Casepoint IQ
by Casepoint
Governed agentic AI for eDiscovery, FOIA and investigations, on FedRAMP High and DoD IL6 infrastructure
Casepoint IQ is the AI framework Casepoint launched across its legal and government platform on 9 September 2026. It unifies semi-autonomous review agents, embedded generative assistants, a decade-old predictive-coding engine and an MCP server under one governance layer, so eDiscovery, legal hold, FOIA and investigations teams get defensible, auditable AI rather than a shelf of point tools.
Casepoint IQ is the intelligence framework Casepoint launched on 9 September 2026 across its unified platform, consolidating what had been separate AI capabilities into one governed layer spanning eDiscovery, legal hold, FOIA, investigations, regulatory and compliance workflows. It has four components: IQ Agents, semi-autonomous agents that move work from analysis to determination; IQ Assist, more than a dozen embedded generative-AI tools for natural-language search, summarisation and classification; CaseAssist, the decade-old machine-learning and predictive-coding engine for relevance prioritisation and privilege identification; and a Model Context Protocol server that lets customers connect approved external AI models with built-in authentication and audit logging. The first two agents — relevance determination and issue coding — shipped in alpha to selected customers, with a beta expected before the end of 2026 and eight further agents in development covering privilege, PII detection and document redaction. The relevance agent follows the workflow review teams already use: it ingests the complaint, production requests and review protocol to build an editable case context of key players, terminology and relevance definitions, samples 50 documents and predicts relevance with written rationales for human feedback, and will not proceed to the full population until it clears thresholds of 90% recall and 80% precision, closing with random-sample validation that reports recall, precision, elusion and F1 scores with confidence intervals. In a vendor demonstration, 10,000 documents took roughly an hour and were scored at 190-plus review hours saved. Every action is permissions-aware, logged and traceable to source; agents cannot independently create, redact or produce documents; and because the framework sits inside Casepoint's existing infrastructure it inherits FedRAMP Moderate and High, ISO 27001 and DoD Impact Level 5 and 6 authorisations, so new agentic workflows do not trigger a fresh vendor security review.
The general counsel, legal operations lead or government agency CIO already running high-volume discovery under FedRAMP or DoD authorisation, who needs AI to absorb first-pass review without creating a defensibility problem in front of a regulator or a court.
Agentic first-pass review that clears stated recall and precision thresholds before it touches the full document set, with every decision logged and traceable — and no new security authorisation to obtain, because it runs inside the platform already accredited.
At a Glance
- Category
- Industry & Government
- Pricing
- Contact for pricing, Subscription
- Target Market
- General Counsel, Legal Operations Leaders, Government Agency CIOs, Compliance Officers, Litigation Support Managers
- Deployment
- Cloud-only
- Founded
- 2008
- Team Size
- 500+
- Customers
- 200+ corporate and enterprise customers; public-sector users include the SEC, Department of Defense, Veterans Affairs, USDA and the U.S. Courts, plus Marriott International, Mayo Clinic and Duke University
Key Features
- ✓IQ Agents with tiered autonomy
Semi-autonomous agents propose actions for human review at the initial level, with higher autonomy tiers planned as the product matures.
- ✓Threshold-gated review
The relevance agent must clear 90% recall and 80% precision on a sampled set before it is allowed to process the full population.
- ✓Statistical validation reporting
Final random-sample QC produces recall, precision, elusion and F1 scores with confidence intervals suitable for defending the process.
- ✓IQ Assist
More than a dozen embedded generative-AI tools across the platform for natural-language search, summarisation and classification of documents.
- ✓CaseAssist predictive coding
A decade-old active-learning engine for relevance prioritisation and privilege identification, now governed under the same framework as the agents.
- ✓MCP server for customer-choice AI
Connects approved external AI models to the platform with built-in authentication and audit logging, so model choice is not locked to the vendor.
- ✓Inherited government accreditation
Runs inside infrastructure authorised at FedRAMP High, ISO 27001 and DoD Impact Level 5 and 6, avoiding a new vendor risk assessment.
Capabilities
Use Cases
- •First-pass relevance determination
An agent reviews a full document population after clearing sampled recall and precision thresholds, reporting hours and cost saved against manual review.
- •Issue coding at scale
A second agent applies issue tags consistently across a matter, removing the reviewer-to-reviewer variance that usually drives re-review.
- •FOIA and public records processing
Government teams accelerate records request handling inside accredited infrastructure with every AI action logged and traceable to source.
- •Internal investigations
Investigators use embedded generative search and summarisation to find and understand relevant material across large collections quickly.
- •Bring-your-own-model deployments
Connect an organisation's own approved LLM through the MCP server so legal AI runs on a model the security team has already cleared.
Ideal For
Best For
- ✓First-pass relevance review on large document populations where recall and precision must be reported defensibly
- ✓Government agencies needing agentic AI inside FedRAMP High and DoD IL5/IL6 accredited infrastructure without a new authorisation cycle
- ✓Issue coding across a matter where consistency between reviewers is the usual quality problem
- ✓Regulated investigations and FOIA workloads requiring permissions-aware, source-traceable audit trails for every AI action
- ✓Legal teams that want to bring their own approved external model through an MCP server rather than accept the vendor's default
Not Ideal For
- ✗Small practices and simple commercial matters — independent analysis rates Casepoint's fit poorly here, and the platform expects standardised workflows and implementation effort
- ✗Buyers who need transparent self-serve pricing: there is no public rate card, pricing transparency is the vendor's weakest independently scored dimension, and IQ's own credit model is not finalised until beta
- ✗Teams wanting fully autonomous production today — the shipped agents are alpha, run at an autonomy level where a human approves proposed actions, and cannot create, redact or produce documents on their own
- ✗Organisations looking for legal billing, time tracking or portfolio-level legal operations reporting, which independent scoring flags as weak or absent
Deployment
Market & Ratings
200+ corporate and enterprise customers; public-sector users include the SEC, Department of Defense, Veterans Affairs, USDA and the U.S. Courts, plus Marriott International, Mayo Clinic and Duke University
Market Analysis
Pros
- ✓Agentic AI arrives inside already-accredited FedRAMP High and DoD IL5/IL6 infrastructure, so no fresh vendor risk assessment is needed
- ✓Quality gating is explicit and measurable — the agent does not touch the full population until it clears 90% recall and 80% precision on a sample
- ✓Human-in-the-loop is structural, not advisory: agents cannot independently create, redact or produce documents
- ✓Consolidates agents, assistive AI, predictive coding and external-model access under one permissions-aware, fully logged governance layer
- ✓Independent reviewers consistently praise review speed, search, analytics and security fit, and the all-inclusive model avoids analytics surcharges
Cons
- ✗The two launch agents are alpha with selected customers only; beta is promised before year end, so this is a roadmap purchase today rather than a deployed capability
- ✗Pricing is opaque — no public rates, and independent RFP scoring puts pricing transparency at 2.5 out of 5, the platform's weakest dimension
- ✗The IQ credit model is not finalised until beta, so total cost of an agentic review cannot be forecast yet
- ✗Reviewers report slow uploads and production preparation on large matters, and note advanced configuration depends on vendor assistance
- ✗Migration, training and production support on large matters are repeatedly cited as major cost drivers beyond the licence
- ✗Portfolio-level legal operations reporting is weaker than dedicated suites, and it is not a billing or time-tracking system
Pricing
Casepoint IQ (platform framework)
Contact for pricing
- ✓Live and generally available to existing Casepoint customers
- ✓IQ Assist, CaseAssist and MCP server included in the platform
- ✓No published rate card; quote-based all-inclusive model
IQ Agents (alpha, beta due 2026)
Contact for pricing
- ✓Alpha exposes raw token consumption in the interface
- ✓Beta converts tokens into credits
- ✓Planned: unmetered access for the largest customers, allocated credits plus overage for smaller subscriptions
Casepoint does not publish rates, and independent RFP analysis scores its pricing transparency at 2.5 out of 5 — every buyer goes through a custom quote, which makes budgeting hard. The offsetting point reviewers make is that the model is all-inclusive, so using analytics does not trigger escalating add-on charges. For IQ specifically the commercial model is still forming: the alpha shows raw token consumption on screen, the beta will convert tokens to credits, and the stated plan is unmetered access for the largest customers with allocated credits plus overage for smaller subscriptions. Migration, training and production support on large matters are cited as the real cost drivers.
Security & Compliance
Sources
This page was written from 5 sources, 3 on domains other than casepoint.com.
Stay Ahead of the Curve
Weekly enterprise AI insights for technology leaders. No spam, no vendor pitches—unsubscribe anytime.
SubscribeRelated Products
Sixfold
An AI underwriter that learns your book and takes submissions through to bind-ready
Aderant Agent Center
Purpose-built AI agents for law firm back-office operations, on the Stridyn platform
DISCO Advanced Research
Agentic deep research for litigation — autonomous multi-step fact investigation with an auditable reasoning trail
Dili
AI-powered prevailing wage and certified payroll compliance for federally funded infrastructure projects