A project meeting should have room for decisions, not just the assembly of an update.
For property and construction teams, preparing that update can involve chasing progress, reconciling spreadsheets, reading meeting notes and explaining changes. AI may help with parts of this work, provided the resulting information is dependable.
The Association for Project Management’s public AI resources bring together learning, research and case studies. Its guide, How to Use Artificial Intelligence in Your Projects, is presented as a practical introduction covering value across the project life cycle, responsible adoption and the continuing importance of human judgement.

01 / THE OUTCOME
Start with the project outcome
Choose the improvement you want the team to experience. That might be less time preparing a monthly report, quicker access to agreed actions or a clearer view of unresolved information.
Define the current process before choosing a tool. Who supplies information? Who reconciles it? Where does it wait for clarification? Who uses the finished output, and what decision does it support?
For a reporting assistant, a useful first objective could be to prepare a source-linked draft from approved project records, with missing information highlighted for the project manager. Agree which parts of the report remain outside the trial.
Measure the time and problems in the existing workflow so you have a basis for comparison.
02 / INFORMATION QUALITY
Improve the information first
APM’s Network Rail case study describes work in its Wales & Western Region to bring fragmented project information together, improve data quality and support teams. It presents those foundations as preparation for further AI adoption, rather than attributing every reported improvement to AI.
For a smaller project team, the practical starting point can be one report and a few agreed sources. Confirm the reporting cut-off, revision status, owners and definitions.
If one system records committed cost and another records a forecast, make that distinction explicit. If progress has not been updated, show it as missing or stale. A well-written summary should not hide that uncertainty.
Fix repeated information problems as part of the trial. Otherwise the team may simply spend its saved drafting time investigating the same inconsistencies.
03 / REPORTING WITH CONTEXT
Separate facts, forecasts and decisions
Project records contain statements with different levels of certainty and authority. A suggested action in a meeting note is not necessarily an agreed action. A proposed variation is not an approved change.
- Recorded fact
What has happened? - Check the dated source, reporting period and project.
- Forecast or assumption
What might happen? - Identify the basis, uncertainty and person responsible for the forecast.
- Decision or approval
What has been authorised? - Confirm the decision record and the person with authority to approve it.
Ask the system to preserve those distinctions and point to the underlying record. Give reviewers a way to see what changed since the last report, including corrections to earlier information.
The person issuing the report should resolve material ambiguities. If an issue needs a sponsor or change authority decision, route it through that process and record the outcome.
04 / PERMISSIONS
Control access and actions
A project assistant may connect to minutes, commercial records, correspondence and programme information. Define which records it can access and make sure existing project boundaries remain effective.
Reading information, updating a live register and sending a report are separate permissions. Start the trial with the actions needed for the defined task, and agree who can authorise any wider access.
Check how the chosen service stores, processes and retains information. Where personal data is involved, use the ICO’s UK AI resources to inform the assessment.
For a meeting assistant, also agree how participants will be informed, which meetings are suitable and how recordings or transcripts will be handled.
05 / PEOPLE AND ADOPTION
Support the people doing the work
Bring the people who prepare and use the information into the trial. They can identify awkward exceptions that a demonstration might miss.
Train reviewers to challenge a plausible summary, identify an unsupported statement and send an issue back for clarification. Show users what the system can do, where it needs help and how to continue the work if it is unavailable.
Keep a simple route for feedback. If people repeatedly rewrite one section or avoid a feature, investigate the cause before expanding it to more projects.
Assign a business owner who can make decisions about the process, alongside the support needed to keep the service working.
06 / EVIDENCE OF VALUE
Measure the finished process
Compare the complete reporting cycle, including input preparation, checking, correction and issue. Also record omissions, outdated information, unclear actions and unsupported claims.
Ask whether the finished report helps its audience. Can the sponsor identify the decision required? Can a workstream owner find the action they need to take? Is the project manager spending more time resolving the project’s issues?
Agree success and stop criteria before testing, then decide whether to refine, extend or stop. Our AI ROI guide explains how to include review effort and the full cost of implementation.
Through BUILD™, we use that evidence to develop a useful process and support the people who depend on it.