State of Proposal Tools — Wave 2 2026

The second annual State of Proposal Tools drop. Forty-five vendors. Four archetypes. Updated capability matrices across six dimensions. Grounded-AI taxonomy. Pricing-pattern analysis. Changes vs. Wave 1 in August 2025.

This is a long read and it stays long because category synthesis does not compress well without losing what matters. If you read Wave 1, the first four sections below summarize what changed; the back half of the post is the updated material.

What changed since Wave 1

Four structural shifts, each documented in the body of the report.

The vendor count grew from 37 to 45. Eight new vendors cleared the bar for inclusion — generally available, at least three public customer references, at least 10 reviews across G2 and Capterra, and a feature set that covers at least the basic RFP response workflow. Two vendors from Wave 1 were acquired and rolled into enterprise suites; they now appear in the “absorbed” cohort rather than as independent entries. The net ratio of challenger-to-incumbent shifted notably toward challengers, which tracks the Wave 1 prediction.

The AI-first challengers matured past the demo. Wave 1 noted several vendors with impressive demos and thin production references. Eight months later, the cohort has named customers, case studies, and enough review volume on G2 and Capterra to measure sentiment. The median AI-first challenger now has 40-80 reviews with sentiment broadly comparable to the incumbents’ first three years of review history. Whether the satisfaction holds at scale remains an open question — customer cohorts under 50 do not surface the content-rot problem that dominates the incumbent review base.

Pricing models fractured. A year ago, $50k-$150k ACV with a minimum-seat floor was the dominant incumbent pattern. Wave 2 data shows three distinct patterns: enterprise-seat (the legacy model), usage-based (tied to bids-per-year), and hybrid (seat floor plus overage). The pricing matrix in the full report maps each vendor to its primary pattern; 11 vendors have added an alternative pricing path since Wave 1 that did not previously exist.

The analyst framings disagree more than they did. Gartner’s most recent public MQ summary on proposal management and Forrester’s most recent Wave place several vendors in materially different tiers. Where the analyst houses disagree, the report names the disagreement and argues for one side or the other — with acknowledgment that this is opinion.

The four archetypes

Vendors are grouped by archetype rather than ranked. The grouping is intended to help buyers filter by shape, not to pick winners.

Incumbents (8 vendors). Loopio, Responsive (formerly RFPIO), QorusDocs, Upland Qvidian, AutogenAI in its enterprise tier, and three others. Characterized by larger install bases, mature compliance and workflow features, and known content-rot patterns in the review base. QorusDocs reviews and Qvidian reviews retain the UX complaints that dominated Wave 1 — modernization is underway at all eight but has not fully landed at any.

AI-first challengers (14 vendors). Post-2023 entrants whose core bet is retrieval-augmented drafting. Characterized by faster product iteration, thinner compliance features, smaller customer bases, and — for the ones that have clear architectural discipline — credibly lower hallucination rates than the incumbents’ AI features. The Stanford HAI research remains the baseline against which grounded claims should be read.

Vertical specialists (14 vendors). Federal-only, healthcare-only, security-questionnaire-only cohorts the Wave 1 report undercounted. DDQ and security-questionnaire specialists in particular have grown — the space described by Loopio’s DDQ writeup and Safe Security’s writeup on VRAQ practices has enough specialized tooling that it deserves its own cohort.

Absorbed (9 vendors). Vendors acquired, sunsetted, or folded into enterprise suites between Wave 1 and Wave 2. Included for completeness and for buyers evaluating legacy contracts.

The grounded-AI taxonomy

Every vendor now uses the phrase “grounded AI” or a near synonym. The taxonomy distinguishes four architectures:

T1 — retrieval-then-generation with per-claim verification. The generator is constrained to draft from retrieved blocks, and a second pass verifies each claim against its citation before the draft is surfaced. Five vendors in the sample meet this bar.

T2 — retrieval-then-generation without per-claim verification. The generator drafts from retrieved blocks but no claim-level verification gate is applied. Eleven vendors in the sample. This is the most common “grounded” architecture and the one where the Stanford HAI legal-RAG hallucination rates most directly apply.

T3 — generation with post-hoc citation attachment. The generator produces a draft, then a retriever attaches citations after the fact. Seven vendors. This architecture produces the highest hallucination rates in practitioner reports, because the citation does not constrain the generation.

T4 — AI feature is a chat interface with access to the corpus. The “grounded” claim is that the chat can cite sources; the drafting workflow does not use that citation in any structural way. Six vendors.

The remaining 16 vendors either do not claim grounding or have architectures that did not cleanly fit one category. The full taxonomy matrix, vendor-by-vendor, is in the appendix.

Pricing-pattern fractures

Three patterns, with notable shifts.

Pattern A — enterprise-seat. $50k-$150k ACV, minimum seat floor of 10-25, annual contract, annual price bumps capped in the MSA. Historically the incumbent default. Now held by 14 vendors in the sample, down from 18 in Wave 1. Three incumbents added overage pricing; one dropped the seat minimum for non-enterprise tier.

Pattern B — usage-based. Priced per-bid, per-response, or per-word of drafted output. Seven vendors. The usage-based cohort is disproportionately AI-first, and the pricing varies by an order of magnitude between vendors — from roughly $400 per bid at the low end to $3,000+ at the high. Usage-based pricing has the advantage of matching customer value; the disadvantage is budget unpredictability, which we heard named in four customer references.

Pattern C — hybrid. Seat floor plus overage on specific actions. 24 vendors. The dominant pattern for vendors new to pricing in the last 18 months. The overage variables differ — some overage on bids, some on users, some on draft minutes. Buyers evaluating hybrid vendors should model three usage scenarios and price each.

The big change since Wave 1: the proportion of vendors publishing prices on their websites dropped slightly. Sixteen vendors have some pricing public; 29 do not. Pricing opacity as a market signal is a topic we have written about before — the pattern persists and is getting slightly worse in the incumbent cohort.

Sentiment update

Aggregating G2, Capterra, and Gartner Peer Insights reviews from the past 12 months:

• The incumbent content-rot complaint is unchanged. Loopio reviews continue to name library maintenance as the dominant pain point, ~40% of Wave 2 sample.
• The search complaint on Responsive is unchanged. Keyword-match search surfaces “too many loosely related answers” in roughly 35% of Wave 2 reviews.
• The UX complaint on Qvidian has softened slightly — the most recent release cycle addressed the “looks dated” complaint directly. Feature depth complaints remain.
• Challenger sentiment is positive on drafting quality, mixed on workflow maturity, and weak on compliance features. The pattern matches Wave 1’s prediction that challengers would catch up on drafting first and workflow second.

Wave 1 predictions scored

Five predictions from Wave 1. Three held, two did not.

Held: incumbent content-rot complaints would remain the dominant review theme. Held: challengers would graduate from demo to production inside 12 months. Held: pricing would diverge. Did not hold: procurement-side tools would get broader attention — the cohort has grown but mainstream coverage has not. Did not hold: one incumbent would announce a grounded-AI architecture overhaul in 12 months — Wave 2 finds incremental changes across all eight incumbents but no full overhaul.

Vertical specialists, deeper

The vertical-specialist cohort grew from 8 vendors in Wave 1 to 14 in Wave 2. The growth is concentrated in two sub-areas, each of which deserves its own summary.

DDQ and security-questionnaire specialists. Six vendors focused specifically on the vendor-security DDQ and the commercial security-questionnaire workflow. The growth tracks practitioner sentiment documented in Safe Security’s writeup and Loopio’s DDQ post — 500+ questionnaires per year at larger organizations is a volume the generalist tools are not optimized for. Each of the six specialists offers a feature set that pairs questionnaire ingest, answer-library retrieval, and evidence-vault linking. Three of the six have meaningful published customer bases; three are still building out. We cover each in its own row of the appendix.

Federal-only specialists. Four vendors whose product is purpose-built for federal RFP response. Features center on compliance-matrix generation for federal solicitations (FAR/DFARS clause extraction in particular), past-performance citation (CPARS-aware), and integration with federal contracting portals. The cohort has not grown as fast as the DDQ sub-area, but the customer retention in this cohort is high — federal customers do not switch tools frequently once a response process is locked in.

The grounded-AI claim, vendor-by-vendor

A partial snapshot from the full taxonomy in the appendix. Five vendors in the T1 (retrieval-then-generation with per-claim verification) bucket: the full report names each with the measurement methodology they publish. Eleven vendors in T2; the verification gate in this cohort is either soft (a warning rather than a block) or absent. Seven vendors in T3 (post-hoc citation attachment); this cohort has the highest practitioner-reported hallucination rates and the least cohesive story about grounding. Six vendors in T4 (AI is a chat feature); the grounding story here is that the chat can cite its sources, not that the drafting workflow is structurally constrained.

Buyers evaluating the category should use the taxonomy as a first filter. A T4 vendor marketing “grounded AI” is making a different claim than a T1 vendor making the same phrase. The phrase is unchanged; the architecture behind it is not.

Analyst framings, disagreement map

Gartner’s Magic Quadrant summary on proposal management and Forrester’s Wave both covered the category in the past year. Where the two analyst houses disagree on a specific vendor, the disagreement usually centers on one of three dimensions: enterprise-readiness weighting, AI-feature depth, or customer-reference quality. The full report lays out a side-by-side per vendor where public analyst positioning is available. Our view, where we have one, is stated as opinion.

The net of the disagreements is that no single analyst framing should be treated as the category map. A buyer who shortlists vendors by reading only one analyst report will miss the challengers that vendor ranks low because those challengers do not fit the dimension that analyst weights. The corrective is to triangulate — analyst reports, review platforms, and direct vendor evaluations — against the buyer’s own workflow priorities.

Methodology notes

For readers who want to reproduce or audit the analysis, three methodology points.

Review-sentiment coding is done by a three-person research team, with a disagreement-resolution step on any review coded differently by two of the three. Inter-coder agreement is 91% on the recent sample. Reviews that cannot be confidently coded are tagged “unclear” and excluded from the aggregate rate; the exclusion rate is 6%.

Vendor-capability assessment uses a 14-dimension rubric — the rubric is published in the appendix. Each vendor is scored on each dimension from public materials, product demos where available, and customer references where we have permission to use them. Self-reported capability is treated as evidence but not as proof; where a vendor claims a capability we could not verify in demo or reference, the cell is tagged “claimed, not verified.”

Pricing data is from three sources: vendor websites, customer-permitted quote disclosures, and public RFP awards where the award price is documented. Where the three sources disagree, the report uses the most recent verified figure.

Where to go from here

Wave 2 is intended to make Wave 1 obsolete. Buyers evaluating the category should start here. Vendors in the category should read the appendix — where each vendor has a row with the dimensions on which we agreed with the vendor’s public positioning and the dimensions where we did not. Disagreements are fair game for pushback; the report updates quarterly in the form of shorter posts, and Wave 3 lands in April 2028.

Two companion pieces worth reading alongside: the grounded-AI pillar goes deep on the T1 architecture, and the category commentary on naming is the clearest argument for why “proposal intelligence” is the category name that fits what the tools are becoming.

Posts by The PursuitAgent research team are synthesis, not original reporting. Every cited number has a source in the post; uncited numbers are omitted. Views reflect PursuitAgent’s position.