--- title: PM Workflow Pattern — LFCS Bid Prep + Project Setup created: 2026-05-03 created_by: cowork-hq (proof-of-concept built on 2602 Munmorah Bridge bid) status: draft-v1 purpose: Replicable AI-assisted workflow for setting up a new LFCS project from incoming tender documents through to bid submission. Pattern derived from 2602 Munmorah Bridge prep on 2026-05-03. Refine after running through 2-3 jobs. related: - "[[lfcs-standards/folder-spec]]" - "[[2602 - Dec Projects - Munmorah Bridge/00-brief]]" audience: LFCS PMs (Rocky + the other guy in the office, plus future hires) --- # PM Workflow Pattern — LFCS Bid Prep + Project Setup > First proof-of-concept on the **2602 Munmorah Bridge** bid (Sun 2026-05-03 evening). Pattern documented here so any LFCS PM can replicate it on any incoming tender. Refine after we've run 2-3 more jobs through it. --- ## Why this exists Robert's idea (per 2026-05-03 conversation): for each project, get AI to read every drawing, extract structured summaries, and put them in a queryable form (Airtable + Obsidian). So that: - Pricing decisions are faster (you can grep "Renderoc" across all jobs to see how it was priced before) - Inclusions/exclusions are tighter (every item references the source drawing) - Onboarding new PMs is faster (the structure is the same on every job) - The QS call goes smoother (you walk in with a structured question list, not just a "we'll see what comes up") The 2602 Munmorah job was the first run. This doc captures what worked. --- ## When you trigger this workflow - New tender documents arrive (drawings + BOQ + scope brief, typically PDF + xlsx) - Bid window is short (1-2 weeks) - You want to bid competitively without missing scope or eating uncosted risk - You want a defensible inclusions/exclusions list when the head contractor pushes back If a job is awarded (no bid required), skip to **Stage B** (project intake). --- ## Stage A — Bid prep workflow (1-2 evenings, AI-assisted) ### Step 1 — Receive + register the tender pack - Tender docs arrive via email - File them under a temporary location (Downloads / desktop) — you'll move them into the vault in step 3 ### Step 2 — Open the LFCS Cowork project + tell it the job ``` "Working on a new bid — [Job number] [Client] [Site]. Got drawings and BOQ. Want to do bid prep." ``` Cowork (cowork-lfcs) will ask: scaffold the folder? answer yes. ### Step 3 — Scaffold the job folder Cowork creates the full 11-section folder structure under: ``` 11_OpsMan_LFCS/Operations/Jobs/Upcoming/[NNNN] - [Client] - [Site]/ ``` per the canonical `lfcs-standards/folder-spec.md`. You drop the source files into: - Drawings → `03 - Plans and Specifications/03a - Issued for Construction/` - BOQ (issued) → `02 - Scope and Pricing/02b - BOQ and Rate Schedules/` - Working copy of BOQ (your priced version) → `02 - Scope and Pricing/02a - Quotes and Estimates/` ### Step 4 — Extract drawing data (the queryable bit) Cowork runs text extraction on every drawing PDF. For each drawing, it builds a row in `Drawing-Index.md` capturing: - Drawing number, title, discipline, sheet size, rev, date, drafter - Geometry / dimensions extracted from text labels - Materials and specifications referenced - Construction notes (verbatim, important) - Existing services / risk items flagged - LFCS-scope tag (Yes / Reference / NO) Plus cross-cutting tables: - All materials referenced (with LFCS scope flag) - All dimensions captured - All risks identified - All standards referenced **Caveat to remember:** text extraction misses geometric content that's drawn but not text-labelled. You MUST do a vision-pass on each drawing yourself before final pricing. The AI extract is a head-start, not a substitute. ### Step 5 — BOQ parse Cowork parses the BOQ xlsx and writes a quantities-at-a-glance table into `00-brief.md` (concrete, reo, formwork, strip totals). All quantities, no rates yet. ### Step 6 — Bid prep doc Cowork writes `00-Bid-Prep.md` with: - Scope summary (1-2 paras) - Tender document register - Bridge / structure specs (geometry, codes, loading) - Quantities (by element, with totals) - Risks identified at quote stage - Cost framework skeleton (pointing to `lfcs-standards/rate-card.md`) - Variation rate template - Exclusions headline - Submission strategy ### Step 7 — Pre-call questions doc Cowork writes `06 - Correspondence/06c - Meeting Minutes/Pre-Call-Questions-[date].md` with the QS call agenda — every must-have-clarified item with answer slots. You take the call with this open and type answers in-line. Categories of questions: - **Scope clarifications** (supply boundaries, cranage, specs, drawing currency, programme) - **Commercial** (variation process, retentions, LDs, defects period, insurance, submission format) - **Site / risk** (existing services, site access, geotech, latent conditions) - **Nice-to-have** (alternatives, pour planning, subbie approvals) ### Step 8 — Inclusions/Exclusions draft Cowork writes `02 - Scope and Pricing/02a - Quotes and Estimates/Inclusions-Exclusions.md` with: - Inclusions (what LFCS will price + perform), grouped by BOQ section - Exclusions (cranage, supply items, other trades, design, PM obligations, latent, commercial) - Qualifications and assumptions (drawing basis, specs assumed, programme, services, rate validity, insurance) - Variation rate schedule (placeholder rates to fill) - Submission checklist Items pending Tuesday-call confirmation are flagged in italics for easy spotting. ### Step 9 — You read the drawings yourself Non-negotiable. AI extraction is good for structured text, weak on drawn geometry. You eyeball every page. Annotate in the drawings PDF or in `Drawing-Index.md` if anything's wrong. ### Step 10 — QS call (Tuesday evening typically) Walk through `Pre-Call-Questions-[date].md`. Type answers in-line. After the call, propagate confirmed answers into `Inclusions-Exclusions.md` and the pricing working sheet. ### Step 11 — Pricing You (or cowork-lfcs with your direction) fill rates into the BOQ working copy. Cross-check with `lfcs-standards/rate-card.md`. Cowork can sanity-check totals against the quantities-at-a-glance in `00-brief.md`. ### Step 12 — Submission package - Cover letter - Priced BOQ (PDF) - Inclusions / Exclusions / Qualifications (signed) - Variation rate schedule - Programme assumption (rough sequence + duration) - Reference projects + insurance certificates Cowork can draft the cover letter. You sign and send. ### Step 13 — Log it Append a one-liner to `11_OpsMan_LFCS/_Brain/_Log.md` with the bid date and outcome status (`submitted`, `awarded`, `lost`). --- ## Stage B — Project intake (after award) If awarded: 1. Move the folder from `Jobs/Upcoming/` to `Jobs/Ongoing Jobs/` 2. Update `00-brief.md` frontmatter (status: ongoing, contract_price: $XXX, start_date: YYYY-MM-DD, target_pc: YYYY-MM-DD) 3. Receive the scaled-CAD drawings (Northrop's promise post-award) — drop new revs into `03a` with rev suffix 4. Run a re-rate pass — compare scaled CAD scope against tender-day scope, lodge VOs for any deltas using the pre-agreed rate schedule 5. Stand up SWMS, JSAs, inductions per `05 - Safety and Compliance/` 6. Open the Job Tracker xlsx (copy `lfcs-standards/Job-Tracker-Template.xlsx`) into the project root If lost: 1. Move folder to `Jobs/Completed Jobs/[Job number] - [Client] - [Site] (LOST)/` or archive it under `_Archive/Lost-Bids/` 2. Append outcome to `_Log.md` --- ## Stage C — Scaling to the office (Robert's full vision) Once 5+ jobs have been run through Stage A, evaluate: - Is `Drawing-Index.md` actually getting referenced when pricing new jobs? - Are PMs grepping across jobs for "how did we price X last time"? - Are inclusions/exclusions getting tighter? If yes → build the Airtable layer: - One base, one row per drawing per job - Frontmatter from `Drawing-Index.md` rows feeds it (manual import or scripted) - Lookups: "every job using AS p/n 040910R bearings", "average $/m³ priced for 50 MPa deck concrete" If no → drop the Airtable idea. Markdown + Obsidian search is enough. Don't over-engineer. **Decision point: 5 jobs in. Date estimate: ~2026-08 if bid cadence is 1-2 per month.** --- ## Files this pattern produces (per job) ``` Jobs/Upcoming/[Job number] - [Client] - [Site]/ ├── README.md ← folder map + status ├── 00-brief.md ← frontmatter, scope, quantities, risks, dates ├── 00-Bid-Prep.md ← full bid analysis ├── 01 - Contracts and POs/ (after award) ├── 02 - Scope and Pricing/ │ ├── 02a - Quotes and Estimates/ │ │ ├── Inclusions-Exclusions.md ← bid submission inclusions/exclusions │ │ └── Pricing-Working-[date].xlsx ← your priced BOQ working copy │ └── 02b - BOQ and Rate Schedules/ │ └── [original BOQ from QS] ├── 03 - Plans and Specifications/ │ └── 03a - Issued for Construction/ │ ├── [original drawings PDFs from designer] │ └── Drawing-Index.md ← AI-extracted queryable per-drawing summary ├── 04 - Programme and Scheduling/ (post-award) ├── 05 - Safety and Compliance/ (post-award) ├── 06 - Correspondence/ │ ├── 06a - Emails/ (export important emails as PDF) │ ├── 06b - RFIs/_REGISTER.md │ └── 06c - Meeting Minutes/ │ └── Pre-Call-Questions-[date].md ← QS call agenda + answer slots ├── 07-11 ... (post-award per folder-spec) ``` --- ## Lane reminders - **cowork-lfcs** owns this workflow end-to-end once the LFCS PM Cowork project exists. No code work, no on-site/voice work (that's Hermes via Telegram). - **cowork-hq** stayed out of this lane except for the one-time scaffold pass on 2026-05-03 evening (bridge bid was a fire). - **Hermes** handles on-site dictation, photo filing, EOD summaries, RFI status queries, docket scanning. Different surface, complementary. --- ## Open improvements (queue for v2) - [ ] Vision-pass on drawings (PDF → images → Claude vision) — for geometric / dimensional content text extraction misses. Need pdftoppm + Read tool to play nicely. - [ ] BOQ-vs-drawings cross-check — script that flags BOQ items with no drawing reference + drawing items with no BOQ line - [ ] Pricing-history lookup — a `Pricing-History.md` at LFCS Operations level pulling rates from past priced BOQs - [ ] Subbie performance log — supplier ratings tied to past jobs - [ ] Standardised cover letter template - [ ] Auto-generation of submission email body from the inclusions/exclusions doc