Why Enterprise AI Needs a Common Language for Work

The Missing Rosetta Stone of Workforce Planning

Every meaningful workforce decision begins with the same core questions:

• How does the workforce we have today compare to the workforce in the market or at our competitors?

• How does today's workforce compare to the workforce we'll need tomorrow to execute our business strategy?

Everything else – hiring, learning, workforce planning, internal mobility, automation, organizational design – is simply a different way of asking or answering one of those two questions. Which makes the next fact all the more remarkable.

After nearly three decades of HR technology innovation, no one system has ever been able to answer those two questions. Not because the questions are too difficult. It's because the answers were written in different languages.

For years we've assumed workforce planning was fundamentally a tooling problem.

• We bought better labor-market intelligence.

• Better skills platforms.

• Better people analytics.

• Better talent marketplaces.

• Better planning software.

Every generation of tech promised to solve the problem the previous generation couldn't. Yet somehow the work never became dramatically easier. If anything, it became harder.

Analysis still takes weeks. Data still arrives late. Executive teams still question the assumptions. HR teams still spend enormous amounts of time reconciling spreadsheets before they can begin producing insight.

That's usually interpreted as evidence that we still haven't found the right tool. I don't think that's true. I think we've been trying to solve an architectural problem with applications.

Imagine asking five people to jointly write a report.

• One writes in English.

• Another writes in Mandarin.

• Another writes in Arabic.

• Another writes in Greek.

• Another writes in Hindi.

Now imagine asking someone else to combine those documents into a single coherent strategy paper every quarter. You wouldn't blame Microsoft Word for how difficult that exercise is. You'd recognize the obvious problem: no one ever agreed on a common language.  

That is where workforce planning has quietly lived for the last twenty years.

There is finally some chatter among thought leaders that suggest that there are two worlds that need to be connected.

• Your internal workforce.

• The external labor market.

If only it were that simple. The external market speaks one language – supply, demand, compensation, hiring velocity, talent availability, competitor behavior, emerging skills.

But the moment you step inside the enterprise, you enter a multilingual environment. There isn't one internal language. There are many.

• Your HRIS defines work one way.

• Your learning platform defines it another.

• Your talent marketplace infers something different.

• Your recruiting platform describes it differently again.

• Compensation has its own structures.

• Job architecture has another.

And then there are tasks – the smallest unit of work and the unit where AI and automation actually create value – which, in most organizations, barely exist as structured data at all.

Every one of those systems is internally logical. Every one was designed for a legitimate purpose. Every one is speaking a different language.

This is the insight that quietly reframes nearly every conversation about workforce technology. The problem has never been a shortage of applications. It has been the absence of a shared semantic model of work.

Internal systems were designed to optimize internal workflows. External data was designed to describe external markets. Neither was ever intended to become the connective tissue between them.

So we ask disconnected systems to answer integrated questions. Then we blame the software when they can't.

That distinction matters because it changes where the problem actually lives.

For years we've assumed the missing layer in workforce planning was another application. Another dashboard. Another AI assistant. Another planning engine.

In reality, the missing layer has been much deeper. It is the absence of a canonical representation of work – a common set of definitions for roles, tasks, skills, and work itself that every system inside and outside the enterprise can understand.

Without that shared foundation, every comparison requires translation. Every translation introduces interpretation. Every interpretation introduces inconsistency. And every inconsistency compounds until strategic workforce planning becomes less about planning and more about reconciliation.

We call it workforce planning. Too often, it's really translation.

Why No One Talks About This

If this diagnosis is right, it raises an obvious question. Why has almost no one in our industry framed the problem this way?

The answer isn't that people don't see the fragmentation. It's that almost everyone experiences only one piece of it.

• The labor-market team sees labor-market data.

• The learning team sees learning data.

• Talent acquisition sees recruiting data.

• People Analytics sees HRIS data.

• Compensation owns job architecture.

Every function is solving a real problem. Every software vendor is building a valuable capability. Every analyst has developed deep expertise within their domain.

The problem is that the fragmentation isn't visible from inside any one of those domains. It only becomes obvious when you step back far enough to see the entire enterprise.

Ironically, nearly every major category in HR technology is built around one of these languages.

• A labor-market platform speaks the language of external supply, demand, cost, competition, and skill evolution.

• An HRIS speaks the language of employees, positions, organizational structures, and systems of record.

• A learning platform speaks competencies, skills, curricula, and capability development.

• A recruiting platform speaks requisitions, skills, candidates, responsibilities and tasks, sourcing pipelines, and hiring workflows.

• A talent marketplace speaks inferred skills, aspirations, and internal mobility.

• People Analytics attempts to synthesize pieces of several of these worlds but is ultimately constrained by whatever data model happens to exist underneath it.

Every category is fluent. None of them is multilingual.

That isn't a criticism of the vendors. It's a reflection of what each category was designed to do. You don't criticize a dictionary for not being a novel. You don't criticize a GPS for not being an accounting system.

Likewise, you shouldn't expect a recruiting platform to become the enterprise's semantic model of work. Nor should you expect your Talent Marketplace solution to be that. Nor your LMS. Not your HRIS. None of these solutions were built for that.

This is also why organizations continue buying more technology while seeing only incremental (if any) improvements in strategic workforce planning. Every new application makes one language richer. Very few make the languages more compatible. In fact, the opposite often happens. More languages and "capability" actually means less ability to "connect the dots" and see the big picture.

• Every implementation introduces another taxonomy.

• Another ontology.

• Another definition of roles.

• Another interpretation of skills.

• Another set of task descriptions.

• Another data model that must somehow be reconciled with every other one already in the organization.

The enterprise becomes increasingly intelligent within individual systems while becoming progressively less coherent and less insightful across and between them. That is a remarkable paradox. Worth maybe a beat or two to really let that sink in...

We have never had more workforce data. We have never had more workforce intelligence. And yet, many organizations still struggle to answer the most basic strategic questions with confidence because intelligence without shared meaning cannot become shared understanding.

Artificial intelligence is making this gap impossible to ignore. Large language models are remarkably good at reasoning. Agentic AI promises to orchestrate increasingly sophisticated workflows across enterprise systems.

But neither changes a fundamental constraint. AI can only reason consistently if the underlying concepts are consistent. Ask five AI agents to analyze five systems that each define "Software Engineer," "Customer Success Manager," or "Senior Analyst" differently, and they will faithfully reason across five different realities. Welcome to the HR Multiverse.  

The AI isn't confused. The enterprise is. That may be the most important implication of all.

For years, semantic inconsistency was an HR operations problem. Now it is becoming an enterprise AI problem. Because every intelligent system ultimately depends on a common understanding of the thing it is reasoning about.

This is why I increasingly believe Work Architecture is the foundational solve to this problem. But we're mislabeling and miscategorizing it.

Most organizations still think of it as an HR project.

• A job architecture refresh.

• A compensation exercise.

• A leveling exercise.

• A skills initiative.

• A taxonomy.

It is none of those things. Or more accurately, it is all of those things – but they are merely the visible manifestations of something much more fundamental.

Work Architecture is enterprise infrastructure. It is the semantic layer that allows every application, every AI agent, every planning process, and every business function to reason from the same understanding of work.

Not simply because consistency is cleaner. Because consistency is what makes enterprise intelligence possible.

Without it, every system continues optimizing its own interpretation of reality. With it, they begin reasoning from a shared one. That is a profoundly different operating model.

The Rosetta Stone, after all, wasn't valuable because it translated three ancient languages. It changed history because it revealed that they were describing the same underlying ideas. That is the opportunity sitting in front of workforce planning today.

Not another dashboard. Not another skills graph. Not another AI copilot. A shared language for work itself.

The Missing Layer

We've spent decades integrating systems and passing data back and forth. We never standardized what they were actually describing.

That missing layer is what I believe Work Architecture ultimately becomes. Not another taxonomy. Not another skills framework. Not another job catalog.

The canonical representation of work across the enterprise. A shared semantic model that defines work once and allows every system to reason from the same foundation.

That distinction matters. Because semantic consistency is fundamentally different from systems integration.

Integration allows applications to exchange data. A shared semantic model allows applications to exchange meaning. Those are not the same thing.

Imagine an enterprise where your HRIS, recruiting platform, learning ecosystem, talent marketplace, compensation systems, organizational design tools, workforce planning platform, finance models, and labor market intelligence all describe work using the same underlying definitions.

• Roles

• Tasks

• Skills

• Capabilities

• Career paths

• Automation opportunities

• Work expectations

• Work value and outcomes

Not merely linked together through APIs. Actually speaking the same language.

In that world, reconciliation doesn't become faster. It disappears. The enterprise no longer spends thousands of hours translating between incompatible taxonomies because the translation requirement disappears entirely. Every application begins from the same understanding of reality.

The implications multiply remarkably quickly.

• Labor market intelligence stops being an external benchmark and becomes a continuous planning input.

• Skills stop being isolated taxonomies and become a common currency connecting hiring, learning, workforce planning, compensation, career mobility, and organizational design.

• Tasks become first-class enterprise data, allowing organizations to reason explicitly about automation, augmentation, productivity, and work redesign rather than approximating them through job titles alone.

• AI shifts from isolated copilots embedded inside applications to intelligent agents reasoning across the enterprise from a shared understanding of work.

• Scenario planning shifts from quarterly exercises to continuous decision support.

• And strategic workforce planning evolves from "stitching together disconnected reports" to a "living model of work that updates as the business and labor market change."

That is not simply better planning. It is an entirely different operating model.

It also changes the role AI plays inside the enterprise. Today, most AI systems inherit the fragmented view of work that already exists. They become extraordinarily capable assistants operating inside disconnected realities.

But give those same systems a canonical representation of work – a shared semantic foundation spanning every role, task, and skill – and something different happens. AI agents stop reasoning inside applications. They begin reasoning across the enterprise.

• A recruiting agent understands the same role definitions as a learning agent.

• A workforce planning agent reasons from the same task model as Finance.

• An organizational design agent evaluates the same skills taxonomy used by internal mobility.

Every agent shares context because every agent begins from the same understanding of work. That is a fundamentally different future than simply embedding AI into existing software. It is giving AI a coherent model of the enterprise itself.

This is why I increasingly believe Work Architecture is not an HR capability. It is enterprise infrastructure. It becomes the semantic operating system for work. The canonical representation that every HR application, every AI agent, every finance model, and every business planning process can reason over.

Workforce planning simply happens to be the first place where the value becomes obvious. It will not be the last.

We've Been Buying the Wrong Things

A growing number of Fortune 1000 organizations are beginning to recognize this shift. Not because they've discovered another workforce planning application. Because they've started asking a fundamentally different question.

For years, enterprise buying decisions have revolved around applications.

• Which recruiting platform?

• Which labor market intelligence provider?

• Which skills platform?

• Which talent marketplace?

• Which workforce planning tool?

Those are reasonable questions. They just shouldn't be the first questions.

The first question should be architectural: "What foundation allows every system that describes work – inside and outside the enterprise – to speak the same language?"

Because once that foundation exists, the conversation changes. Everything becomes faster. More consistent. More explainable. More trustworthy.

Not because the algorithms improved. Because the enterprise finally agreed on what it was talking about.

The last twenty years of HR technology have largely been about digitizing processes. The next twenty years will be about creating a digital representation of work itself.

Those are profoundly different ambitions. One automates workflows. The other creates a model of reality.

And once that model exists, every application built on top of it becomes smarter. Every AI agent becomes more capable. Every planning process becomes more connected. Every workforce decision becomes more defensible because every participant – human and machine – is reasoning from the same underlying understanding of work.

Which brings us back to where we started. Every meaningful workforce decision ultimately asks two deceptively simple questions.

• What do we have?

• What do we need?

For decades, we've assumed those questions were difficult because workforce planning is inherently complex. I no longer believe that's the primary reason. I think they have been difficult because every answer had to be translated before it could be understood.

A shared Work Architecture becomes the canonical representation of work across the enterprise – the semantic layer that every HR application, every AI agent, every finance model, and every business planning process can reason over.

Instead of every system maintaining its own interpretation of roles, tasks, and skills, they all operate from the same underlying definitions. And the business gains something it has never really had before: A common understanding of work itself.

So I'll leave you with the question I now ask in almost every executive meeting.

"When your organization compares its workforce to the labor market – or compares one internal system to another – or your today state to your desired future state –

What language are you actually doing it in?

One shared language?

Or five different languages...

...a spreadsheet...

...and a prayer?"