Request for Information (RFI)

Definition — what it is

An RFI is a formal, numbered written question issued by a contractor or subcontractor to the design team or owner seeking clarification of the contract documents. It exists because drawings and specifications are never complete: they are a representation of intent produced under time pressure, and the field inevitably encounters conditions the documents do not resolve. The RFI creates a written record of the question asked, the date asked, the answer given, and the date answered — which is precisely why it is as much a risk instrument as a communication tool. An RFI is not a change order and does not by itself authorize additional work or cost, but the answer to one frequently triggers both.

Also known as: Request for Interpretation, Clarification Request, Information Request

Why it matters — what it protects

The RFI is the primary mechanism by which design ambiguity is converted into a documented decision. Without it, field crews interpret drawings on their own, and those interpretations become permanent, expensive, and unattributable. A contractor who builds from an assumption rather than an answer owns the consequence of that assumption.

RFIs are the leading indicator of schedule risk. An unanswered RFI on a critical-path activity stops work as surely as a missing material delivery, and the aging of the open RFI log is often the earliest quantitative signal that a project is heading for delay. Experienced project executives read the RFI log before they read the schedule.

RFIs are evidence. In a dispute, the RFI log establishes what the contractor knew, when they raised it, and how long the design team took to respond. Time-to-response on an RFI is a routine element of delay claims and time impact analyses, because contracts typically specify a response window and the record shows whether it was met.

Volume itself is a signal about document quality. An unusually high RFI count per million dollars of contract value indicates incomplete or uncoordinated design, which is a predictor of change order volume, rework, and margin erosion. Owners increasingly track this metric across their design consultants.

Lifecycle — how it moves

1. Identification

   A field engineer, superintendent, foreman, or subcontractor encounters a condition the documents do not resolve — a dimension that does not close, a detail referenced but not drawn, a conflict between architectural and structural sheets, or a specification that contradicts the drawings.

2. Screening

   A good project team filters before it writes. Many apparent RFIs are answered by reading the specification section, checking a prior addendum, or searching the existing RFI log for the same question already asked. Screening is where mature teams cut volume dramatically.

3. Drafting

   The question is written with the drawing and specification references, a description of the condition, photographs or a marked-up detail, and — critically — a proposed solution. RFIs that propose an answer are resolved materially faster than RFIs that only ask.

4. Internal review and issue

   The project manager reviews for clarity, cost and schedule implication, and whether the question should instead be a change notice. The RFI is numbered, logged, and transmitted through the contractually specified channel with a required-by date tied to the activity it affects.

5. Design team review

   The architect coordinates with the relevant consultant — structural, MEP, civil — and issues a response. Complex questions may be routed to multiple consultants, which is where response time typically degrades.

6. Response and distribution

   The answer is returned, logged against the original, and distributed to everyone whose work it touches. Distribution failure is a common and expensive breakdown: the answer exists but the crew never receives it.

7. Downstream action

   If the answer changes scope, cost, or duration, it triggers a change event and potentially a PCO. If it clarifies without changing scope, it is filed as a document interpretation and may need to be reflected in as-builts.

8. Closure and record

   The RFI is closed, incorporated into the as-built record set, and retained. The closed log becomes part of project closeout and the evidentiary record for any later dispute or warranty question.

Anatomy — the data it carries

RFI number

Sequential, project-unique identifier. Often prefixed by discipline or subcontractor for sortability.

Subject

A short, specific question title. Vague subjects are the primary cause of misrouting and delay.

Date submitted / required-by date

The required-by date should be derived from the schedule activity it blocks, not chosen arbitrarily.

Originator and company

Who is asking — often a subcontractor asking through the general contractor, which creates a two-tier response chain.

Discipline / responsible party

Architectural, structural, mechanical, electrical, plumbing, civil. Determines routing and heavily influences response time.

Drawing and specification references

Sheet numbers, detail callouts, and spec sections. Missing references are the single most common reason an RFI is returned unanswered.

Question / description of condition

The factual condition encountered, ideally with photographs, survey data, or a marked-up detail.

Proposed solution

The contractor's recommended resolution. Converts the RFI from an open question into an approval decision, which is far faster to answer.

Cost and schedule impact flags

Whether the contractor believes the answer will carry cost or time. Sets expectations and preserves notice rights.

Response and responder

The design team's answer, who authored it, and the date returned.

Attachments

Photos, sketches, marked drawings, product data, survey results.

Linked records

Related change events, PCOs, submittals, drawing revisions, and ASIs generated by the answer.

Status and aging

Open, answered, closed, void — plus days outstanding, the field that actually drives management attention.

Failure modes — how it breaks

The RFI that is really a change notice

Teams sometimes use an RFI to raise what is plainly extra work, because an RFI feels less confrontational than a change notice. This can forfeit notice rights: contracts usually require written notice of a change within a defined window, and an RFI may not satisfy it. Ask the question, but preserve notice separately.

Duplicate and re-asked questions

On large projects with multiple subcontractors, the same question is asked repeatedly because nobody searches the existing log. Duplicates inflate volume, consume design-team capacity, and slow response times for genuinely new questions.

Missing references and context

An RFI without sheet numbers, spec sections, or photos forces the design team to reconstruct the condition. It gets returned with a request for clarification, and a week evaporates on a round trip that added no information.

Answer received, never distributed

The response is logged in the project management system but never reaches the foreman who raised the condition. The crew proceeds on its original assumption and the rework is discovered at inspection.

Aging invisible until it is critical

Open RFIs are reviewed weekly in a meeting rather than monitored continuously against the schedule. By the time an overdue RFI is discussed, the activity it blocks is already delayed and the float is gone.

Non-answers accepted as answers

Responses that say 'refer to contract documents' or 'per plans and specs' close the RFI in the system without resolving the condition. Status shows answered; the field is no better off.

No linkage to cost

The RFI answer changes scope, but no change event is created. The cost surfaces weeks later as an unexplained variance in job cost, long after the leverage to negotiate it has passed.

Metrics — how it is measured

Average response time

Days from submission to substantive answer, measured against the contractual response window. The core accountability metric on both sides.

Open RFI aging

Distribution of days outstanding for open items, typically bucketed 0-7, 8-14, 15-30, 30+. The 30+ bucket is where schedule damage lives.

RFI density

RFIs per million dollars of contract value, or per thousand square feet. Benchmarks design completeness and predicts change volume.

Critical-path RFI count

Open RFIs tied to activities with little or no float. Far more actionable than total open count.

Duplicate rate

Share of RFIs answered by pointing to an existing answer or document. Measures screening discipline.

Cost conversion rate

Percentage of RFIs that generate a change event or PCO, and the aggregate dollar value. Links the log to margin.

Rejection / clarification-return rate

Share returned incomplete. Measures the quality of RFI drafting.

The AI shift — what actually changes

Conversational

The RFI log stops being a spreadsheet you read and becomes something you interrogate. Instead of filtering columns, you ask which open RFIs touch critical-path activities, which have been outstanding past the contractual window, and which relate to the same detail — and get an answer with the underlying records cited.

Generative

Drafting shifts from a blank form to a reviewed draft. Given a photograph, a sheet reference, and a sentence of context, a model produces a properly structured RFI with the specification sections identified, the condition described in contract language, and a proposed solution — which the engineer edits rather than composes.

Orchestrated

The RFI stops being an isolated document. Its answer is automatically checked against the schedule activity it blocks, matched to related submittals and drawing revisions, screened against the existing log for duplicates, and — where the answer changes scope — used to open a linked change event so cost never gets separated from the question that caused it.

Autonomous

The routine motion runs without a person driving it: new RFIs screened for duplicates and completeness on submission, references validated against the current drawing set, aging monitored continuously against schedule float, reminders escalated on the contractual clock, and answers distributed to exactly the crews whose work they affect — with humans deciding the substance and approving anything that carries cost.

Prompts — put it to work

Tool-agnostic and copy-ready. Adapt the specifics — thresholds, contract windows, cost codes — to your own project before you run them.

Conversational — Monday morning triage of an RFI log you did not write.

Review our open RFI log. Identify every RFI that is (a) outstanding longer than the 10 business-day contractual response window, or (b) linked to a schedule activity with less than 5 days of total float. For each, give me the RFI number, subject, responsible discipline, days outstanding, the activity it affects, and the schedule consequence if it is not answered this week. Rank by risk, not by age.

What good output looks like: A ranked table of at-risk RFIs with the schedule linkage made explicit, not just a list sorted by date — plus a clear separation between items that are merely old and items that are actually blocking work.

Follow-ups:

• Which of these are waiting on the same consultant? Group them so I can make one call instead of six.
• Draft a short escalation email to the architect covering only the items past the contractual window.
• Which of these look like they will carry cost, and what is the likely exposure?

Generative — A foreman sends a photo of a conflict from the field and needs a real RFI written.

Draft a formal RFI from the following field condition. Photo attached. Context: at grid line C-4, level 3, the 12-inch supply duct shown on M-301 conflicts with the W21x44 beam shown on S-204; there is roughly 4 inches of clearance where 14 inches is required. Write it as a complete RFI: subject line, description of the observed condition in contract-appropriate language, all relevant drawing and specification references, a clearly labelled proposed solution, cost and schedule impact flags, and a required-by date given that ductwork on level 3 starts in 8 working days. Keep the tone factual and non-accusatory.

What good output looks like: A submission-ready RFI with correct references, a specific proposed resolution, and a required-by date derived from the schedule — not a generic template with blanks.

Follow-ups:

• Rewrite the proposed solution to offer two options with different cost profiles.
• Add the notice language we need to preserve our rights if this turns into extra work.
• Produce a one-paragraph version I can text to the superintendent.

Orchestrated — An RFI answer just came back and you need to know everything it touches.

RFI 214 was answered today. Read the response and trace its full impact across the project: does it change scope relative to the contract documents; which schedule activities are affected; are there open submittals, shop drawings, or material procurements that now need revision; does it contradict any previously issued RFI answer or ASI; and does it require a change event. Return a single impact summary with each conclusion tied to the specific document or record that supports it, and flag anything you are not confident about rather than guessing.

What good output looks like: A cross-referenced impact assessment spanning schedule, submittals, procurement, and cost — with citations to the underlying records and explicit uncertainty flags.

Follow-ups:

• Open a change event for the scope delta and draft the cost narrative.
• List every person and subcontractor who needs this answer, and draft the distribution note.
• Does this answer invalidate any work already installed?

Orchestrated — Screening a new RFI before it consumes design-team capacity.

Before this RFI is issued, screen it: search the existing RFI log, issued addenda, ASIs, and the specification sections for an answer that already exists. If the question has already been answered, tell me where and quote the answer. If it has not, check that the drawing and specification references are correct against the current drawing set, verify the required-by date is consistent with the schedule activity it affects, and tell me what is missing that would cause it to be returned incomplete.

What good output looks like: Either a citation to the existing answer, or a specific completeness checklist with each gap named — the difference between screening and rubber-stamping.

Follow-ups:

• Rewrite it to fix everything you flagged.
• How many RFIs in the last 90 days were duplicates of an existing answer?

Autonomous — Standing policy for how the RFI process should run itself.

Operate our RFI process continuously under these rules. On submission: screen for duplicates against the full log, validate drawing and specification references against the current issued set, and return anything incomplete to the originator with the specific gaps identified. While open: monitor aging against the 10 business-day contractual window and against schedule float, escalating to the project manager at 7 days, the design team lead at 10, and the owner's representative at 15. On response: distribute to every affected crew and subcontractor, flag any answer that appears to change scope, and open a linked change event when it does. Never close an RFI whose response does not substantively answer the question, and never approve cost — route every cost implication to me with your reasoning.

What good output looks like: A running process with a complete audit trail, where the human sees a short exception queue rather than the entire log — and where cost decisions never happen without a person.

Follow-ups:

• Show me everything you handled this week and everything you escalated.
• Which of your escalations did I override, and what should you learn from that?

Get the full Construction AI Prompt Catalog — every prompt in the library in one document.

Maturity — locate yourself honestly

0. Level 0 — Email and memory

   RFIs move as emails and attachments. There is no reliable log, aging is unknown, and the record has to be reconstructed from inboxes when a dispute arises.

1. Level 1 — Logged

   A central register exists with numbers, dates, and statuses. Aging is visible in a weekly report. Screening and distribution are still entirely manual.

2. Level 2 — Linked

   RFIs are connected to schedule activities, submittals, and change events. Aging is measured against float rather than just calendar days, and cost implications are traceable.

3. Level 3 — Assisted

   Drafting is model-assisted from field input, duplicates are caught on submission, references are validated automatically, and impact analysis of answers is generated for human review.

4. Level 4 — Operated

   The routine loop runs unattended inside defined guardrails — screening, validation, aging escalation, distribution, and change-event creation — while humans own substance, cost, and anything the system flags as uncertain.

Common questions

Does an RFI authorize extra work?

No. An RFI is a question and its answer is an interpretation of the contract documents. Authorization for additional scope, cost, or time comes through a change order or a construction change directive. Proceeding on an RFI answer alone, without converting a scope change into a priced and executed change, is one of the most common ways contractors perform work they never get paid for.

What is a reasonable RFI response time?

Most contracts specify between 7 and 14 calendar days, with 10 business days common on commercial work. The contractual window is the floor, not the target: what actually matters is whether the answer arrives before the activity it affects reaches the field, which for a long-lead procurement item may require an answer far sooner than the contract requires.

Is a high RFI count a sign of a bad contractor?

Usually the opposite. High volume more often indicates incomplete or uncoordinated design documents, and a contractor asking questions is a contractor not guessing. The more diagnostic metrics are duplicate rate and completeness-return rate, which measure the contractor's own discipline, and RFI density against comparable projects, which measures the documents.

Who owns the cost of answering RFIs?

Design teams generally absorb RFI response within their fee, which is precisely why response capacity is finite and why excessive or poorly written RFIs degrade turnaround for everyone. Some contracts include provisions for charging back excessive or duplicative RFIs, though these are contentious and rarely enforced cleanly.