AlliCrew Get in touch →
A worked example · fictional data

A new RFQ comes in. What prior work is closest — and where would copying it go wrong?

Here's what you're looking at: on the left is a messy folder of old quotes and job records. The system read it and wrote the brief in the middle. Every line in the brief shows its work — click one to see the files behind it on the right.

A quantity-12 stainless coolant manifold RFQ lands at a fictional shop, Riverbend Manifold Works. The estimator knows there's relevant prior work in the archive. This is a worked example of an Engineering Knowledge System reading that messy archive and returning a source-backed similar- configuration brief — one where every claim traces back to a file.

Everything below is fictional but realistic. The folder is a real RFQ packet, the brief is the real generated output, and the evidence is the real source text — all built on invented data. The real version runs this on a representative folder of your own records.

Panel 1
The messy folder
The new RFQ packet, plus the prior-job archive it gets matched against.
The new RFQ — folder OMS-2026-0314
Harborline Marine Services · qty 12 · 316L coolant manifold · 21-day quote window
MD
cover_email.md
The written ask. Budgetary quote, 14-week first article target, identify risks before committing.
CSV
parts_list.csv
Five line items, 316L throughout. Note column already says "verify vendor and acceptance history."
MD
drawing_notes.md
1.25" tube, 2.5" bend radius. Reviewer flags: don't assume bend geometry is reusable.
MD
schedule_note.md
14 weeks is a target, not a commitment. Confidence: medium-low until checks close.
MD
verbal_note_log.md
Customer mentioned chloride pitting on a call. Verbal only — never entered the written RFQ.
Verbal note · not a written requirement
The prior-job archive — what the system searches
Three relevant prior jobs surface. None is a clean match.
RMW
RMW-2024-1182 — "Soundreach"
Best geometry & mounting match. But it's 304L, not 316L — and the quote and a later field note disagree on the pressure-test value.
Closest configuration Internal conflict: 1,200 vs 1,450 psi
RMW
RMW-2025-0617 — "Briarwater"
Best 316L material precedent. But fewer outlets, no isolation grommets. Quote summary is a stale draft superseded by a later inspection record.
Best material precedent Stale draft in archive
RMW
RMW-2023-2240 — "Northline"
Strong grommet-vendor and acceptance-test precedent. But it's 6061 aluminum — useless for 316L corrosion claims.
Grommet / acceptance precedent
STD
standing_docs/ — WPS-RMW-014, IP-MAN-03, VND-LORD-01
Welding procedure, manifold inspection plan, grommet vendor file. Reference material the brief draws on.
+
scanned legacy PDFs, duplicate records, near-duplicate draft chains
Several archive files are flagged scanned-legacy with medium OCR confidence and unreliable table columns.
This is the substrate, shown before any output. Three prior jobs are relevant and not one of them is a clean match — the closest configuration is the wrong material; the right material is the wrong configuration. Handling that honestly is the work.
Panel 2
The generated brief
The actual brief the system produced. Click any underlined line to see the files behind it.
Similar Configuration Brief — OMS-2026-0314
Built from 15 source documents · Same inputs produce the same brief, every time
What the system caught

Two files in the same job folder disagree on the pressure-test value — the RMW-2024-1182 quote says 1,200 psi; a later field note says 1,450 psi. The brief surfaces the conflict instead of picking one.uncertainty

The closest prior job by configuration, RMW-2024-1182, is the wrong material — it was 304L, this RFQ is 316L. The obvious thing to copy is the thing to be careful with.sourced

The ask

OMS-2026-0314 asks for quantity 12 vibration-isolated 316L stainless coolant manifold assemblies, with a hard-quote decision window of 21 days.sourced

The chloride-pitting note is verbal — it's risk information for engineering review, not a written RFQ change.uncertainty

Closest-prior-job retrieval

The three closest prior jobs, ranked: 1. RMW-2024-1182 (Soundreach), 2. RMW-2025-0617 (Briarwater), 3. RMW-2023-2240 (Northline).sourced

What does not carry forward
  • Material differs — OMS-2026-0314 is 316L; the closest-configuration job, RMW-2024-1182, was 304L.sourced
  • Tube bend geometry differs — 1.25" tube / 2.5" bend radius now, versus 1.5" tube / 3.0" radius on RMW-2024-1182.sourced
Uncertainty flags

A pressure-test conflict sits inside the archive — RMW-2024-1182's quote says 1,200 psi; its later field note says 1,450 psi.uncertainty

Two referenced documents are missing — the Soundreach fixture memo and the Briarwater pressure-test report are cited but not in the archive.uncertainty

The 14-week first-article timing is a schedule risk, not a committed delivery outcome.uncertainty

  • In-house bender capability needs confirmation.uncertainty
  • 316L raw-material lead time needs revalidation.uncertainty
  • The customer chloride-pitting note is verbal / low confidence.uncertainty
  • Fixture condition needs physical inspection.uncertainty
  • Schedule confidence is medium-low.uncertainty
Human-review checklist
  • Fixture and bend tooling need confirmation before quote release — confirm fixture condition, bender setup, and geometry fit.uncertainty
Panel 3
The evidence behind it
Every claim traces to its source documents. Click a claim to see them.
← Click any highlighted claim in the brief to see the source files behind it — the actual document text, and why it supports the claim.
"OMS-2026-0314 asks for quantity 12 vibration-isolated 316L stainless coolant manifold assemblies, with a hard-quote decision window of 21 days."
Source documents
OMS-2026-0314 / cover_email.mdRFQ ask
Riverbend Manifold Works received RFQ OMS-2026-0314 from Harborline Marine Services for a quantity-of-12 vibration-isolated 316L stainless coolant manifold assembly. The customer is asking Riverbend to provide a budgetary quote and to state whether a hard quote can be released within 21 days after engineering review.
OMS-2026-0314 / parts_list.csv316L & quantity evidence
line 1 — formed coolant manifold header, material 316L, quantity 12. line 2 — branch outlet tube set 1.25" OD, 316L, qty 144 (twelve outlets per assembly).
Why these sources: The cover email is the written ask; the parts list independently confirms 316L and the quantity of 12. Two documents agreeing is stronger than one asserting.
"The chloride-pitting note is verbal — it's risk information for engineering review, not a written RFQ change."
Source documents
OMS-2026-0314 / verbal_note_log.mdverbal chloride note
During a follow-up call, Harborline mentioned prior chloride pitting on a different coolant manifold. The note was verbal only and was recorded by Riverbend estimating as low-confidence context. The caller did not revise the written RFQ, did not add a chloride-soak test, and did not provide a separate corrosion specification.
OMS-2026-0314 / cover_email.mdwritten RFQ context
The written RFQ package requests a drawing review, a parts list check, and a first article target. The cover email does not list a chloride-soak test as a written requirement.
Why this is flagged uncertainty: The system holds the line between what was said on a call and what's in the written RFQ. It surfaces the chloride risk — but refuses to promote a verbal note into a customer specification. The audit explicitly tests that it does not overclaim here.
"The three closest prior jobs, ranked: 1. RMW-2024-1182 (Soundreach), 2. RMW-2025-0617 (Briarwater), 3. RMW-2023-2240 (Northline)."
Source documents
RMW-2024-1182 / quote_summary.mdrank 1
The work used a similar outlet count and the same general vibration-isolated mounting pattern later seen in OMS-2026-0314, which makes it the best overall configuration precedent.
RMW-2025-0617 / quote_summary.mdrank 2
The job remains the strongest Riverbend precedent for 316L material handling, chloride-resistance discussion, passivation, and final rinse documentation.
RMW-2023-2240 / quote_summary.mdrank 3
The job is the strongest precedent for isolation grommet selection, vendor file VND-LORD-01, and acceptance testing language.
Why this ranking: Each job earns its rank for a different reason — configuration, material, grommet precedent. The ranking is repeatable: the same archive and the same RFQ produce the same three jobs every time, in the same order.
"Material differs — OMS-2026-0314 is 316L; the closest-configuration job, RMW-2024-1182, was 304L."
Source documents
OMS-2026-0314 / parts_list.csv316L source
Every line item — header, outlet tube set, mounting foot bracket, passivation — specifies 316L.
RMW-2024-1182 / quote_summary.md304L source
The quoted material was 304L stainless, not 316L. The tube package used 1.5" outside diameter tube and a 3.0" centerline bend radius, so geometry does not carry forward cleanly.
Why this matters: RMW-2024-1182 is the best configuration match in the archive — same mounting pattern, similar outlet count. It would be the obvious thing to copy. The system flags that the obvious thing is the wrong material.
"Tube bend geometry differs — 1.25" tube / 2.5" bend radius now, versus 1.5" tube / 3.0" radius on RMW-2024-1182."
Source documents
OMS-2026-0314 / drawing_notes.mdquery bend geometry
The tube callout uses 1.25 inch outside diameter tube with a 2.5 inch centerline bend radius at the two tight return bends. The mounting pattern resembles prior work, but the bend geometry should not be assumed reusable.
RMW-2024-1182 / drawing_notes.mdprior bend geometry
The drawing used 1.5 inch outside diameter tube with a 3.0 inch centerline bend radius for the two return legs. The drawing reviewer warned that bend tooling and tube geometry should be checked before reuse.
Why this matters: Both drawing files independently warn against assuming the bend geometry transfers. The system isn't inferring a risk — it's surfacing one that both source documents already named.
"The 14-week first-article timing is a schedule risk, not a committed delivery outcome."
Source documents
OMS-2026-0314 / schedule_note.md14-week schedule target
Target is first article in 14 weeks from PO release. Riverbend estimating should describe this as a target with schedule risk, not as a guaranteed delivery commitment. Overall schedule confidence is medium-low until checks close.
RMW-2024-1182 / quote_summary.mdprior schedule risk
First article review and the 12-to-14-week schedule range were both marked as risk items tied to fixture readiness and stainless tube availability.
Why this is flagged uncertainty: The schedule note itself instructs that 14 weeks be described as a risk, not a commitment — and the closest prior job hit the same schedule-risk pattern. The audit explicitly tests that the brief does not claim 14 weeks is "guaranteed."
"A pressure-test conflict sits inside the archive — RMW-2024-1182's quote says 1,200 psi; its later field note says 1,450 psi."
Source documents
RMW-2024-1182 / quote_summary.md1,200 psi value
The quote says the released configuration pressure test was 1,200 psi, while a later field note should be checked before relying on that value.
RMW-2024-1182 / field_note.md1,450 psi value · scanned legacy PDF
Riverbend service noted that the same configuration was actually tested at 1,450 psi during the witnessed field-support package, which conflicts with the earlier quote summary value.
Why this is the demo's centerpiece: Two documents in the same job folder disagree on a number that matters. The system does not pick one and move on, and does not average them. It surfaces the conflict and routes it to a human. The field note is a scanned legacy PDF with medium OCR confidence — exactly the kind of file that gets skimmed and missed.
"Two referenced documents are missing — the Soundreach fixture memo and the Briarwater pressure-test report are cited but not in the archive."
Source documents
RMW-2024-1182 / quote_summary.mdmissing fixture reference
The quote also references fixture inspection memo F-1182-FIX-INS; that referenced fixture inspection memo is missing from this archive snapshot.
RMW-2025-0617 / test_summary.mdmissing test report reference
It references final pressure test report PT-0617; that referenced pressure test report is missing from this archive snapshot, so the summary should not be treated as a complete test package.
Why this is flagged uncertainty: The system tracks what the documents point to, not just what they contain. When a referenced file isn't in the archive, it becomes a review flag — the brief asks for the missing document rather than pretending the chain is complete.
"In-house bender capability needs confirmation."
Source document
OMS-2026-0314 / drawing_notes.mduncertainty flag evidence
Engineering should confirm in-house bender capability and fixture reach before quote release.
Why flagged: The drawing notes raise this directly. The brief carries the open question forward rather than assuming the shop can make the bend.
"316L raw-material lead time needs revalidation."
Source document
OMS-2026-0314 / schedule_note.mduncertainty flag evidence
Purchasing should revalidate 316L raw-material lead time before hard quote release. 316L raw-material lead time is listed first among the main schedule risks.
Why flagged: The schedule note names 316L lead time as a live risk and explicitly assigns the revalidation to purchasing. The brief preserves that assignment.
"The customer chloride-pitting note is verbal / low confidence."
Source document
OMS-2026-0314 / verbal_note_log.mduncertainty flag evidence
The note was verbal only and was recorded by Riverbend estimating as low-confidence context for materials and inspection planning. Riverbend should treat it as a risk flag for engineering review, not as a written customer requirement.
Why flagged: The same verbal note appears as both a risk worth surfacing and a claim worth bounding. The system does both — it raises the chloride concern and labels its confidence level honestly.
"Fixture condition needs physical inspection."
Source document
RMW-2024-1182 / fixture_or_tooling_note.mduncertainty flag evidence
Fixture F-1182-MAN was stored after final inspection with a tag requiring physical inspection before reuse. The old fixture was built around 1.5" tube and 304L assumptions; the new RFQ uses 1.25" 316L tube, so bend tooling, clamp inserts, and inspection datums need review.
Why flagged: The fixture note itself carries a physical-inspection-required tag. The brief doesn't assume the fixture is usable — it forwards the tag's instruction.
"Schedule confidence is medium-low."
Source document
OMS-2026-0314 / schedule_note.mduncertainty flag evidence
Overall schedule confidence is medium-low until these checks close — 316L lead time, in-house bending confirmation, fixture availability, passivation queue capacity, and inspection plan approval.
Why flagged: The brief uses the schedule note's own words. "Medium-low" is not the system hedging — it's the source document's stated confidence level, carried through unchanged.
"Fixture and bend tooling need confirmation before quote release — confirm fixture condition, bender setup, and geometry fit."
Source documents
OMS-2026-0314 / drawing_notes.mdquery fixture & bender review
Engineering should confirm in-house bender capability and fixture reach before quote release.
RMW-2024-1182 / fixture_or_tooling_note.mdprior fixture note
Do not release a hard quote until fixture condition and bender setup are confirmed. Bend tooling, clamp inserts, and final inspection datums need review.
Why this is the closing item: The human-review checklist is where the brief hands back to a person. Both the new drawing and the prior fixture note say the same thing — confirm the tooling before quoting. The brief makes that the explicit gate.
Every claim in this brief was checked before you saw it.

The run ships with two layers of checks. The first confirms every factual claim in the brief resolves to a real source document. The second goes further — it requires the brief to surface specific conflicts, and forbids it from overclaiming: it must not assert the chloride-soak test is an RFQ requirement, or that the 14-week schedule is guaranteed. Both layers passed, with zero failures.

14
claims, all sourced
0
unsourced / overclaims
2×
layers of checks, both pass
On your data

This one was fictional. Yours wouldn't be.

The RFQ, the archive, and the brief above were built from invented data. The real version runs this on one representative folder of your own quote and job records — and the brief comes back drafted against your work, with the same discipline: every claim sourced, every conflict surfaced, every uncertainty named rather than smoothed away.

Request a working sample →

One folder. A two-page brief back within a week. No charge for the first one.