People & leadership
Video · 9 min read
The short answer
Watch: getting construction teams to use AI
Watch the explanation, then explore the practical implications below.
What makes AI adoption work in a construction business?
AI adoption becomes useful when people can complete a real piece of work more reliably and with less unnecessary effort. For a construction business, that might mean recording site observations once, preparing a scope with the supporting evidence beside it, or bringing the right exceptions to a project manager for a decision.
Access to a tool is only the starting point. The system also needs usable information, clear responsibilities and a way to retain approved lessons. If each person starts a fresh chat and rebuilds the same context, the business can remain dependent on repeated explanations from its most experienced people.
Turn experienced judgement into approved company knowledge
An estimator’s correction can be valuable beyond the immediate draft. It may identify a recurring exclusion, a missing assumption or a particular interpretation of the company’s scope. Before that correction is reused, someone needs to establish where it applies and whether it is safe to share.
CSM would approach this as a managed process: capture the correction and its source; ask the responsible person to approve it; record its project, client or work-type boundaries; make it available to the relevant workflow; and test whether a later draft applies it correctly. Give the guidance an owner and a review point so outdated advice can be amended or withdrawn.
A comment in a chat should not silently become company policy. Some comments are tentative, some are wrong and others are valid only for one project. Staff need to understand the difference between a suggestion, an approved rule and a project-specific exception.
A practical example: an exclusion in an AI-generated scope
Suppose an experienced estimator notices that a draft scope includes work that should be excluded for a particular job type. Correcting that document solves the immediate problem. Retaining the approved lesson gives the next estimator a chance to avoid repeating it.
The useful record would explain the exclusion, the circumstances in which it applies, the evidence supporting it and who approved it. A general lesson may be suitable for a shared workflow after confidential project details have been removed. Client-specific information should remain within the appropriate project and access boundaries.
A later draft should surface the relevant guidance and supporting evidence for the estimator to review. Test both a case where the exclusion applies and one where it does not. This illustrative workflow needs to demonstrate correct reuse, rather than simply showing that a note has been saved.
Design around the different people doing the work
Site teams and surveyors: make capture practical
Observe what someone can realistically record while carrying out the job. Keep required fields focused on what the next person needs, and make it clear which project and location an observation belongs to. Account for poor connectivity and agree what happens when a record cannot be submitted. Check that people are not also being asked to enter the same observation into another system.
Estimators: put the evidence beside the draft
An estimator needs a practical way to challenge a result. Show the relevant drawings, specifications, observations and approved assumptions alongside the proposed scope. Make missing information and uncertainty visible. The process should make a correction easier to review and reuse without taking away the estimator’s responsibility for the judgement.
Project managers: make exceptions and decisions clear
A project manager may need a short queue of decisions rather than another long report. For each exception, make the issue, supporting information, proposed action and decision owner clear. Define which commitments need approval and how the approved decision reaches the people carrying out the work.
Train people on complete activities
Training should follow a real job from start to finish. Show a surveyor how to record an observation, an estimator how to check the resulting draft and a manager how to resolve an exception. Include what happens when information is missing, the answer is wrong or the system is unavailable.
Explain why the system is being introduced, the repetitive work you want to remove and who remains accountable. Be candid that roles may change. Give people a clear route to raise concerns, and show how their feedback influences the design. Training will have limited value if the workflow continues to create avoidable work.
Start with a small project team and improve the workflow
Choose a bounded process and involve the people who capture the information, check the output and act on the result. Include an experienced sceptic who can identify where the design falls short. Agree the existing process as a baseline and the outcomes that would justify continuing.
Review actual work during the pilot. Keep a visible record of reported problems, who owns each fix and what changed. If people return to a spreadsheet, investigate the reason: it may be quicker on site, easier to correct, more dependable or better suited to an exception the new workflow has missed.
Measure adoption through completed work
Useful measures include whether work is completed correctly, the time spent reviewing and correcting it, repeated errors, duplicate entry and the quality of the handover. Track whether approved corrections are used appropriately in subsequent drafts. Logins and message counts can show activity, but they cannot establish whether the process is helping.
CSM’s starting point would be the gap between the individual value people already get from AI and the dependable workflow the business needs. The aim is to keep useful knowledge, respect its boundaries and make it easier for site and office teams to do good work together.
Questions for your construction leadership team
Where are experienced people repeating the same explanation?
Identify one recurring correction in scoping, estimating or handover. Where is it recorded today, and how would a colleague find the approved version tomorrow?
Who decides what can become shared guidance?
Name the person who approves the lesson, defines its project or work-type boundaries and keeps it current. How will confidential client information stay within the right project?
What does each role need to complete the job?
Check the capture experience on site, the evidence available to the estimator and the decisions presented to the manager. Where are people still entering information twice?
Can the team challenge a result and recover from a problem?
Ask people to demonstrate what they would do with missing information, an incorrect draft or poor connectivity. Is the route to a responsible person clear?
What would tell us the pilot is helping?
Agree how you will compare completed work, review time and repeat errors with the current process. What feedback would make you change the design before expanding it?
Full video transcript
The full video transcript is included below for reference.
CSM Construction FAQ 05 · Adoption
[0:01] Your most experienced estimator corrects an AI-generated scope.
[0:05] The next day, a colleague opens a different chat and makes the same mistake. So the estimator explains it again.
[0:12] A month later, the team is still pasting in company background, finding the same documents and repeating the same corrections.
[0:19] People may be getting individual value. But the business has no dependable process for retaining what they have learned.
[0:26] That is a problem worth solving.
[0:28] Shared AI features and connected knowledge already exist. The work is deciding which information is approved, how corrections are reviewed, and how the next person benefits from them.
[0:39] At CSM, we want an experienced person’s judgement to become something the team can reuse, with the right boundaries.
[0:48] For example, an estimator spots an exclusion that applies to a particular type of job. Capture the correction. Have the responsible person approve it. Record which project or type of work it applies to. Do not transfer confidential client information into another project’s knowledge. Make it available to the relevant workflow. Then check that future drafts use it properly.
[1:14] That is managed company knowledge. It should not happen by silently treating every chat message as a new rule.
[1:21] How do you get people to use the system?
[1:23] Involve them before the design is finished. Sit with the surveyor doing the capture, the estimator checking the result and the manager making the commitment. Find out what would make each person’s working day easier.
[1:36] A surveyor may need fewer fields and a practical way to record observations. An estimator may need the original evidence beside the draft. A manager may need a short queue of exceptions requiring a decision.
[1:50] Train those activities. Show someone how to complete a real job, challenge an answer and recover when something is missing.
[1:57] Also be direct about why you’re introducing it. Explain the work you want to remove, how roles may change and who remains responsible. Do not promise that nobody’s role will ever change. Do give people a clear way to raise concerns and influence the implementation.
[2:15] Start with a small project team that includes an experienced sceptic. Account for poor connectivity on site and avoid making people enter the same observation twice. Give them a route to report problems, and show them what changed because of their feedback.
[2:32] Measure completed work, review time and repeat errors. Logins alone will not tell you whether the system is helping.
[2:42] If people keep returning to a spreadsheet, investigate what that spreadsheet still does better. Adoption is feedback about the design as well as the training.
[2:52] The aim is a construction business that keeps its useful knowledge and makes it easier for people to do good work together.
[3:00] If that is the gap you’re seeing, book a call with CSM. We can help you connect individual AI use to a repeatable way of working.
[3:08] End card: Book a call with CSM · calendly.com/rory-csm/csm-discovery-call
Start with one process that matters
Bring a workflow where delays, repeated work or unclear decisions are limiting the business. We can explore the outcome you want, what needs to change and where AI could help.