixed formats. Bid is due Friday." Type it into the Context box above.
Step 02 · 1 min
Set the role
"Veteran estimator with 25 years catching scope gaps in CMU bids." Type it into Role.
Step 03 · 1 min
Define the task
"Build a side-by-side comparison, flag silent exclusions, suggest plug numbers." Interview field is already locked, do nothing there.
Step 04 · 2 min
Test it live
Click Test it live. Watch the AI ask the room three questions back. Watch the lightbulb go on.
Mock files for the demo
Project folder includes 00_Scope_of_Work_RFP.xlsx, 01_Bid_DesertMasonry.pdf, 02_Bid_PhoenixBlockworks.docx, and 03_Bid_SonoranStoneAndBlock.xlsx. The third bid is the lowball with multiple silent exclusions. That's the punchline. Use these to demo the same exercise in Cowork after the break (Section 07 has the full Cowork prompt).
Section 06 / 09 · Three places to earn its keep
Three places it earns its keep this week.
Scheduling, bids, and marketing. Three workflows where AI turns a few hours of work into ten minutes, and where the output is good enough to use, not just look at. Click through the tabs.
None of these replace your judgment. They get you to a working draft fast. You still review, edit, and own the output.
Section 07 / 09 · Prompt library
A library of masonry CRIT prompts.
Pre-built prompts for the situations you'll face this week. Click a card to see the full prompt. Each prompt is editable, so you can paste in your own job details, names, or numbers right here, then copy or test it live. Click "Reset to original" any time you want to start over.
Want the whole library as a takeaway?
I have all 14 of these prompts in a printable PDF and editable markdown. Send the link to your team or hand it out at training. Email me at bg@polrai.com to grab a copy.
Section 08 / 09 · Coming next, search and research
What we'll cover in Session 2.
AI is great at speed. It is dangerous at codes, specs, and standards if you trust it without verifying. Session 2 is the safe-research playbook. Here's the preview.
01
Searching with AI
Where each platform actually searches the web vs. answers from memory, and how to tell the difference.
02
Deep research mode
When to ask AI to spend 5 to 30 minutes pulling sources before answering, and how to read what it finds.
03
Asking it to confirm sources
Never trust an AI answer on a code, spec, or standard without making it tell you exactly where it came from.
04
Verification habits
A short workflow for confirming what AI says against authoritative sources before relying on it.
A prompt to remember between now and Session 2
"Where did you find this? Provide the references."
Add this to any AI answer that touches a code, spec, or standard. If the AI can't show you the source, treat the answer as a hypothesis, not a fact.
Section 09 / 09 · Where you start Monday
Pick one thing. Try it tomorrow.
The biggest mistake people make is trying to roll out AI across the company on day one. The firms winning with AI are running small, low-risk pilots and growing what works.
Leadership
Run one weekly leadership prompt with CRIT for a real decision
Audit which tools your team already pays for that have AI built in
Set one rule: nothing confidential gets pasted without thinking
Estimator
Try AI takeoff on one closed bid to see how close it gets
Use AI to generate bid clarification questions before you submit
Draft scope assumptions and exclusions language with AI
PM
Turn one week's field notes into clean daily reports with AI
Draft your next RFI with AI, review it, then send
Use AI to summarize a long spec section into bullets for your foreman
Field
Voice-to-text into a chatbot to clean up a daily log on your phone
Ask AI to translate a spec into plain English before you push back
Draft tough conversations with subs or foremen before you have them
A quick word on safe use.
AI sounds confident even when it's wrong. Don't paste confidential project info, client data, or anything you wouldn't want in a public doc. Don't trust AI on codes, specs, or compliance without verifying the source. Human review stays your job. Session 2 of this series goes deep on safe research and verification, that's the right place to take code questions.
Common questions, answered
You're probably wondering one of these.
Click any question on the left, the answer appears on the right. These are the questions every contractor in every session asks.
Click a question to see the answer.
Open Q&A
This is the part of the session you came for. Bring your real situations, your weirdest workflow questions, the thing you've been wanting to try but weren't sure how. Let's go.
Two ways to keep going
Today is a start.
Real movement comes from working through your specific situation. One-on-one or with your team. Pick whichever fits where you are.
