The RFI, drafted for you.
The RFI Drafter is the Trueleveler engine that detects when a spec section disagrees with a drawing detail or another spec section, then drafts the complete RFI: structured question, cited evidence (verbatim quotes + drawing snippets), cost + schedule exposure, and routing. Every field is confidence-gated — anything below 70% blanks with a "?" prompt instead of guessing. The PM reviews, edits if needed, and sends.
Try it now
RFI Drafter runs on a spec section and a drawing reference and needs a free account (no card). It has no sample data yet, and the no-signup demo only levels bids.
Drafting is the easy part. Catching the conflict is the win.
An RFI exists because someone noticed a conflict. The drafter doesn't just save typing — it surfaces conflicts you'd otherwise miss until the day the field crew is standing in front of an impossible partition. Spec section 09 21 16 calls for an STC-50 demising assembly. Drawing A-301 shows a 5/8" one-layer-each-side configuration that only achieves STC-45. The conflict is real, the cost is real, and the day you find it determines which side covers it.
RFI Drafter reads the spec and drawings together. When it detects a conflict, it composes the full RFI: a 2–4-sentence structured question with verbatim quotes from both source documents, a proposed resolution cited to UL assemblies / ASTM standards / prior RFIs from this firm, cost exposure pulled from the awarded subcontract or RSMeans, schedule exposure computed against the critical path, and the routing list resolved from the contract's notice provisions.
The moment the conflict surfaces. Not the morning of the install.
Field PMs and project engineers run RFI Drafter the moment a conflict appears in any of the project's signals. The engine takes whatever input you have and fills the rest.
- Spec/drawing conflict detected by Doc Compare — the conflict-reconciler engine flags a disagreement; one click drafts the RFI with both quotes already wired in.
- Submittal flagged a deviation — the Submittal Reviewer found a sub proposing a substitution. Draft the RFI to confirm whether the architect accepts the deviation before procurement commits.
- Open question in meeting minutes — "Do we know whether spec or drawing governs the demising wall?" Voice Synthesis extracts the question; RFI Drafter writes the formal request.
- Field photo + voice note — site walk capture surfaces an ambiguity (e.g., a beam pocket that doesn't match the structural drawing). Photo, voice note, and the conflicting reference auto-attach.
- Manual entry — PM types the one-line question; drafter fills the four supporting fields.
Spec text. Drawing reference. The rest is project memory.
Two required inputs and a project context the engine reads automatically. No template forms.
Spec section
CSI-formatted spec section (PDF or pasted text). The engine identifies the controlling clauses and pulls the verbatim wording into the question.
Drawing reference
Drawing sheet (PDF) or page reference. The engine cites the specific sheet number and detail callout in the drafted question — an auto-cropped image snippet of the conflict region is on the roadmap, not in the product today.
Project context
Open RFIs (avoid duplicate filings), prior architect responses (predict the answer), the awarded subcontract for the affected trade (compute cost exposure), the master schedule (compute critical-path delta).
A complete RFI. Every field, confidence-tagged.
Sample drafted RFI from a spec/drawing STC conflict on a 14-floor commercial tenant fit-out:
Hallucinated routing or a fabricated spec citation is worse than a slow RFI. Every field carries a confidence score. Anything below 70% renders blank with a "?" prompt asking the PM to fill it in. The engine never invents a UL number, a cost figure, or a spec section that isn't in the source documents.
The draft flows. To the tracker, to the CO, to the schedule.
RFI Drafter is one step in the field-phase chain. The drafted RFI doesn't sit in a Word doc — it's filed in the tracker and feeds the project's risk model the moment it's open.
- RFI Tracker — the drafted RFI flows straight into the tracker for routing, aging, and response logging. The carries-data badge fires on this chain.
- Submittal Reviewer — submittal-flagged deviations are a top input. A substitution request becomes an RFI Drafter run that asks the architect to confirm acceptance.
- Schedule Risk — open RFIs age into the schedule risk model: an RFI that stays open past this project's median response time becomes a driver the next time Schedule Risk runs. The drafter's schedule-exposure figure is its own estimate; it does not change the forecast.
- CO Review — if the resolution lands as a scope add, the CO Review engine inherits the drafted question and proposed resolution as the change-order narrative.
Things everyone asks first.
Does the engine ever invent a spec citation?
No. Every quote in the drafted question is pulled verbatim from the uploaded spec section — we render the exact text with the section + paragraph reference, not a paraphrase. If the engine can't find a controlling clause, the question field stays blank with a prompt for the PM to clarify. Drawing details work the same way in text: the engine cites the sheet number and detail callout it drew from, not a paraphrase — a real cropped image snippet of the drawing region is on the roadmap, not in the product today.
How is cost exposure computed?
Two-tier. If the affected trade's subcontract is in the project, we use awarded unit pricing × affected quantity (a real number on this job). If the subcontract isn't loaded, we fall back to RSMeans or a confidence-scored estimate flagged as "needs PM input." Either way, the field is PM-editable — the PM is the cost authority, the engine is just suggesting a starting figure.
What if the spec and drawing don't actually conflict?
The engine won't draft an RFI if both documents agree. If a PM forces a draft on an ambiguous case, the question gets composed but a header note marks "possible non-conflict — review before sending" with the specific lines the engine compared. The drafter is biased to under-file rather than over-file.
Does it predict the architect's answer?
The Proposed Resolution field is the engine's best guess at what the architect is likely to approve, based on UL/ASTM standards, the contract's substitution clause, and prior RFIs on this firm (when the project history is loaded). Approval rate varies by project, trade, and architect, and Trueleveler doesn't track a single number across customers — so the PM still reviews it before it's sent, we just give them a starting position instead of a blank page.
What happens when the architect responds?
The response is logged in the RFI tracker. If approved, the resolution converts to a directive — if the change has cost or time impact, the CO Review engine inherits it as a draft change order. If denied or modified, the original RFI marks resolved and the PM decides whether to refile or escalate. The RFI's drafted cost + schedule exposure becomes baseline metadata for the next step.
Can subs draft RFIs the same way?
Yes — the engine runs the same flow for subcontractors. The subcontractor's bid + executed sub agreement carries forward as the routing context (sub routes to GC, GC routes to architect). The Pro plan tier covers both GC and sub roles.
Drop a spec/drawing conflict. See the drafted RFI in 60 seconds.
Bring one real conflict to a 15-minute call — a spec section and the drawing that disagrees with it. We'll draft the RFI on screen, walk through every field, and you keep the output.
Try Free for 14 Days →