Somewhere in the last eighteen months, the document management system quietly became the most important piece of legal AI infrastructure most firms already own — and most people barely think about. In a recent Above the Law interview, iManage CEO Neil Araujo made the case plainly: the firms best positioned to succeed with AI aren’t necessarily the ones buying the flashiest new point solution. They’re the ones that have spent decades building disciplined information management practices, and are now discovering that discipline is exactly what AI needs to be useful.

Araujo’s framing is worth sitting with: “your data foundation, plus an LLM, plus how you change your process is what gives you AI outcomes.” Note the third variable. It’s not just about the platform and the model — it’s about process, which is really shorthand for people who know what to do differently. A document management system re-engineered around AI, permissions, and knowledge graphs is a genuine leap forward. But a leap forward in the tool doesn’t automatically translate into a leap forward in outcomes if the humans profiling documents, running searches, and reviewing AI-generated summaries haven’t been trained on what changed and why it matters.

The Foundation Is Only as Good as What Gets Built On It

This is the theme running underneath the whole DMS-reimagined conversation happening across the industry right now — not just at iManage, but at NetDocuments and other platforms making similar moves toward “context graphs” and AI-ready knowledge layers. The pitch is consistent: AI output is only as good as the context it can reach, and that context lives in how well a firm’s documents, matters, and institutional knowledge are organized, tagged, and governed.

Here’s the training gap that pitch tends to skip over. Firms have had document management discipline “for governance reasons, for improving collaboration,” as Araujo puts it — largely because they had to, for compliance and business continuity. That legacy discipline is a genuine asset in the AI era. But asking staff to now think about documents not just as records to be filed correctly, but as signals that feed an AI system’s judgment, is a real behavior change. Metadata quality, consistent matter organization, and thoughtful version notes stop being administrative housekeeping and start being inputs that directly shape what an AI assistant tells a partner about a matter. Most people doing that filing today have never been told that’s now the stakes.

Three Roles, Three Different Training Needs

A reimagined DMS creates new responsibilities up and down the org chart, and they’re not the same responsibilities:

• Administrative and support staff need to understand why consistent profiling and metadata practices matter more than ever — their day-to-day filing habits now directly shape AI search quality and matter summaries firmwide.
• Associates and paralegals need training on how to use natural-language search and AI-generated matter overviews critically — treating them as a fast on-ramp to a matter, not a substitute for actually reading the file.
• Partners and practice leaders need to understand the governance side: what “permission-governed” AI access really means for ethical walls, client confidentiality, and how to supervise associates who are now getting AI-assisted matter summaries instead of building institutional knowledge the old-fashioned way.

None of that training happens automatically because a firm signed a new DMS contract. It happens because someone builds and delivers it.

Turning Platform Investment Into Actual ROI

This is squarely where Savvy Training & Consulting comes in. We help firms and legal departments close the gap between “we bought the AI-ready platform” and “our people actually use it the way it was designed to be used” — through the SavvySMART Content Library, role-specific modules delivered via SavvyAcademy ProPlus LMS, and custom on-site or hybrid training built around your firm’s specific DMS, its new AI capabilities, and your existing technology stack.

Firms have spent decades earning the “good information management” reputation that’s now paying off in the AI era. Don’t let that advantage sit half-realized because the humans running the system day to day were never brought up to speed on what it can now do.