Summary

  • Casepoint introduced IQ on September 9 as a shared framework for several AI capabilities, with early access for selected customers.
  • Reusing platform context and controls may simplify expansion; it does not establish universal availability, accuracy or a cost-free path to adoption.

Buying another AI tool often means buying another set of connections, permissions and working habits. Casepoint is trying to change that purchasing decision. Its new IQ framework asks existing customers to expand inside a platform they already use for document review, information requests and related work, rather than assemble a separate AI application around it.

The September 9 launch announcement combines agentic AI, assistive AI, CaseAssist machine learning and predictive coding, and the Casepoint Model Context Protocol server. These are different kinds of capability brought into one framework, not four newly invented products. The announcement's closing qualification also matters: IQ is running in early access with selected customers and is available for demonstrations. A broad launch headline is not evidence that every agent is generally available.

The sale is continuity between tasks

Casepoint's proposition is that teams can extend their AI use while retaining a common setting for context, permissions and oversight. For a buyer, the potential gain is not simply fewer logins. It is less repeated work connecting a new tool to the information, roles and records that an existing workflow already uses.

The IQ product page gives those roles different jobs. IQ Agents are presented as moving from analysis toward decisions with reasoning and human approval. IQ Assist provides natural-language search, summarisation and classification. CaseAssist supplies established machine-learning and analytical functions, including relevance prioritisation and precision-and-recall reporting. A common framework does not make an assistant's output, a ranking and an agent's proposed decision equivalent evidence.

That distinction limits the economic inference. Shared controls could reduce some integration and administrative work, but the announcement supplies no measured implementation saving, new subscription value or improvement in review outcomes. A log can show what happened without establishing that the conclusion was right. Human approval is a place to exercise judgement, not a measured reduction in the effort that judgement requires.

Model choice stops short of platform independence

The external connection is not a September invention. Casepoint launched its MCP server on July 30. The initial functions described retrieval of operational information such as review-batch status, productivity metrics and information-request status. The company said requests use the authenticated user's existing permissions and that its function library would expand. That is narrower than permission for any external agent to perform any action.

The IQ page also distinguishes locally hosted models, described as running inside Casepoint's perimeter, from connections to an organisation's approved external AI ecosystem. The local-hosting statement cannot establish the data path of every external configuration. Neither route was tested for this report.

For customers, the useful question is therefore what can be added within the existing environment, with which supported functions and decision responsibilities. For Casepoint, the opportunity is to make successive AI purchases extensions of the same customer relationship. Early access provides a starting point for that proposition; it does not yet demonstrate how widely it can be repeated.