Blog · Page 25
Field notes.
Page 25 of 31. Browse the archive of RFP workflows, grounded-AI architecture, and proposal operations notes.
The grounded drafting loop, step by step
Retrieve, draft under constraint, verify, emit — or refuse. The four-step loop that produces every drafted answer in PursuitAgent, and the failure mode each step exists to prevent.
Reading the RFP the procurement lead actually wrote
RFPs are procurement documents written by named humans with known constraints, not sales documents. Read them that way and you respond 40 to 60 percent better. A preview of next week's pillar piece.
The chunk size ablation: 256, 512, 1024 tokens on RFP text
We ran the same retrieval pipeline at three chunk sizes against our RFP-text gold set. Directional results, the tradeoffs that surfaced, and why we don't ship a single global chunk size.
Stop announcing features, announce what changed for the reader
An announcement that names a feature is a press release. An announcement that names what changed in the reader's day is a useful one. A short field note on what we'll publish under 'shipped' from now on.
Anatomy of a mid-sized municipal RFP: Chicago-shaped
A walk through the structure of a typical mid-sized municipal RFP — front matter, technical scope, evaluation, contractual riders — using a representative Chicago-shaped solicitation as a composite worked example.
Shipped: answer-block diff view for reviewers
When a drafted answer is regenerated against a different KB block, reviewers now see a side-by-side of the previous version and the current version. Shipped this week.
A bid/no-bid scoring rubric we actually use
Five dimensions, a 1–5 score on each, a written floor, and a no-bid decision that is as cheap to defend as a bid decision. The rubric I bring into every kickoff.
The Responsive teardown: what 'enterprise-grade' means
A feature-by-feature look at Responsive (formerly RFPIO) — content library, AI, workflow, reporting — using public review sites as the primary signal. Where they win, where they don't.
Win themes when you're the incumbent defending a contract
Defending a contract is a different proposal than winning one. The win themes that worked four years ago will lose you the renewal. Here's what changes when you're already inside.
Our eval harness, on the command line
A walkthrough of the dev loop for retrieval changes — one command to baseline, one command to re-run, one to diff. The CLI ergonomics that keep us from tuning by feel.
Content library vs. knowledge base is not semantics
The vendors call it a content library. We call it a knowledge base. The two words name two different products. Why I think the distinction is the most important one in this category.
Federal RFP word counts, 2024 to 2025
What two years of public federal RFPs on SAM.gov tell us about response-document length, page-count caps, and the directional drift of complexity. Research note with sample-size caveats.
Prefer to see the product?
Take the 5-minute tour, or start a trial workspace and see PursuitAgent draft answers with citations.