An RFP arrives Monday.
A defensible first draft exists by lunch.

ProjEx Propose ingests your Knowledge Base and the client's RFP, extracts every requirement, drafts a compliant outline, and generates the proposal — with a human checkpoint at every stage. Runs on your machine. Your data never leaves.

Request Beta Invite See how it works ↓
On-premise · zero cloud dependency Human-in-the-loop by design Auditable at every step

The Workflow

Seven steps. One afternoon.

The full pipeline at a glance — from raw documents to a compliant first-draft proposal, with a human decision at every step.

Proposal generation workflow: Load Documents → Knowledge Base + RFPs → RFP Analysis → Proposal Outline → Cover Letter, Compliance Report, Draft Proposal → Post-Generation AI-Assisted Editing
1

Ingest your knowledge base

Point ProjEx at your existing corpus — past proposals, personnel bios, case studies, methodologies, safety plans. Documents are chunked and embedded locally with a small on-device model.

Why it matters: nothing leaves your machine. No SaaS ingest, no third-party embedding endpoint, no data-processing agreement to negotiate.

2

Chat with your KB to curate it

Ask questions in plain English against your own KB before you ever look at an RFP. Confirm what's there, what's missing, what needs an update.

Why it matters: the quality of every proposal downstream is capped by the KB's quality. This step is where SMEs shape that ceiling.

◉ Human decision
3

Semantic + hybrid retrieval

Vector search over embeddings, blended with BM25 keyword scoring, gets the right chunks to the LLM at the right moment. Zero manual prompt-stuffing.

Why it matters: your existing evidence — bios, case studies, past section text — flows into every generated section without a human bookkeeper.

4

Read the RFP end-to-end

A four-pass extractor reads every page (text + tables + figures) and produces a structured Analysis grid: requirements, personnel constraints, schedule milestones, evaluation criteria, submission rules.

Why it matters: the "did we miss something?" question is settled at extract time, not the night before submission. Every requirement is tagged and searchable.

◉ Reviewable + editable
Analyst time: 4–6 hours of RFP read → 8 minutes.
5

Draft a compliance-bound outline

The Outline Creator proposes sections tied to specific requirements from the Analysis. A Python validator refuses any outline that leaves a mandatory requirement unbound.

Why it matters: the outline is auditable before a single sentence of body text gets written. Fix it in five minutes, not five hours.

◉ Approves outline
6

Generate the first draft

Per-section Writer runs against KB retrieval + Analysis context. In parallel: Cover Letter (iterative refinement), Draft Proposal body, and a per-requirement Compliance Report scoring the finished draft against the RFP.

Why it matters: a coordinated first pass — not a wall of AI text you have to reassemble. Compliance scoring flags gaps the same run.

Drafting time: ~40 hours of writer + reviewer → 20 minutes generated + 2 hours human polish.
7

SME polish, not blank-page

AI Assist lets your subject-matter experts refine any section with a targeted instruction: tighten the safety language, add a specific case reference, tone-match the client. Every edit is versioned.

Why it matters: your SMEs are reviewing and refining, not staring at an empty document at midnight. Every version is diff-able.

◉ SME final pass

The same RFP. Two different Mondays.

Without ProjEx

3–5 days
  • RFP read + extract 4–6 hrs
  • Outline + scoping 2–4 hrs
  • SME drafting (parallel) ~40 hrs
  • Compliance check 3–5 hrs
  • Fully-loaded analyst cost $8–15K

With ProjEx

1 afternoon
  • RFP read + extract 8 min
  • Outline + validation 3 min
  • Generation (Cover + Body + Compliance) 20 min
  • SME polish (targeted) 2–3 hrs
  • Cloud LLM cost per proposal ~$5–20

Designed for procurement to say yes.

Every design choice is optimized for regulated-industry buyers: defense, healthcare, legal, energy, and finance.

🔒

On-premise, air-gappable

Runs entirely on your Windows machine. Local LLM option for zero-cloud installs; cloud LLM (OpenAI, Gemini) is opt-in per-consumer.

📄

Auditable by design

Every generated section binds to specific requirements from the Analysis. Every claim in the proposal traces back to a KB chunk you approved.

👤

Human checkpoints, always

KB curation, Analysis review, Outline approval, SME polish. The AI drafts; your team decides. No autopilot mode.

⚙️

Calibrated per-vertical

Prompts, thresholds, and evaluation criteria are tuned for your industry — not a generic "AI writing tool" wrapper.

📊

Cost transparent

Before every run, see estimated token cost. Rolling calibration converges the estimate on actual OpenAI billing.

🛡

Data sovereignty guarantee

RFP text and KB content stay on your machine at ingest. Only the specific prompt going to a cloud LLM (if you enable one) leaves — never bulk data uploads.

See it work on your RFPs.

Beta program is open to proposal teams in regulated industries. Two months free, then $500/month for the following ten months.

Request Beta Invite