IA before AI: the critical oversight most organizations are missing

There’s a statistic from the iManage Knowledge Work Benchmark Report 2026 that I keep coming back to. Eighty-five percent of legal business and technology decision-makers are already piloting, implementing, or using AI. Yet only 17 percent say their AI tools are fully integrated and widely used. Surprising though that stat may be, what’s more surprising is that the gap isn’t a technology problem; it’s an information architecture (IA) problem.

The models are capable, and if you've sat through enough vendor demos, you've seen what's possible when the inputs are clean and the context is rich.

The problem is that most organizations’ knowledge doesn't look like a vendor demo. 

Documents sit in the DMS with limited metadata, disconnected from business, operational people, and matter context, and largely invisible to any AI tasked with reasoning over them. Without that foundation, the gap between what AI promises and what it delivers shows up fast: answers from irrelevant past work, embarrassing hallucinations, governance blind spots that create risk, and tools that ask lawyers to change how they work rather than fitting into how they already do.

Clearly, for AI to function properly, the way organizations approach their knowledge foundation must change.

Information architecture before artificial intelligence

At a recent conference, I spoke about how iManage Insight is purpose-built to unlock richer context fabric™ capabilities, transforming your firm's most important historical and current content into a structured, actionable knowledge base that powers trustworthy AI. Insight+ was built years before generative AI entered mainstream conversation, grounded in the belief that having structured context alongside the outputs of knowledge work (contracts, motions, advice notes) would determine how well firms could leverage their knowledge. Gartner has since put a number on that thesis: by 2027, they suggest, organizations that design their context fabric in advance rather than discovering it at runtime will see an 80 percent increase in generative AI model accuracy and a 60 percent reduction in costs.

The reason organizations are stalling today is straightforward: When AI draws from content that lacks context, it produces output that’s plausible, but wrong. Lawyers can’t defend it, they stop trusting it, and adoption collapses. The fix isn’t a better model. It’s better input.

What the context fabric™ means in practice

iManage context fabric has two layers that are part of a broader platform capability. The first is organization and business context: Insight+ connects to an organization's data store and applies approximately 200 context fields across people, matters, and clients, propagating that meaning to every document and email in the DMS. The second is knowledge context — curated best practice or AI-ready collections: more than 100 extended metadata fields applied to an organization's most valuable institutional content, including precedents, playbooks, templates, and practice updates.

The broader platform capability we are building includes these two capabilities as well as AI Enrichment that can be seen in Insight+ and Work, and more context from within active work or matters — producing structured, trusted, enriched data. That is the foundation that makes better search, smarter AI, and reliable discovery work in a way that lives up to the AI hype.

Getting your information architecture right before layering in AI delivers compounding returns across your organization:

• Increases trust in AI output across every tool you invest in. When AI draws from structured, enriched, governed content, it produces results lawyers can verify and defend. That trust is what drives adoption, regardless of which AI tools your organization chooses.
• Improves search quality right now, not just when AI matures. Properly organized content with rich metadata aligns with how your people actually work, surfacing the right knowledge at the point of need and reducing the time spent hunting for documents that should be easy to find.
• Reduces the risk of hallucinations and ungoverned AI usage. Structured context filters out irrelevant, outdated, or contradictory content before it reaches an AI model, significantly reducing the likelihood of outputs that mislead or create compliance exposure.
• Protects and multiplies the value of every future AI investment. A well-built context fabric does not just support the tools you have today. It means every new AI capability your organization adopts lands faster, performs better, and earns broader adoption from day one.

Why search improves first, and why that matters

Prioritizing deeper context today isn’t just an AI story; it’s also a search story. The combination of meager metadata, minimal curation, and no business context means that most DMS users’ keyword retrieval is unlikely to yield the best results. When context is prioritized, that changes. Users can easily filter by practice area, matter type, deal value, jurisdiction, and dozens of other fields that reflect how legal work is actually organized.

Natural language search over the knowledge library, Ask Knowledge, is now in early access and lets people ask questions in plain language and get answers grounded in the organization's curated content, with verifiable citations. This will expand to asking broader knowledge “collections” as that feature goes live. When search improves, trust improves, and that trust is the precondition for adopting AI tools built on the same foundation.

Knowledge Collections: turning institutional knowledge into an AI-ready asset

Knowledge Collections is the capability I’m most excited about right now. Organizations have valuable institutional knowledge: precedents, checklists, playbooks, and gold-standard examples. But currently they exist across myriad solutions, administrative matters, shared drives, and so forth, which lawyers can’t reliably navigate or easily point an AI tool at, and they often can’t be sure the information they locate is complete. Collections are curated, structured, governed sets of content built around a specific purpose: to provide practice teams, partners, and legal teams a definitive, maintained resource for a practice area, client, or transaction type.

Collections become the trust layer for AI. When a legal AI tool draws from an ungoverned pool of documents, the output quality is unpredictable. When it draws from a Collection, the output is grounded, defensible, and aligned with how the organization realistically practices. Before the end of 2026, AI-ready Knowledge Collections will be accessible via Model Context Protocol (MCP) to tools such as Harvey, Legora, Microsoft Copilot, and Claude, through the iManage MCP Server. The organizations building their Collections now will be the ones whose AI outputs are defensible in 12 months.

iManage MCP Server: Putting knowledge into motion

One of the most important shifts with Insight+ is the move from knowledge as a destination to knowledge as a service. Through iManage MCP Server, governed knowledge and context from Insight+ can now flow directly into the AI tools lawyers already use: Harvey, Copilot, Legora, Claude in Word, and custom agent pipelines. A lawyer drafting in Word can ask a question and receive an answer grounded in the organization's knowledge library, and without leaving their document.

This means the curation work organizations do today have a much longer reach. Curate once, govern always, discover instantly, and create with confidence: the MCP integration is what makes that promise real across every surface.

A practical note on implementation

There’s no question about it, building the context fabric requires upfront work. Content needs to be organized with the right metadata, knowledge needs to be curated and enriched, and collections need to be structured and governed. I know that creates hesitation in a market where AI is often positioned as instant and out-of-the-box. But the more important question is: what’s the cost of deploying AI without this foundation? That 85-to-17 percent gap we noted in our benchmark data between orgs that are piloting, implementing, or using AI and those that are fully integrated and widely used provides a sobering answer.

The organizations that will get the most from AI are the ones investing in their information architecture now, not as a prerequisite that delays AI, but as the layer that makes every AI investment more accurate, more trusted, and more widely used. Done right, prioritizing the increase of your investment in knowledge has the power to pay dividends not once, but with every new AI capability your firm adopts going forward, setting your knowledge into motion.

The organization that knows more wins more

Context Search, Matter Search, Ask Knowledge, Collections, and iManage MCP Server integration are all live or in advanced access today. The question for your organization isn’t whether AI matters, but rather whether the knowledge your AI depends on is ready. If it isn’t, the path forward to realizing the benefits of iManage context fabric starts with Insight+ and AI Enrichment, and it starts now.

Making the next step

If your organization is ready to build the foundation for more trusted, effective AI, schedule a personalized demo to see how Insight+ can help.