RFI & Document-Control AI Agent

What this agent does

Routes the question; it does not answer it

Log every inbound RFI against the drawing, specification and location it cites.

Log a repeat against the original thread rather than opening a second one.

Reason

Route each RFI to the named responder the project responsibility matrix sets.

Route anything carrying design content to the professional in responsible charge.

Report ageing, response-time breaches and the RFIs sitting on open delay events.

Decide

Report every out-of-date file in circulation against the register the CDE holds.

Route the drafted wording to the responder, who decides what is released.

Out

Log the question, its routing, every status change and the responder's decision.

Execute write actions only inside the approval boundaries agreed during implementation.

→ Product statement

The agent logs and routes; the responder in responsible charge answers — and once the field has built to an answer, the remedy is physical work, not an edit.

Example workflow

One RFI, inbox to response

AgentHuman

1RFI receivedContractor query, site question, submittal comment or coordination issue

2Register entry madeThe drawing, specification, location and revision cited, with the status the CDE holds for each

3Routing computedNamed responder, discipline, ageing clock, design-content flag and confidence

4Controls appliedResponsibility-matrix checks, design-content and scope tripwires, ageing clocks and confidence threshold

No human action required

Stages 1 to 4 run unaided, and nothing is answered or published at any of them — the agent is logging and routing, and the responder's lane opens at the confidence gate.

5DecisionBranches at the confidence threshold

High confidence

Goes to the named responder.

Low confidence

Adds a contracts read first.

Responder review

The RFI is held with the documents it cites, its flags and the confidence.

Answer · Reassign · Send to contracts

Answered — released by the responder▼

6Register and CDE updatedOnly where write access and approval policy allow it

7Outcome evaluatedResponse time, reassignments, design-content catches and answers disputed afterwards

Reassignments

Every reassignment made in review is counted in the evaluation.

What should not run autonomously

Human approval stays in control

Outside the boundary — human approval required8 items

Answering an RFI with design content.

Changing an information container's status code.

Issuing or distributing a drawing for construction.

Closing, voiding or merging an RFI.

Automation boundaryAgent acts unaided

✓Log, deduplicate and index inbound RFIs against drawings, specs and locations.

✓Route to the named responder and chase on the agreed clock.

✓Report ageing and response-time breaches.

✓Reconcile the register and flag stale files.

Any write happens inside the boundaries agreed at implementation, and never onto a status code.

Declaring a response a variation, a change, or no change.

Asserting that deemed acceptance has occurred.

Writing to or amending golden-thread records.

Deleting or suppressing a superseded revision or comment.

Example output

One RFI, annotated

Everything the agent logs is attached to the documents the question cites.

Package

Question

Open for

Routing

Confidence

Status held

Façade subcontract

Bracket fixing at the slab edge is not shown for this bay

6 days

Design content — engineer of record

95%

Shared, per the project BEP

As receivedTaken from the register and the CDE as it currently stands — nothing on this side is written by the agent.

What was read Cited drawing revision Specification section Responsibility matrix

Why this routingAn answer here would carry a dimension, so it goes to the licensee in responsible charge.

ActionAnswerReassignSend to contracts

What the score decidesBelow the configured threshold the RFI picks up a contracts read before the responder sees it.

Value

Where AI adds value

The same four claims, placed at the point in the workflow where each one applies.

Where the value landsValue 01 – 04

Every inbound RFIFrom the CDE and the register

03Routing

Apply the project's own protocol

Status codes and responder duties come from the project's BEP and information protocol, which is what governs here.

01Approved path

Route it to the right desk

Routine logging, deduplication and chasing happen before anyone opens the register.

02Human review

Send review to the design questions

Design-content and scope-change flags are marked, so the responder's read starts where licensure sits.

04Build an evidence trail

The question, its routing and every status change stay on the register.

Integrations

Typical integrations

Five system groups connect to the same agent. Which of them are in scope is decided in discovery.

CDE and document controlAconex · Autodesk ACC
Procore · Newforma

RFI workflowProcore · Newforma
Trimble Viewpoint For Projects

Model authoring and coordinationAutodesk Revit · Navisworks
Solibri · Revizto

Agent

RFI & document control

Reads the register
Routes the question
Holds for the responder

Markup and handoverBluebeam Revu · Dalux
ISO 19650 codes · project BEP

Observability & evaluationOpenTelemetry · Langfuse
Supported monitoring/evaluation sources

Integration availability depends on the client's existing systems and API access.

Agent controls

Six layers between the model and the field

Each control wraps the one inside it. What a layer does not catch is named in the failure-mode map below it.

L6 · Outermost — last line of defenceInward → L1 · closest to the model

L6Rollback / safe modePull the agent back to logging and routing only when evaluation or production signals degrade.Roll back

L5Version monitoringTrack model, prompt, matrix and CDE-configuration changes.Track

L4TraceabilityRecord the revision read, the routing, the responder and the decision behind each RFI.Record

L3Responder reviewHold responses for the named responder; it governs release, not whether the answer is sound.Gate

L2Policy guardrailsTest drafts against the design-content and scope tripwires; a failure returns the draft.Restrict

L1Confidence thresholdsRoute low-confidence RFIs to a contracts read before the responder sees them.Require review

Model coreDraft produced — routing, cited documents, flagged content and confidence

L1 – L2Test whether a draft may stand

L3Puts the release in the responder's hands

L4 – L5Keep the register and its status changes legible

L6Drops to routing only when signals degrade

How Nestack evaluates it

Evaluate the RFI workflow — not only the finished response.

Coverage runs the whole depth of the workflow, and every layer is cut by slice.

Surface — the answer the field builds to

Depth of coverage ▼

E1Final-output evaluationDid the response stay clear of design content and scope change?

E2Step-level evaluationDid the agent use the current revision, the matrix and the project protocol?

E3Tool evaluationDid it read and write the correct RFI and the correct register?

E4Confidence calibrationDo low-confidence RFIs actually attract more reassignments?

E5Slice evaluationHow does performance change across specific package types?

E6Business outcomeHow many responses needed rework, or were disputed once built to?

Floor — the work the field actually builds

Failure modes

Where each failure originates in the agent

Seven failure modes, each shown at the stage where it originates.

Agent lifecycleDirection of processing →

01 · Retrieval1 mode

RQ-01

Superseded revision returned

Sorted by upload time rather than the status the CDE holds.

Stage gathersRFIs, cited drawings, the register and the CDE status

02 · Reasoning2 modes

RQ-02

Scope change called minor

A scope-changing answer passes as a clarification and the work is unpaid.

RQ-03

Deemed acceptance asserted

Acceptance is claimed without the step the contract requires first.

Stage proposesRouting, cited documents and confidence

03 · Tool / write2 modes

RQ-04

Status code promoted

A shared container is moved to a published code and becomes an issue.

RQ-05

Ageing RFI auto-closed

A bare acknowledgement closes the item and stops the delay clock.

Stage writesOnly where write access and approval policy allow it

04 · Output1 mode

RQ-06

Design drafted as a response

A dimension or code reading is attributed to the responder.

Stage returnsThe response the responder signs and the field builds to

05 · Change / Version1 mode

RQ-07

Status semantics remapped

A platform update publishes containers under the old label.

Stage tracksModel, prompt, matrix and CDE configuration

Sev-1 · a held act performed by the agent Sev-2 · a wrong answer reaches the field Sev-3 · the register degrades, RFI routes to review

See the complete Construction failure-mode catalogue →

Affected slices

A clean aggregate can sit on one package

Aggregate response quality can look acceptable while a few packages carry nearly all of the releases that never reached a licensed responder. Nestack reports the rate by slice, not only in total.

Slice performance — reported separately, not only in aggregateIllustrative example

Slice	Failure rate	Lift	Status
Delegated-design packages	3.2%	3.6×	Review
Higher-risk buildings	2.5%	2.8×	Review
Traditional design-bid-build	1.6%	1.8×	Watch
Repeat-typology work	0.8%	0.9×	Normal

Bar: unsigned-release-rate lift vs. an all-response baseline · scale 0–4.0× · tick marks the 2.0× review threshold 2 of 4 slices over threshold

Evidence-linked improvement

A cycle closes on a test, not an answer

Closing means the failure is now a case the next release is measured against, not that someone has written up what happened.

Improvement cycle · five stagesSwitchback — the path turns at Improve and returns at Learn

01Detect

Unsigned-release rate rises in a package slice.

02Diagnose

Which step let it out — the routing, the tripwire, the register or the queue?

03Improve

A version stamp and the RFI thread travel with the correction.

04Verify

Release is held until the affected register cases clear.

05Learn

Both the suite and the responsibility matrix carry it from then on.

Learn → DetectThe return edge. The next detection runs against a suite one case longer.

Typical build scope

Twelve workstreams across six weeks

The build scope read against the delivery timeline. Week structure follows the six-week plan — discovery, sources, RFI workflow, evaluation, integration, then production validation and handover.

Workstream Week 1Week 2Week 3Week 4Week 5Week 6

01RFI workflow discovery and boundary definition.

02CDE, RFI and model source assessment.

03Responsibility-matrix and status-code mapping.

04RFI ingestion, deduplication and indexing.

05Routing logic and design-content tripwires.

06Confidence scoring and RFI routing.

07Responder review workflow.

08CDE and project-system integration.

09Response-quality regression cases.

10Guardrails and release controls.

11Register and status instrumentation.

12Deployment, documentation and Agent Care handover.

12 workstreams · 6 weeks · bar shows the weeks a workstream is active — several run in parallel Final scope and sequence confirmed in discovery

Engagement tiers

What each tier includes

Rows are the capabilities named in each tier's scope. Higher tiers include everything below them.

Capability✓ in scope · — not at this tier PilotOne project, one CDE ProductionProduction CDE and RFI systems AdvancedMultiple projects / CDEs

Introduced at Pilot

Logging and routing on your matrix✓✓✓

Responder review✓✓✓

Response-quality baseline✓✓✓

Introduced at Production

Reporting by package—✓✓

Review workflow in your systems—✓✓

Approved write-back—✓✓

Analytics feed for design teams—✓✓

Introduced at Advanced

Multi-project CDE rules——✓

Multi-stage responder review——✓

High RFI volume——✓

Enterprise CDE controls——✓

Build price From $5,000 From $8,000 Custom quote

Final build priceConfirmed after discovery based on integrations, workflow complexity, RFI volume, approval controls and deployment requirements.

Separate from buildBuild pricing is separate from recurring Agent Care, which covers managed monitoring, evaluations, incidents and verified improvements after launch.

What we need from you

What you bring, and what we build with it

Each input maps to a piece of build scope and a week in the delivery timeline.

You bringWe build with it

01Your responsibility matrix and named responders → Routing rules and responder mappingWeek 1

02Representative RFIs and their responses → Routing baseline, deduplication and design-content tripwiresWeek 2

03Your BEP and information protocol → Status-code mapping and automation-boundary definitionWeek 1

04Access to relevant APIs, feeds or exports → CDE, RFI and model system assessment, then integration setupWeek 2

05RFIs that turned out to be design questions → Response-quality cases and failure-mode testingWeek 4

06Which responses may never leave without a licensed read → Confidence scoring, RFI routing, guardrails and approval controlsWeek 3

07Named responders to review the routing → Responder review workflow, then pilot and production validationWeeks 5–6

Nothing else is required Deployment, documentation and Agent Care handover are ours.

Delivery timeline

Four phases across six weeks

Each phase sits on the weeks it really occupies. Week 5 runs evaluation and pilot at the same time.

Phase W1W2W3W4W5W6

Discovery W1

Build W2 – W3

Evaluate W4 – W5

Pilot & Launch W5 – W6

Week focus W1RFI workflow discovery, protocol mapping and the automation boundary W2Source integration and the routing baseline W3Tripwires, confidence logic and review controls W4Evaluation suite, guardrails and failure-mode testing W5CDE integration, pilot packages and targeted corrections W6A live RFI cycle answered under the responder, then handover

Reading the bandEach bar covers the weeks its work is named in and no others. The week 5 overlap is real, not padding.

At the end of W6Sign-off on the pilot register, then continuous monitoring under Agent Care.

DurationSix-week plan shown · typical delivery 4–6 weeks depending on scope confirmed in discovery.

Next step · Construction AI agent

Build an RFI agent around your responsibility matrix.

Show us your CDE, your matrix and who sits in responsible charge. If your RFIs carry design content, the licensee answers and seals — we map what the agent hands them, and what it never touches.

Discuss this agent →

Nestack Agents · RFI and document controlAGT-CN-04 · Agent Care available after launch