Proven Experience

A Personal Journey Toward Applied-AI

By David H. Deans

I've spent much of my professional life working with business technology companies and the people responsible for turning applications into business outcome value.

Looking back, there is a thread connecting that work that wasn't entirely obvious to me at the time.

I've been interested in a deceptively difficult question:

How can I capture what people know—and turn that knowledge into something others can use?

I didn't start asking that question because of artificial intelligence (AI). I was thinking about it long before Generative AI and Large Language Models became part of the enterprise conversation.

Where We've Been

Earlier in my career, I spent considerable time working with telecommunications and IT organizations. One recurring problem became increasingly apparent to me: some of the most valuable knowledge in an organization wasn't contained in its databases, manuals or formal processes.

It was contained in people's heads. Memories of best practices that were proven by experience.

Experienced practitioners knew how customers behaved. They knew which objections mattered and which didn't. They recognized patterns that weren't obvious to less experienced colleagues. They knew when to follow a process and when circumstances required an exception.

Much of that knowledge was difficult to document because it was inseparable from experience.

I became interested in ways of connecting people to that expertise and capturing some of what experienced practitioners had learned. That thinking eventually contributed to ideas around expertise networks and what I called an ExperienceBase.

The objective wasn't simply to create another knowledge repository.

It was to make the experience discoverable and actionable.

That work eventually led to GeoNetworker, my R&D exploration into expertise networks, knowledge transfer and organizational intellectual capital.

At the time, the technology available to us imposed significant limitations. We could connect people. We could organize information. We could create communities of practice.

But capturing the deeper reasoning of an expert was much harder.

The problem was not information.

It was context.

Where We Are Today

Generative AI changes that equation.

For the first time, we have technology capable of interacting with enormous amounts of unstructured material and helping us identify relationships, patterns and meaning within it.

That creates an interesting possibility for enterprises.

Instead of beginning with the question: "What information can we give the AI?" perhaps we should begin with: "What does our organization know that isn't captured in our information systems?"

Consider the experienced employee who has spent twenty or thirty years working with customers.

Their skills may be scattered across presentations, emails, proposals, meeting notes and project files.

But the most valuable part may not exist in any of those things.

It may be the judgment developed through thousands of interactions.

Why did they make that decision?

What did they recognize that others missed?

Which exceptions mattered?

What questions did they learn to ask?

What signals told them that something was about to go wrong?

This is what I have come to think of as Tacit Knowledge Harvesting.

The objective isn't simply to interview people and produce transcripts. It is to uncover the reasoning, context, heuristics, assumptions and practical lessons embedded in their experience — and then determine how that knowledge can be validated, structured and made useful to the organization.

This brings me back to the ExperienceBase idea, but in a very different technological environment.

The original question was: How can we connect people with the expertise they need?

The emerging question is: How can we make an organization's collective expertise available to its AI systems without losing the context that makes that expertise valuable?

I think that is a much more consequential question.

From Institutional Memory to Organizational Intelligence

I've become increasingly interested in the distinction between preserving knowledge and creating organizational intelligence.

A traditional knowledge-management system might preserve an expert's documents when that person leaves the organization.

That's useful.

But it doesn't necessarily preserve the reasoning behind those documents.

Generative AI potentially gives us an opportunity to do something different.

We can begin to capture and organize the accumulated experience of people in ways that allow other people — and potentially AI systems — to access it.

That could change how we think about institutional memory.

The objective isn't merely to prevent knowledge from disappearing when an employee retires or moves on. It is to transform individual expertise into a valuable organizational asset.

I think this has particular relevance to the emerging discussion around Sovereign and Private AI.

Much of that discussion understandably focuses on infrastructure, data residency, security, governance and control. Those are essential considerations.

But I believe there is another dimension of sovereignty that deserves more attention:

Who owns the intelligence expressed through the AI?

An organization can operate a private AI model and still rely primarily on generic knowledge.

The more interesting opportunity is to create AI that understands the organization's own experience — its customers, markets, processes, decisions, exceptions and accumulated expertise.

In other words, AI that understands what makes the organization different.

Where We're Going from Here

This is where my thinking about Applied-AI Initiatives is taking me.

I don't see Applied-AI simply as the implementation of another generation of technology.

I see it increasingly as a question of strategic foresight.

What is becoming possible?

What will become economically practical?

Which forms of human expertise will become more valuable rather than less?

Which organizational knowledge is at risk of disappearing?

What proprietary intelligence could become a competitive asset?

Where should an organization use a large general-purpose model, and where might a smaller, specialized model be more appropriate?

How should people, knowledge graphs, retrieval systems, AI agents and language models work together?

And perhaps most importantly: What should an organization begin doing today to prepare for capabilities that will become possible tomorrow?

These questions sit at the intersection of technology, strategy and human experience.

That is where I've spent much of my career.

The technology has changed dramatically.

The underlying question has changed much less.

I've been interested for a long time in how organizations discover, capture, share and apply what their people know. Today, AI gives us tools that make that problem both more difficult — and potentially much more valuable.

The opportunity I see ahead is not simply artificial intelligence.

It is organizational intelligence augmented by AI.

And perhaps the most valuable AI asset an enterprise will develop won't originate in a model at all.

It may already exist inside the heads of its people.

The strategic challenge is figuring out how to bring it out, preserve its context, and put it to work.