AI-enabled knowledge management: A new imperative for government

Knowledge management has been around for decades and has become somewhat of a self-standing discipline – that is, until generative artificial intelligence arrived. For years, knowledge management often struggled for executive attention because its value was indirect.

AI changes that. When leaders see an assistant who can find, explain, and even execute work based on organizational knowledge, knowledge management becomes a strategic capability. But the same power that increases knowledge management’s impact also increases the cost of weak governance.

The future of knowledge management is not “a smarter repository.” It’s an organizational capability in which knowledge is continuously improved, securely accessible, evidence-based, and increasingly actionable — with humans still accountable for what the organization claims to know and for what it decides to do with that knowledge.

For a long time, knowledge management felt like a noble — but sometimes thankless — mission. We built repositories, launched portals, argued about taxonomies, pleaded with busy experts to “please update the FAQ,” and celebrated every time someone actually found the right document before calling a colleague. knowledge management was (and still is) about helping people do their work better by making hard-won experience easier to reuse.

As a result of AI, knowledge management is making a comeback with greater visibility, greater usefulness, and — if we’re honest — more complexity. It’s like someone took the old knowledge management promise (“the right knowledge to the right person at the right time”) and turned up the volume. Now the question isn’t just whether people can find information. It’s whether the organization can trust what it’s putting into circulation, and whether it’s ready for knowledge to show up not as a link… but as an answer.

The Old World: People Had to Go Looking

In the classic model, knowledge lived somewhere “over there.” A shared drive. A wiki. A knowledge base. A portal with a search bar. In theory, staff would go there, search, skim, and apply what they found.

In reality, most of us know what happened: people asked the person down the hall, or posted the question in a chat channel, or reinvented the wheel because finding the “right” document took too long. So knowledge management teams kept trying to make the library easier to use — better labels, better search, better structure — while the work kept moving faster.

The New AI World: Knowledge Looks For You

AI flips that. Increasingly, you don’t have to go to the knowledge base. Knowledge comes to you—inside your email, your chat, your case management system, your document editor, your calendar invites. You ask a question the way you’d ask a colleague:

“Do we have a standard response for this?”

“What’s our policy on that?”

“How did we handle this last time?”

“Who’s the right person to contact?”

Ideally, the system tries to answer in plain language. That’s the first big change: knowledge management isn’t just a place anymore. It’s becoming a presence — a kind of organizational voice that travels with staff through their daily work. It also raises the stakes, because a link to a document is one thing. A confident, polished answer — delivered instantly — carries more authority, whether it deserves it or not.

AI Learns From What You Give it

One misconception is the idea that AI will magically fix messy knowledge. It won’t. If anything, it will highlight it.

If your policies conflict, AI may blend them together. If your documents are outdated, AI may serve them as if they’re current. If your “final” version is hiding in someone’s inbox, AI won’t find it. If your staff never trusted the knowledge base, AI won’t change that overnight.

AI is like a bright flashlight in the attic. It helps you see what you have — but it doesn’t clean it up for you unless you put real effort behind it. This is where knowledge management becomes newly valuable. In the past, some leaders saw knowledge management as “nice to have.” In an AI world, good knowledge management becomes the difference between a helpful assistant and a confident misinformation machine.

Knowledge is Shifting

Here’s another change that knowledge management people find genuinely hopeful: AI can take some of the burden off humans when it comes to creating and updating knowledge. In many organizations, that work is too much for the few people assigned to it — and too annoying for busy experts who’d rather do their real job.

AI can help by drafting the “first version” of a how-to, summarizing patterns from past cases, and proposing updates when things shift. That’s not the same as “AI writes our policies.” It’s more like: AI becomes the intern who’s always available, who can produce a decent draft in minutes — then a human reviews it, corrects it and takes responsibility for it.

That human responsibility part is non-negotiable. But the speed change is real. And it can make knowledge feel more “alive,” rather than frozen in time.

When AI Starts Doing Things

A subtle but important turning point is when AI stops being a helpful reference and starts taking action. This is where knowledge management and operations start to blend. Knowledge isn’t just guidance; it becomes instructions for how work gets done. And that’s where organizations need to slow down just enough to do it responsibly.

Trust becomes the main event

People will forgive a knowledge base that’s hard to navigate. They will not forgive an AI assistant that gives them a wrong answer with a confident tone — especially if it leads to a bad decision, a compliance error, or a public embarrassment. So trust becomes the centerpiece of modern knowledge management.

Knowledge Management Becomes a Strategy, Not a Side Project

AI is forcing organizations to confront something they could have postponed for years — what they actually know, what they think they know, and where knowledge lives when it matters. When AI works well, it can reduce rework, shorten onboarding, improve service consistency, and help experienced staff scale their expertise. It can also relieve the “knowledge tax” that falls on a few go-to people who answer the same questions over and over.

But when AI sits atop messy knowledge, it can spread confusion at scale.

What Leadership Model Works Best in Most Governments

A proven structure is a hub-and-spoke model:

1. Executive Sponsor (Deputy/COO/Administrator): sets expectations, removes barriers, ties knowledge management to agency priorities.

2. knowledge management Program Lead (CKO or knowledge management Director): runs the program day-to-day—standards, content lifecycle, training, measurement, and (now) AI-ready knowledge practices.

3. Governance council: CIO + HR/Training + Legal/Records + Security/Privacy + key operating departments. APQC’s research stresses enterprise governance and sponsorship as the “glue” for sustained knowledge management.

4. Department knowledge stewards: each department owns its “authoritative” content and updates—knowledge management doesn’t succeed if it’s “someone else’s job.”

What Kind of Person is Best to Lead it?

Whether or not you give them the CKO title, pick someone who can do three things:

1. Translate mission into “what we need to know” (frontline service, compliance, emergency response, procurement, HR, etc.)
2. Drive adoption across silos (credible with leadership and staff; can negotiate ownership)
3. Operate with discipline (content owners, review cycles, “one source of truth,” feedback loops)

The AI Twist: Adding Accountability

With AI-enabled knowledge management, it is critical for knowledge management leadership to be tied into an organization’s AI governance—whether that’s a Chief AI Officer, data/analytics leadership, or a responsible AI committee in order to best manage:

• permissions and sensitive data exposure,
• “source of truth” and citations,
• human review for high-stakes answers,
• audit trails when AI starts taking actions.

Governance Considerations

There are three practical moves to start:

1. Decide what you want the organization to be able to answer well.Pick the high-value questions people ask constantly: onboarding questions, policy interpretations, customer service responses, procurement steps, IT help, HR basics, and emergency procedures.
2. Clean up and label what matters most.Don’t try to boil the ocean. Take the top knowledge areas and make them reliable: one authoritative version, named owners, clear review dates, and plain language.
3. Build feedback into the system from day one.The fastest way to improve AI-enabled knowledge management is to make it easy for users to say “this helped” or “this was wrong,” and route that feedback to the content owners. knowledge management becomes a loop, not a publishing event.

In government, the “best” leader is usually a dedicated knowledge management lead (often titled chief knowledge officer or knowledge management director) backed by a senior executive sponsor — because knowledge management is as much about culture, process, and accountability as it is about tools.

Alan R. Shark is an associate professor at the Schar School for Policy and Government, George Mason University, where he also serves as a faculty member at the Center for Human AI Innovation in Society (CHAIS). Shark is also a senior fellow and former Executive Director of the Public Technology Institute (PTI). He is a National Academy of Public Administration Fellow and Founder and Co-Chair of the Standing Panel on Technology Leadership. Shark is the host of the podcast Sharkbytes.net.