Every RFP, tracked from intake to submission.
One dashboard for every opportunity — status, client, site, due date. Drill in for RFP intake, analysis, proposal generation, and the audit-ready compliance matrix.
ProjEx Propose reads the client's RFP, extracts every requirement, drafts a compliance-bound outline, and generates a proposal grounded in your team's real track record — with a human checkpoint at every stage. Runs on your machine. Your data never sees the cloud.
One dashboard for every opportunity — status, client, site, due date. Drill in for RFP intake, analysis, proposal generation, and the audit-ready compliance matrix.
Every architectural choice is defensible to legal, IT, and evaluation committees. Not a general-purpose AI wrapper.
The Systematic Reader architecture reads every chunk of every RFP — not the top-N most semantically similar. Commodity retrieval-based RAG caps at ~80% recall on structured extraction; a 100+ deliverable EPC RFP loses 30 in the gap.
~95% extraction recallHybrid retrieval fuses dense vector search + BM25 via reciprocal rank fusion. Map-Reduce synthesis reads full documents in disciplined windows. Per-category multi-label extraction dispatches one small-schema call per category — 60× per-call headroom. Iterative self-critique rewrites high-stakes prose before you see it. Universal self-heal absorbs context overflow, truncation, and network drops — every consumer × every backend.
Five techniques, one integrated pipelineEvery output is quality-gated — confidence scoring, hallucination detection, self-review. Every finding is source-attributable. The Compliance Traceability Matrix auto-maps every RFP requirement to the proposal section that addresses it, with evidence quotes.
Kills a 4-12 hr manual auditSingle-resident-model architecture with an activation-aware VRAM solver that refuses doomed loads before the inference engine starts. Self-healing pipeline auto-recovers from context overflow, truncation, and retrieval imbalance. No Windows GPU freezes.
Runs on one 32 GB GPUYour past-project details, personnel qualifications, proposal templates, and methodologies stay inside your infrastructure. No training on your data. No telemetry captures RFP or KB content. No vendor employee ever reads your proposals. On the LOCAL tier, air-gap the machine — the vendor has zero visibility into what you generate.
Zero training-data leakageVendor calibrates per-vertical on public data, ships an encrypted profile as the license key — one installer serves LOCAL or CLOUD tiers, customers upgrade by swapping keys, no reinstall. Every pipeline change is scored against gold datasets before shipping.
Same installer, LOCAL or CLOUDEvery design choice is optimized for regulated-industry buyers: defense, healthcare, legal, energy, and finance.
Runs entirely on your Windows machine. Local LLM option for zero-cloud installs; cloud LLM is opt-in per-consumer.
Every generated section binds to specific requirements from the Analysis. Every claim traces back to a KB chunk you approved.
KB curation, Analysis review, Outline approval, SME polish. The AI drafts; your team decides. No autopilot mode.
Prompts, thresholds, and evaluation criteria are tuned for your industry — not a generic "AI writing tool" wrapper.
Before every run, see estimated token cost. Rolling calibration converges the estimate on actual OpenAI billing.
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 uploads.
Beta program is open to proposal teams in regulated industries. Two months free, then $500/month for the following ten months.
Request Beta Invite