Daily field briefing
POLR AIPOLR AI

AI + AEC Today

Daily AI intelligence for the built world

Weekly digest

Weekly AI in AEC stories

← Archive

Edition · Tuesday, July 21, 2026

AI & Construction

Weekly Pulse · Industry actionability

The field is piloting, not just experimenting.

56%
ExperimentalPilotingField-ready
  • Big money is flowing into AI that runs real job machines
  • AI bid and estimating helpers cut weeks of work down to days
  • Free, open AI models are catching up to the top paid ones
  • The AI model race stays tight, so keep your options open

Practical Use of the Day

Let AI sort your next bid package in minutes

Estimators lose hours just sorting plans and specs before the real work starts. An AI chat tool can do that first pass for you. Give it the drawing list and specs, and ask it to pull out the scope, the key dates, and what needs a price. Check its work, then send your quote requests. This gets you to the real thinking a lot faster.

Try this prompt

You are helping a construction estimator start a new bid. I will paste a list of drawings and spec sections. Do three things. First, list the scope items I need to price, grouped by trade. Second, pull out any key dates, like the bid due date, the RFI cutoff, and the job start. Third, flag anything unclear that I should send an RFI on. Keep it short and plain, and put it in a simple table I can copy. Do not guess at prices. Here is the bid info:

Tool — Claude (or any capable LLM)

Stories

5 curated
ConTechHigh impact

Startup raises $115M to run heavy equipment from a screen

TerraFirma is a young company in Austin, Texas. It adds new gear to excavators, dozers, and loaders so a worker can run them from a screen, not the cab. That lets one skilled operator run more than one machine at a time. The company just raised $115 million to grow.

Why it matters to GCs

Skilled operators are hard to find, and this could stretch the ones you have. The firm says one operator can do up to three times the work. If it holds up, you could move more dirt with a smaller crew and keep people out of harm's way.

The Robot Report·Jul 14·Read source →

LinkedIn
ConTechModerate

AI bid helper raises $4M to speed up glazing quotes

Guthrie AI trains a virtual bid assistant for glass and glazing crews. The helper pulls project files, finds the scope, and sends out price requests on time. It works inside tools teams already use, like Bluebeam, Outlook, and Excel. The company raised $4 million to grow.

Why it matters to GCs

Chasing bids eats hours that estimators do not have. Guthrie says its users bid 70% more work without adding staff, and cut bid time from weeks to days. More bids out the door can mean more jobs won.

VentureBeat·Jul 13·Read source →

LinkedIn
ConTechModerate

Buildots ties AI progress tracking to crew and safety data

Buildots uses cameras and AI to track how a job is going. Now it added a new part called Buildots Field. It watches the workforce, safety, and where crews and gear are on the site. This came from a company Buildots bought last year.

Why it matters to GCs

Seeing progress and people in one view helps you spot slowdowns fast. It can flag safety risks and show where a crew is stuck. That means fewer surprises and less rework on your schedule.

Buildots·Jul 15·Read source →

LinkedIn
AI FrontierModerate

China's Moonshot puts out the largest open AI model yet

A Chinese company called Moonshot AI put out a new model named Kimi K3. It is the biggest open model ever shared, with 2.8 trillion parts. Moonshot says it does about as well as the top paid models from US firms. It plans to share the full model files by July 27 so anyone can run it.

Why it matters to GCs

Open models are free to run on your own setup, which can cut cost and keep data in-house. Strong free options give you more choice and more say on price. It is worth knowing these tools exist as you pick AI for your firm.

VentureBeat·Jul 17·Read source →

LinkedIn
AI FrontierLow

Google delays its top AI model over weak coding

Google pushed back the launch of its next big AI model, Gemini 3.5 Pro. The model did not do well enough at writing computer code in tests. Google is trying to fix it before letting people use it. Rivals like Anthropic and OpenAI keep shipping new models.

Why it matters to GCs

The AI race is close, and no one maker stays ahead for long. If you lean on just one tool, a delay or a miss can slow your plans. Keep your options open and do not bet your whole workflow on a single AI.

9to5Google·Jul 16·Read source →

LinkedIn