Project Memory Agent — ai project memory for aec, claim-ready timelines on demand
The Project Memory Agent ingests project correspondence, meeting minutes, RFIs, design submissions, change orders, and contract documents into a structured project memory, then produces decision logs, claim-ready timelines, and recall-on-demand summaries. It focuses on emails, chat channels, minutes, RFI platforms, and contract files. It is currently shipping with 1–3 design partners, with A-1 contracted.
- Project decision log — every decision, every change, every commitment, with the source document, the date it was made, and the parties involved for each entry.
- Claim-ready timeline on demand — when a dispute, claim, or insurance review needs it, the timeline generates in hours, not weeks, calibrated per deployment to the project’s ingested record and document volume.
- Recall-on-demand — practitioners ask plain-English questions (“when did we agree to remove the third basement level?”) and get cited answers from the project record, including document excerpts and timestamps.
What the Project Memory Agent does.
-
Multi-source memory ingestion
The Project Memory Agent reads project correspondence (email threads, Slack and Teams channels per Labs scoping), meeting minutes (Word or Google Docs, PDF packs, and video-meeting transcripts), RFIs from platforms such as Procore and BIM 360 / ACC, design submissions, change orders, and contract documents into a structured project memory. Memory is per-project and scoped to the project record, not the whole firm. Each Beta deployment calibrates which sources are canonical and how frequently the ingestion runs (~daily or ~weekly).
-
Decision-log generation
From the structured memory, the agent extracts every decision it can find — design decisions, scope decisions, payment approvals, schedule moves — into a structured log. Each entry carries the source document, the date, the participants, and a short summary of the decision. Teams use this log at closeout, at each major milestone, and when preparing dispute-evidence timelines so they do not reconstruct history from scattered inboxes.
-
Claim-ready timeline production
When a dispute, claim, owner-side audit, or insurance review needs a project timeline, the agent produces it on demand from the structured memory. The timeline orders events across RFIs, meeting minutes, change orders, and contract notices, and cites each event back to its source document with date and provenance. This same timeline underpins schedule-drift analysis for project timeline drift reviews and can be exported for counsel to annotate without editing the underlying record.
-
Recall-on-demand search
Practitioners ask plain-English questions such as “what was the structural engineer’s response on the cantilever?” or “when did we first flag waterproofing at the podium?” and receive cited answers from the project record. Responses include document snippets, dates, and links back to the original file or RFI. This is especially useful when senior staff turnover loses context that the project memory still holds, so the new lead queries the agent instead of trawling years of email or re-reading every meeting-minutes-to-action-items thread.
-
Personality trait — The Historian
This agent personifies The Historian: it remembers every decision, every change, and every commitment across the project record long after individual inboxes have been archived. The Historian remembers what no one wrote down twice, including side agreements captured only in a single email or meeting note. That persistent recall gives project directors a single place to confirm “what actually happened” before they respond to an owner, contractor, or insurer.
-
Pairs with the Guardian agents (Scope, Contract, RFI)
Memory provides the underlying project record; the Guardian agents — the Scope Agent, the Contract Agent, and the RFI Agent — act on it. Studios deploying Memory alongside the Guardian family see one structured project memory feeding all four agents, with no duplicate ingestion and no inconsistent timelines. The same events that support a claim-ready timeline also drive scope-change alerts, contract-clause checks, and RFI history views for both live coordination and later dispute review.
Project Memory Agent — frequently asked questions
-
What sources does the Memory Agent ingest?
The Memory Agent ingests project correspondence such as email threads and scoped Slack or Teams channels, meeting minutes in Word, Google Docs, or PDF, and video-meeting transcripts where available. It also reads RFIs from platforms like Procore and BIM 360 / ACC, design submissions, change orders, and contract documents stored in your DMS or CDE. Each Beta deployment scopes which of these sources are canonical for the customer’s portfolio so the ai project memory for aec stays aligned with how the firm actually works.
-
Where does the project memory live?
Project memory is maintained per project, not as a single firm-wide blob, and can be hosted in the customer’s preferred environment. Options include a VitruAI-hosted store under MSA, a customer-hosted store via on-prem deployment, or a hybrid model where text is indexed by VitruAI while source documents remain in the customer’s systems. Each Beta deployment picks the canonical model and documents where the memory sits in relation to existing PM tools and file servers.
-
How does this differ from a project-management platform’s audit log?
Project-management platforms record structured events — RFI status changes, submittal approvals, and task movements — but they rarely capture the full correspondence and minutes where decisions are actually made. The Memory Agent reads those emails, chat threads, and meeting notes alongside the PM audit log, so the combined record shows both the formal event and the conversation that led to it. That combined record is what holds up under dispute review or a dispute-evidence timeline request, where the PM audit log alone can miss the email that established the real agreement.
-
What does “claim-ready” mean for the timeline?
Claim-ready means the timeline is defensible across legal or insurance review because every event is cited back to its source document, with the document’s date, author, and provenance recorded. The agent does not make legal judgments or recommend strategy; it assembles the factual sequence that counsel, insurers, or commercial managers work from. Firms use the same claim-ready timeline as the backbone for project timeline drift analysis, extension-of-time discussions, and negotiation prep with owners or contractors.
-
How does this work alongside the Scope, Contract, and RFI Agents?
The Memory Agent provides the underlying structured project record, while the Guardian agents — Scope, Contract, and RFI — act on specific slices of that record. Scope focuses on scope drift and change events, Contract focuses on clause triggers and notice periods, and RFI focuses on question-and-response history. Studios deploying all four see one ingestion pipeline feeding multiple downstream checks, so the same event that appears in a claim-ready timeline also drives a contract notice reminder or an RFI aging alert.
-
What does the Beta cohort look like?
The Project Memory Agent is currently shipping with 1–3 design partners, with A-1 contracted as the first deployment and the same core agent shipping as A-1’s CommsAgent for meeting-minutes-to-action-items. Cohort criteria include an AEC portfolio of at least 5 live projects, willingness to ingest correspondence and minutes as part of the project record, and a project-management or PMO function ready to consume decision logs and timelines. Firms that meet these criteria can apply via the “Apply to the Beta” CTA and scope how Memory supports their dispute, audit, and closeout workflows.