Lawyers are using AI assistants. What could go wrong?

AI tools are changing how lawyers work, and not entirely in the ways knowledge managers planned for. What's emerging at firms across the US and UK isn't as much a failure to adopt AI as what we might call unstructured, or ungoverned adoption.

AI assistants offer lawyers a way to get insights and answers to clients faster. To leverage this new capability, lawyers pull documents out of the governed iManage environment to ground the AI responses, leading to an often unintended and definitely undesirable outcome: now the firm has sensitive client and proprietorial data outside the iManage platform. Unprotected, ungoverned, and at risk of becoming out of date.

A further concern is that the product, or output of the assistant, is not captured and that any modifications made to the original data are lost, eroding the status of the iManage environment as the sole “source of truth.”

This makes us ask, what would it take to modify this new behavior?

Institutional knowledge and “Knowledge Collections”: the governed layer for AI

The institutional knowledge library — the curated best practice documents, precedents, and templates — is exactly what AI tools should be reasoning from. The issue is that getting knowledge into those tools, securely and consistently, hasn't had a clean answer. Until now.

Knowledge Collections is a new feature within iManage Insight+ that lets knowledge professionals build sets of documents that can serve as the source layer for any AI tool the firm uses.

Collections can be practice area-specific or cross-functional; they can guide processes like corporate transactions, be built out as toolkits, or be a repository for information about a particular topic. They can be transitory or permanent. They can be what you need them to be. Light touch or heavily curated – either is fine. Organizations with Knowledge resource can take advantage of the rich feature set that allows them to order and structure the collection for consumption in the Insight+ interface. Where Knowledge resource is scarce, organizations can take advantage of the “virtual” benefits — making the collection available only to the iManage MCP Server with Insight+ Collections tool. Nothing is copied out of iManage. The result? When a lawyer's AI assistant needs to reason about employment law obligations or M&A process, it points at the relevant collection. The answer comes from the firm's knowledge, not from whatever the individual lawyer chose to pull out of the governed iManage environment.

You can see the creation workflow in a demo: a knowledge admin builds a collection from the Insight+ interface, attaches documents with their existing metadata, nests them up to four levels deep, and publishes. The result is searchable by end users and serviceable by AI tools — and the source documents stay in iManage. Your knowledge collections become the trust layer for your AI ecosystem, ensuring your AI tools are always reasoning over the latest information.

For firms not yet at full curation maturity, as well as the virtual benefits, AI-assisted submission and curation help close the gap. Using the iManage MCP Server — the Model Context Protocol (MCP) connectivity layer available to all iManage Cloud subscribers — customers can leverage their AI agent to identify candidates for submission and then populate metadata fields as part of the submission workflow.

What this means in practice

At a recent conference, Tom Humberstone, Knowledge Manager at Womble Bond Dickinson, described the shift he sees underway. The firm has used Insight+ for institutional knowledge for years. Humberstone highlighted that a history of good curation now makes governed access into AI tools possible, whether through Ask Knowledge, a collections interface, or an MCP integration. "Having well-formed, governed content with the appropriate metadata that you can target," he says, "puts you in a good place to address any risks, concerns, or challenges."

Alex Smith, Senior Director of Product – Search, Knowledge, & AI, says the benefits go further; the firms seeing the most from their AI tools are those that have treated knowledge investment as infrastructure rather than overhead. The context fabric — the metadata, the curated collections, the governed relationships between documents, matters, and people — is what separates a trusted AI output from one that might be wrong.

Knowledge Collections for Insight+ are due for release later this year. Organizations already running iManage Insight+ have the foundation in place.

Explore how iManage Insight+ supports AI-ready knowledge management.