“Knowledge is power.” - Francis Bacon
AI was supposed to change everything. For most companies, it hasn’t. Not because the models aren’t good enough. But because the understanding of the business - the actual knowledge of how it operates - is broken, scattered, and inaccessible to the systems that need it most.This isn’t a data problem; it’s a knowledge problem. And solving it requires a role the industry completely forgot:The Knowledge Engineer.
What Is a Knowledge Engineer?
A knowledge engineer is responsible for capturing, structuring, and defining how a business actually works so machines can understand, reason, and act on it.
This includes:
Defining core business concepts - what “customer”, “revenue” and “product” actually mean across the organization
Standardizing meaning across systems - reconciling how different tools represent the same things
Encoding relationships, rules, and logic - how data flows, transforms, and connects
Creating a machine-readable representation of the organization - a structured understanding, not just raw data
A knowledge engineer turns fragmented data into structured understanding. Without that understanding, even the most powerful AI is guessing.
From the earliest days of computing, the idea was clear: If you want machines to behave intelligently, you need to give them structured knowledge about the world. Edward Feigenbaum, one of the pioneers of AI, defined it decades ago:“Knowledge engineering is the process of building an expert system by extracting knowledge from human experts and encoding it into a knowledge base.”
The role was clear. The mission was clear. And then, quietly, it disappeared.
Knowledge Engineering vs. Data Engineering
These two roles are closely related, but they solve fundamentally different problems.
Data Engineering asks: Is the pipeline running? Is the data fresh? Is the schema correct?Knowledge Engineering asks: Does the system understand what this data means? Can it reason about it? Can it act on it correctly in a situation it hasn’t seen before?
Data engineering is about how data moves from A to B. Knowledge engineering is about why it matters and what it means when it gets there. Both are essential. But for the past two decades, we’ve only invested in one. And that’s exactly why AI isn’t working for most organizations.
Why Knowledge Engineering Disappeared
Sounds like a critical concept, right? So where the hell was it all those years?
It didn’t fully disappear. The work got absorbed into data engineering, but in the process, it lost the part that mattered most. The technical side survived, the knowledge side didn’t.
If you look at what knowledge engineers were supposed to do, it’s almost identical to what today’s data engineers and analytics engineers do: connect source systems, unify definitions, and implement how the business is represented in data. But over time, something shifted.
Capturing organizational knowledge was brutally hard. The knowledge needed to run a business doesn’t sit in one place. It’s hidden across thousands of systems, spread across the heads of different people, buried in undocumented processes and tribal conventions.
Capturing it meant pulling data from here to there, unifying definitions across departments, running around the organization interviewing stakeholders - and then doing it all again when something changed. The work translated into a deeply technical, deeply tedious process: connecting source systems, processing data, reconciling conflicting definitions, and constantly updating representations that went stale the moment someone changed a pipeline.
It was just too hard. The cost of building and maintaining a knowledge base exceeded what most organizations could sustain. And even when teams managed to capture some of it, the systems available couldn’t do anything meaningful with it. The knowledge had nowhere to go.
What Replaced it: The Shift From Knowledge to Data
As data systems grew more complex, the work required deep technical skills: ingesting data from multiple sources, processing large volumes efficiently, designing architectures, and navigating an ever-growing stack of tools. The role attracted highly technical people, and the focus moved toward solving these challenges. And slowly, the framing changed.
Instead of thinking in terms of representing how the business works, the work became about tables, pipelines, and infrastructure. The language shifted from knowledge to data.
Rather than capturing the organization’s understanding of itself, the goal became “making data usable” - which was already hard enough. The broader mission - creating a coherent representation of how the business operates - became implicit, fragmented, and unowned.
Over time, the balance within the role shifted. The technical side kept growing - more tools, more scale, more complexity - while the business understanding side gradually faded. What started as an effort to represent how the business operates, became an effort to manage data infrastructure.
This created growing frustration that anyone in data recognizes: despite massive investment in pipelines, models, and systems, teams were still not able to fully capture how the business actually works. The original promise - a structured, reasoned representation of the organization - was never fulfilled. It was just renamed and diluted.
Why Knowledge Engineering Matters Now More Than Ever
Because this gap has never been more costly.
Today, we have incredibly powerful systems that can understand, reason, and act. LLMs and AI agents can process information, draw inferences, and execute complex tasks across tools and environments. But without the right organizational knowledge, that power doesn’t translate into real outcomes.
It’s like handing a Formula 1 driver a school bus and asking them to drive - without telling them which kids to pick up, where to go, or what to do when the kids start fighting in the back. They’re highly capable, but set up to fail.
Every organization runs on concepts that people understand intuitively - what “revenue” means, who a “customer” is, how a process works - even when every system defines them differently.
The ability to engineer this knowledge is what unlocks the real value of AI in enterprises. Without it, even the most advanced systems will produce inconsistent, unreliable results. And this is where companies are stuck today.
According to a recent report by MIT, 95% of organizations are getting zero return from their AI investments, a gap they describe as the “GenAI Divide.” The outcomes are sharply split between the few who succeed and the vast majority who don’t.
This isn’t a model problem, it’s a knowledge problem. The bottleneck is no longer intelligence, it is understanding.
The few organizations succeeding with AI are the ones that have figured out how to package their organizational knowledge - the entities, relationships, rules, and tribal context - into something AI can actually work with. Everyone else is feeding raw data into powerful models and wondering why the outputs are inconsistent.
What Changed: Why Knowledge Engineering is Coming Back
So what changed? Why is knowledge engineering suddenly possible, and necessary, again?
AI can now reason over knowledge.
The old constraint - that no system was smart enough to use structured knowledge - is gone. In the past years, AI agents have become good enough to operate across tools, environments, and workflows. They don’t just generate outputs - they reason through tasks, trace problems across systems and connect cause and effect across pipelines, schemas, and business logic. Knowledge engineering finally has a consumer worthy of the investment.
And the technical barriers that made capturing knowledge impractical? AI is removing those too. Connecting to systems, extracting schemas, tracing lineage, mapping dependencies - much of the heavy lifting can now be abstracted away.
The focus of the role shifts back to where it was always meant to be: the business itself. The people shaping this knowledge don’t need to be experts in data systems. They need to understand how the organization operates, define its core concepts, and decide what should be captured and how it should be structured.
As technical barriers fall, the work moves from managing data to shaping knowledge. And in doing so, data engineering begins to evolve back into knowledge engineering.
The Knowledge Engineer is back - but this time as a core function of every AI-driven organization.
—---
This is the shift we’re building for at Upriver.
—---
FAQs: The Rise of the Knowledge Engineer
What does a knowledge engineer do?A knowledge engineer translates how a business actually works into structured, machine-readable knowledge. They define key concepts, align inconsistent definitions across systems, and encode the relationships and rules AI systems need to reason and act correctly.
How is knowledge engineering different from data engineering?Data engineering focuses on pipelines, infrastructure, and moving data. Knowledge engineering focuses on meaning - ensuring systems understand what the data represents, how concepts relate, and how to use that understanding in real-world decisions.
Why is knowledge engineering critical for AI success?AI systems fail when they lack context, not capability. Without structured knowledge AI produces inconsistent results. Knowledge engineering provides the foundation that makes AI reliable and useful.
Why are most companies not seeing ROI from AI?Most organizations feed fragmented and inconsistent data into AI systems without a unified understanding of their business. This leads to outputs that don’t reflect reality, limiting adoption and preventing meaningful returns.
Why is knowledge engineering becoming important again now?AI systems can now reason over structured knowledge, and new tools make it easier to capture that knowledge from complex systems. This combination makes knowledge engineering both practical and essential for modern AI-driven organizations.
How does a data engineer become a knowledge engineer?
The transition happens as AI platforms handle more operational data engineering work investigation, pipeline maintenance, firefighting. As that burden lifts, data engineers focus on higher-value work: defining business concepts, shaping how the organization's data environment is understood, validating AI reasoning, and making tribal knowledge permanent. It's not a career change. It's the same role, evolved - from doing the work to shaping it.
—-----------