Knowledge / AEC-native data
Meeting minutes and decision records
Minutes of design team, progress and coordination meetings, holding the decisions, actions and agreed changes that rarely make it back into the drawings.
Minutes are the only place many decisions are ever written down. A design team meeting agrees to change a detail, a progress meeting records a verbal instruction, and neither reaches the drawing until someone remembers. Minutes are written by whoever chairs, in whatever format the chair prefers, with numbered actions carried forward week after week until they are closed or dropped. Teams read the current set, look for their own name, and file the rest. Construction meeting minutes AI is useful only if it reads the whole run of meetings as one record, so that a decision, its reversal and the drawing revision that answered it can be questioned together, with every answer traceable to the minute it came from.
What teams need to ask of it
When was the change to this detail agreed, in which meeting, and by whom?
Was that decision later reversed or qualified in a subsequent meeting?
Which actions are still open, who owns each one, and how many meetings has it been carried forward?
Was the agreed change ever instructed, or does it exist only in the minutes?
Which drawing revision responds to this minute item, and has it been issued?
Which decisions in the minutes contradict the current drawings or specification?
What did the client agree to in recent client meetings that carries cost or programme consequences?
Why generic AI gets it wrong
A set of minutes looks like prose and behaves like a register. Items are numbered, actions carry the same reference across twenty issues, and a single line such as "closed, see revised drawing" only makes sense if you know which meeting it appeared in and which line it closes. Flatten a folder of minutes into text and that goes. A general-purpose tool cannot tell whether item 4.3 in the March minutes is the same action as item 4.3 in July or a new one that reused the number, and it has no way of knowing that the drawing the item refers to was revised twice after the action was closed.
The larger problem is that a decision in a minute is not a fact about the building. It is a record that somebody agreed something on a date, which may have been reversed three meetings later, never instructed, or overtaken by a drawing revision that shows something else. A tool reading minutes alone will find the sentence where the decision was made, quote it accurately, and stop. The reversal is in another document, the instruction never came, and the drawing shows the old detail. The answer reads like the minute, cites the meeting, and is wrong in the way that matters most: it tells a design manager a change is settled when it is not.
One question, answered
QUESTION
When was the change from precast to in-situ stair flights agreed, was it ever instructed, and which drawing revision shows it?
ANSWER
The change was first raised at design team meeting 14 as a cost saving proposed by the frame subcontractor, recorded as an action on the structural engineer to confirm feasibility. It was agreed at design team meeting 16, item 5.2, subject to the engineer confirming the landing connection detail, and the action was carried as open through meetings 17 and 18. No instruction has been issued: the change does not appear in the instruction register, and the contractor's proposal has not been formally accepted. Drawing 2301 rev C04, the stair core sections, was issued after meeting 18 and shows in-situ flights, so the drawing has moved ahead of the contractual position. Progress meeting 22 records the client's commercial team asking why the change had not been instructed, and that action remains open.
SOURCES
design team meeting 14 minutes, item 5.2; design team meeting 16 minutes, item 5.2; design team meeting 17 and 18 minutes, carried-forward actions; instruction register to date; drawing 2301 rev C04, stair core sections; progress meeting 22 minutes, item 3.4.
The project, documents and references are illustrative examples only.
We onboard a focused slice of your project data, configure the ingestion paths, extract a structured context graph, and build a workflow against the resulting structured knowledge layer.
Related to
Minutes sit between the documents that raise questions and the documents that answer them. An RFI is often opened because a minute could not settle the point, a change order or variation should follow every change agreed in a meeting, and the construction drawings are where the decision finally has to appear. Project knowledge search uses minutes as the record of intent, and lessons learned capture reads them for the decisions that turned out badly.
Which meeting minute formats does this apply to?
Does it follow an action across several meetings?
How do we see it on our own minutes?
Related pages
RFIs
Formal requests for information raised during construction to resolve ambiguities or conflicts in the drawings and specifications, with their responses.
Construction drawings
The 2D graphical record of design intent: plans, sections, elevations, details, schedules and title blocks, issued as sheets with revisions.
Change orders, variations and compensation events
Instructed or notified changes to scope, time or cost, and the notices, quotations and assessments that follow under the contract.
Project knowledge search
Asking questions across the whole project record, drawings, models, specs, contracts and correspondence, and getting answers that cite their source.
Lessons learned and institutional knowledge capture
Capturing decisions, defects, precedents and outcomes from completed projects into structured knowledge the next project can reuse.