Blog · Page 28
Field notes.
Page 28 of 31. Browse the archive of RFP workflows, grounded-AI architecture, and proposal operations notes.
Hybrid search: dense embeddings plus BM25 for proposals
Pure dense retrieval misses on numeric identifiers, product names, and SOC codes. Pure BM25 misses on paraphrase. The blend ratio we use, how we tune it, and the test set that catches regressions.
The Loopio teardown: what 1,700 customers are actually paying for
Loopio is the category's reference customer. Reviews, pricing signals, and product surface area, taken seriously. What it does well, what it doesn't, and what 1,700 customers are buying when they renew.
DDQ Anatomy, Part 1 of 4: the finance section
What the finance section of a DDQ asks — SOC audit history, revenue recognition, debt covenants, parent-company financials. What evaluators want, where questions repeat, and what good answers look like.
The Responsive pricing trail: what three years of leaks reveal
Responsive (formerly RFPIO) lists no prices. Three years of public data — job postings, G2 reviews, customer signals, and indirect references — let us describe the shape without inventing the numbers.
In preview: DDQ question classification
Every question in an ingested DDQ is classified at intake into finance, legal/privacy, security, or operations buckets. In preview behind a feature flag — DDQ is a pursuit-type the marketed platform does not yet describe.
Shipped: auto-generated compliance matrix from ingested RFPs
Stage 4 of the RFP pipeline — the compliance matrix — is now one click away from intake. Drop a PDF in, get a matrix out. Here's what shipped and where it still needs a human.
Grounded Retrieval 101, Part 4: what we're still wrong about
The closing post of the Grounded Retrieval 101 series. Three failure modes we have not solved — numeric precision, compound claims, synonym drift — with the test cases that surface them and what we are doing about each.
Win themes and the swap-name test
If you can swap your company name with a competitor's and the win theme still makes sense, the win theme is fluff. Six examples of themes that fail the test, six that pass, and what evaluators actually score.
How the citation rendering stack works
From a retrieval hit to a verify button next to a sentence, in four components. The plumbing behind every cited claim PursuitAgent ships, and why we render the source inline instead of in a footnote.
RFP software is a vocabulary problem
The terms vendors use — content library, AI suggestion, workflow automation — are doing too much work. Rename them and the failure modes get obvious. An opinion piece on why the category's marketing language is the bug.
Reviews watch: what G2 and Capterra said this week
Five reviews from G2 and Capterra worth reading if you're shopping the proposal-software category. Loopio, Responsive, QorusDocs, Upland Qvidian — the patterns that recur.
Shipped: content-block freshness scores
Every KB block now carries a freshness score that decays as the source ages, drifts from the company's current marketing language, or contradicts a more recent block. Stale citations get caught at draft time.
Prefer to see the product?
Take the 5-minute tour, or start a trial workspace and see PursuitAgent draft answers with citations.