Radar / Ideas / CounselStack: enterprise-grade…

CounselStack: enterprise-grade contract review for the mid-market at one-tenth the seat cost

daily ideamoderateJEV confidence 0.512026-10-02
Outcome10-200 lawyer firms clear contract backlogs with cited, escalation-gated AI review at roughly one-tenth of enterprise legal-AI seat pricing

The problem

Harvey AI runs an estimated $1,000-$2,000 per seat per month for mid-market firms on $50k-$300k+ annual contracts with 25-50 seat minimums - priced for Am Law firms, locking out the mid-market. Cheaper alternatives are either general-purpose chatbots without legal specialization or narrow tools without review workflows. Mid-market firms and in-house teams drown in NDAs, MSAs, and vendor contracts they cannot afford to have reviewed properly.

The idea

An open-weights contract-review stack: Ivo Sage (open legal specialist, 70% to 91% on the Legal Agent Benchmark Contracts) reviews agreements against the firm's playbook; an agentic extraction layer pulls every clause with bounding-box citations so attorneys can verify; a shared-space embedding index (index once with the quality tier, query with the cheap tier) retrieves the firm's own precedent clauses; and a calibrated decision gate implements Ivo's key insight - most review judgment happens before any edit - auto-accepting low-risk language, proposing redlines on deviations, and escalating true judgment calls to a named attorney. Every output ships with citations and an approver step.

Why now

Ivo just became the first legal-AI company to open-source its contract specialist, explicitly inviting others to fine-tune it on their own playbooks - the frontier-quality legal model is now free. LlamaIndex Extract v2.5 proved cited, confidence-scored extraction at falling cost, and Cohere Embed 5's shared embedding space makes per-matter clause retrieval cheap. The mid-market pricing umbrella ($1-2k/seat/month) has never been wider.

What it combines

Open legal specialist (ivo-sage) + cited agentic extraction (llamaindex-extract-v2-5) + shared-space clause retrieval (cohere-embed-5-pro-fast) + typed escalation gates (kev-open-trainable-jev-like-family-of-small-decision-models-on-qwen3-5-3-8). Ivo's own benchmark argues contract review is misgraded today: the hard part is judgment (prioritize, show restraint, escalate, follow the playbook), not edits. The stack operationalizes exactly that - extraction finds the language, retrieval finds the precedent, the open model applies the playbook, and the decision gate enforces the escalate-versus-accept judgment. No single component delivers a review workflow; together they do.

MVP

Weekend scope: upload an NDA or MSA, review against one playbook with Ivo Sage, output a redline report where every flagged clause carries a citation and a confidence score, and route low-confidence items to an attorney-approval queue. Deliberately skip: multiple playbooks, Word and Google Docs plugins, and firm-specific fine-tuning.

Distribution

B2B2C: white-label the review engine as an add-on inside practice-management platforms mid-market firms already live in (Clio, MyCase, PracticePanther), plus partnerships with legal-tech consultancies serving the 10-200 lawyer band. Per-matter pricing replaces per-seat. Who pays: mid-market law firms and in-house legal teams at mid-market companies.

Why it wins

Harvey, Legora, and CoCounsel sell quote-only enterprise seats with minimums; Ivo's own product is enterprise SaaS. Budget options ($29-99/month prompt tools) lack legal-specialized models and review workflows. CounselStack is the first stack combining an open frontier-class legal model with cited extraction and explicit escalation gates, sold per matter instead of per seat - and firms can self-host the whole thing, which matters for client confidentiality.

Risks

Biggest risk is attorney trust - lawyers will not act on AI output they cannot defend. The MVP de-risks by making the escalation gate and citations the product: nothing ships without a source location and a named human approver, mirroring how firms already work, and Ivo's micro1 Contract Bench gives an independent yardstick for the judgment behavior.

Build it with

Repo to start from

counselstack - open contract-review pipeline: Ivo Sage playbook review plus cited clause extraction plus calibrated escalation gates, per-matter pricing.

Evidence

Get the week's best AI launches, plus 3 ideas worth building

One email every Saturday. Ranked by traction, not hype. Free.