Scope Agent — ai scope drift detection for aec projects
The Scope Agent reads the signed contract scope of services and watches RFIs, meeting minutes, design submissions, and email correspondence for asks that fall outside that scope. It treats the contract as the canonical reference while monitoring every in-flow signal. It is currently shipping with 1–3 design partners, with A-2 contracted under the Beta program.
- Flags each scope drift with the specific RFI, meeting minute, or email that triggered it and the exact clause in the contract that the ask falls outside.
- Prepares a change-order draft automatically when drift exceeds the contract’s de-minimis threshold, ready for the project manager to review, edit, and submit.
- Builds a per-project scope timeline showing what was in-scope, what drifted, and what was change-ordered or absorbed, defensible at closeout and during any dispute review.
What the Scope Agent does.
-
Contract-scope ingestion + canonical reference
The Scope Agent reads the signed scope of services from architectural services agreements, design-build agreements, and owner-architect agreements and turns them into a structured reference for ai scope drift detection for aec projects. Each flag cites the specific article, paragraph, or exhibit the request falls outside. Ambiguous clauses are surfaced at deployment so the project manager can confirm intent before in-flow monitoring starts, and the clarified interpretation is logged in the same project rule pack as the Scope Drift Detection workflow.
-
In-flow scope monitoring
It monitors RFIs, design submissions, meeting minutes, email threads, and owner directives for phrases and requests that imply new deliverables, expanded scope, or accelerated milestones. The agent connects to tools such as Procore for RFIs and meeting minutes, and to shared mailboxes or project portals where owner directives land. Project managers see drift the day it appears in the record, not at month-end billing, so they can decide whether to push a change order, adjust internal effort, or decline the request.
-
Drift classification and severity tagging
Every scope-drift event is classified by type: added scope, expanded scope, accelerated scope, or changed performance criteria. The agent tags each event with a severity score based on contract language, fee structure, and any pre-configured de-minimis thresholds calibrated per deployment. Low-severity items collect in a watch list for pattern review; material items surface in a high-priority queue that can be reviewed alongside Contract Clause Monitoring so commercial risk and scope risk stay aligned for the same project.
-
Change-order drafting
When a flagged drift exceeds the configured threshold, the Scope Agent drafts a change order that cites the relevant contract clauses, describes the added or changed work, and proposes a fee basis and schedule impact grounded in the project record. It can reference prior similar changes on the same project, pulling history from the Memory Agent where available. The project manager edits, negotiates, and submits; the agent never files anything directly with the client or owner system, and it never asserts a fee the underlying documents do not support.
-
Scope timeline and audit log
For each project, the agent maintains a scope timeline starting from contract signing through major milestones, RFIs, directives, and change orders. It records which requests were flagged, which were change-ordered, which were absorbed, and which remain unresolved, with links back to the source emails, RFIs, and minutes. This timeline is designed for project closeout, owner audits, and dispute preparation, and pairs with the RFI Agent to show how unanswered or late RFIs interacted with scope drift over the life of the job.
-
Guardian-family scope protection
This agent sits in the same Guardian family as the Contract Agent, focusing specifically on scope rather than insurance or indemnity terms. It stays inside the four corners of the signed agreement, flags drift, and links every action to a contract clause and a dated project communication. In practice, The Guardian protects the firm from its own nicely-worded emails that quietly accept extra work, and gives project managers a defensible record when they push back on owner or contractor requests.
Scope Agent — common questions
-
How does the Scope Agent learn what’s in scope?
The Scope Agent learns scope from the signed contract documents for each project, treating the scope of services as the canonical reference rather than informal emails. During deployment, it parses the agreement into a structured rule pack that defines deliverables, exclusions, and de-minimis thresholds, calibrated per deployment. Where clauses are ambiguous or conflicting, it flags them for a project manager or commercial lead to clarify before in-flow monitoring starts, and records those clarifications so future flags and Scope Drift Detection decisions stay consistent.
-
What about scope drift on design-build vs. owner-architect contracts?
Both design-build and owner-architect contracts are supported, along with IPD and construction-management-at-risk structures. For each Beta deployment, the implementation team identifies which party’s scope is being monitored—architect, contractor, or both—and configures the rule pack accordingly. That configuration can include different drift thresholds for design vs. construction scope, and can align with how your Contract Agent tracks payment terms, performance criteria, and risk allocations across the same agreement.
-
Does it submit change orders for us?
No, the Scope Agent prepares drafts but never submits change orders on your behalf. It assembles the narrative, clause references, and a fee basis from the project record, giving the project manager a starting point that typically takes single-digit minutes to review per item. The PM or commercial lead then edits language, sets final pricing, and submits through your normal channel—email, owner portal, or a platform such as Procore—so commercial decisions stay with humans while documentation work is automated.
-
How does this work alongside the Contract Agent?
The Scope Agent and the Contract Agent are designed to sit together as part of the Guardian family. Contract Agent watches clause-level obligations such as insurance, indemnity, payment timing, and performance criteria, while Scope Agent monitors deliverable drift against the same documents. Studios running both agents see one shared rule pack, so when a scope drift touches a performance or liquidated-damages clause, both agents flag it and the project manager can see scope, risk, and fee exposure in a single view, backed by the Memory Agent for historical context.
-
What about scope drift the architect causes (designer-driven changes)?
The Scope Agent flags scope drift regardless of who initiates it—owner, contractor, or the design team itself. If a designer upgrades finishes, adds coordination passes, or revises layouts beyond what the contract describes, the agent treats those as potential drift events and logs them against the same scope timeline. That record helps practice leaders see where internal over-design is eroding fee, and supports conversations about design standards, staffing, and when to treat certain upgrades as marketing vs. chargeable scope, especially when combined with the RFI Agent for context on coordination pressure.
-
What does the Beta cohort look like?
The Scope Agent is currently shipping with 1–3 design partners, with A-2 contracted as one anonymised engagement. Beta cohort firms are AEC practices—architects, project managers, or contractors—with documented architectural-services or design-build contracts and an active portfolio of at least 5 live projects. Teams with existing contract workflows in tools like Procore or who already use Guardian-family workflows such as contract clause monitoring are especially well-suited, and can apply through the “Apply to the Beta” CTA on this page.