How should construction project managers learn to use AI?

It depends on where your PMs spend their time. If most of their week goes to writing, such as RFIs, change order narratives, meeting minutes and owner updates, start with short, hands-on sessions on those exact documents using approved AI tools. If they mostly fight scattered information across email, plans and project software, pair training with a simple way to find past project documents first. Either way, train on real project work in small groups, not generic AI demos, and make clear that AI drafts while the PM still checks and signs off.

Who this is for

COOs, operations leaders and senior project managers at general contractors, subcontractors and construction management firms who want their PMs to use AI well, without adding risk.

The short version

PMs adopt AI when it saves them time on documents they already have to write.
Train in small groups on real, approved project material, one task at a time.
Set clear rules on what AI may not do, such as decide on scope, cost or safety.
Measure time and rework on a few named tasks, not general “AI usage”.

Why PM training is different

Project managers sit between the field, the office, the owner and the design team. Their work is heavy on communication and documentation, which suits AI well. But they are also short on time, wary of anything that could create liability, and used to tools that promise a lot and add clicks. Generic AI training that shows a chatbot writing a poem will lose them in ten minutes.

What works is showing a PM how to turn their own messy notes into a clean RFI, or how to check a subcontractor’s proposal against the scope sheet, in less time than they spend today.

PM tasks to start with

Pick two or three of these for the first sessions:

RFI drafting. Turn field notes, photos and plan references into a clear, well-structured RFI. See AI for construction estimating and RFIs.
Meeting minutes and action items. Turn a recording or rough notes from an OAC or subcontractor meeting into minutes with owners and dates.
Owner updates. Draft weekly or monthly updates from schedule notes, logs and open items.
Change order narratives. Draft the explanation of what changed and why, from the facts the PM provides.
Scope and proposal checks. Compare a subcontractor proposal with the scope of work and list gaps or exclusions for the PM to review.
Specification lookups. Ask questions of a long spec section and get pointed to the right paragraph to read.

Leave schedule logic, cost decisions and safety calls out of AI training at first. Those need firm processes before AI touches them.

A training plan that fits PM schedules

Kickoff (60 to 90 minutes). Show two tasks done live with real, approved project documents. Set the rules. Give everyone access to the approved tool.
Practice week. Each PM uses AI on the chosen tasks on a live project and saves examples of what worked and what did not.
Working session (60 minutes). Small group review of real examples. Fix prompts, agree on templates and add one new task.
Shared templates. Save the best prompts and examples where every PM can find them. See how to create prompting standards.
Monthly check-in. Short session to share wins, fix problems and add tasks.

Keep groups small and mixed by experience, so senior PMs bring judgment and newer PMs bring curiosity. For field staff, a different approach works better; see rolling out AI to construction field teams.

Rules to set before the first session

Approved tools only. Project documents, owner contracts and pricing go only into the firm’s approved AI accounts.
AI drafts, the PM decides. Anything that goes to an owner, architect or subcontractor is reviewed and sent by a person.
No AI decisions on scope, cost, schedule commitments or safety.
Check references. If AI cites a spec section or drawing, the PM opens it and confirms.
Respect contract terms. Some owner contracts have confidentiality or technology clauses. See can we use AI on documents covered by an NDA.

Write these into your AI acceptable use policy.

How to tell if it is working

Before training, ask PMs to estimate how long a few named tasks take, such as one RFI, one set of meeting minutes and one owner update. Ask again 30 and 60 days later. Also track:

How many PMs use the templates each week without being asked.
Whether RFIs and updates need fewer rounds of revision.
Whether PMs are suggesting new tasks to add.

If use drops off after the first month, the usual causes are tasks that did not fit, no follow-up session, or unclear rules. See why AI adoption stalls and how to measure whether AI training worked.

Common mistakes

Generic demos. PMs need to see their own documents, not a general tour of features.
Too many tools at once. One approved tool and a few tasks beat a long list.
Skipping senior PMs. If the most respected PMs do not use it, others will not either.
No templates. Without shared prompts and examples, each PM starts from scratch and many give up.

When you do not need us

If you already have a PM who uses AI daily, has time to teach others and works within clear rules, an internal program may be enough. Bring in help when adoption stalls or you want to connect AI to your project documents and software.

How Ruby helps

AI training: hands-on sessions for PMs on their real project documents, with follow-up.
AI strategy and consulting: choosing which PM workflows to change first and setting simple rules.
AI engineering: building agents that find past project documents or draft RFIs and updates from your systems.

If you want your project managers spending less time on paperwork, book a call with Ruby.