Career Highways CEO Liz Eversoll explains why enterprises need more than powerful AI models: clear skills data, governed knowledge, and a human at the wheel.
AI is changing work faster than most organizations can update their roles, development programs, and operating models. As companies begin moving from AI assistants to AI agents capable of taking action, that gap is becoming harder to ignore.
Liz Eversoll, CEO of Career Highways, argues that the challenge isn't simply giving organizations access to more powerful AI, which is why she built Skills Intelligence: a living view of the skills connected to roles, employees, career pathways, learning, and business priorities. Instead of leaving job architecture in spreadsheets and static documents, the platform turns it into something organizations and employees can actually use.
Career Highways' recent announcement of SIGN™, Sigil Intelligence Graph Notation, extends that thinking beyond workforce strategy. SIGN is designed to help AI agents understand the facts, rules, constraints, and authority behind enterprise knowledge, not simply retrieve another document.
In the conversation that follows, Eversoll discusses why Skills Intelligence matters, what SIGN adds, and what organizations need to get right as AI agents move into real work.
Eversoll spoke with TechEchelon about why Skills Intelligence matters, what SIGN adds, and what organizations need to get right as AI agents move into real work. This transcript has been lightly edited for length and clarity.
What is Skills Intelligence?
Most companies still make people decisions off job titles and outdated job descriptions. Skills Intelligence replaces that with a living, governed view of the skills behind every role: where the gaps are, how the work is changing, and how someone moves from one role to the next.
The way I describe it: the road to advancement is paved with skills. Skills Intelligence is the GPS, and career pathways are the routes. It isn't a skills library and it isn't a dashboard. It's the decision layer underneath internal mobility, hiring, reskilling, and workforce planning. And it sits on top of each organization's own enterprise knowledge, its roles, its rules, its definition of what "ready" actually means. You need that foundation before you can honestly call yourself a skills-based organization.
Why is that so important now?
Because AI doesn't hit a whole job at once. A role is a bundle of skills and activities. Some of them get automated, some get augmented, and some, judgment, communication, leadership, matter more than they did before.
So "Will AI replace this job?" is the wrong question. It's too blunt to act on. The better one is, "How does AI change the specific skills inside this job, and what should this person learn next?" That's a question a business and an employee can actually do something with.
What is SIGN in plain English?
SIGN, Sigil Intelligence Graph Notation, is a way to write down an organization's knowledge so that both people and AI agents can read it and act on it the same way.
Every company runs on rules. Definitions, exceptions, approval levels, who is allowed to do what. Most of that lives in documents, in systems, and in the heads of experienced people. An AI agent can go find a document, but finding it isn't the same as knowing which rule applies or what it's allowed to do.
SIGN is a shared language for stating that plainly: here's the fact, here's the rule, here's where it came from, here's the limit, and here's the action an agent can or can't take. We're releasing it as an open standard, the same way SQL became the common language for databases, so knowledge can move across systems instead of every company inventing its own format.
Isn't connecting AI to company documents enough?
No, and this is where most "put AI on top of our documents" projects quietly break. Finding information and knowing how to apply it are two different things. An agent can pull up the right policy and still miss the exception. It can find a job description and not realize it's three versions out of date. Documents describe intent. They can't enforce it.
What you actually need is a canon: a single, governed source of truth for your enterprise knowledge. Your facts, your definitions, your policies, your rules, written down in one owned and current place instead of scattered across PDFs and people's memories. SIGN is the language that canon is written in. Canon is the knowledge; SIGN is how it's expressed so a person and an agent read it the same way.
That's the real shift. Once your knowledge lives in a governed canon, everyone and every agent is working from the same information and the same rules. No more one answer in the handbook, a different one from the manager, and a third from whatever the model guessed.
Why is Career Highways moving into this area?
For us it isn't a pivot, it's the layer we were already standing on. We didn't start with AI and go hunting for a problem. We've spent decades inside large organizations, and we built Career Highways on a governed body of workforce knowledge: roles, skills, policies, pathways. SIGN is the notation we created to express that knowledge, and canon is where it lives.
Our own workforce product is the first thing we built on SIGN and canon. Skills Intelligence is really SIGN applied to one domain, the workforce. Once we saw how well it worked for us, open-sourcing the notation was the obvious next step. Organizations can adopt it without lock-in, and their canonical knowledge stays theirs. SIGN is the language; Career Highways is the governed workforce intelligence system built in it.
How do Skills Intelligence and SIGN work together?
Skills Intelligence is the workforce map: roles, skills, gaps, pathways. Underneath it is your enterprise knowledge, written in SIGN and governed as canon.
Take career pathways. A pathway isn't just "this role leads to that role." It's a set of rules: which skills are required versus nice-to-have, what counts as equivalent experience, how much tenure a move needs, which certifications are mandatory, who has to sign off. Those are exactly the kinds of rules SIGN captures and canon governs. The map shows the route exists; canon tells you the speed limit, the closures, and which turns are allowed.
Workforce is our first implementation, but the same model works well beyond it: compliance rules, risk thresholds, supply chain and manufacturing constraints. Anywhere an organization has knowledge and rules that people and agents both need to apply the same way.
What should responsible AI decision-making look like?
You should be able to explain the decision. What facts did the agent use? Which rule did it apply? Where did that rule come from, and was it the current version? Did an exception apply? Was the agent even allowed to take that action?
And there should still be a person at the wheel. I'm not trying to take people out of decisions that affect someone's career or livelihood. The point is that the person, the agent, and the automation are all working from the same governed knowledge and the same rules, so "a human reviewed it" actually means something, because the human and the machine were looking at the same source. SIGN doesn't do the reasoning for you. It sets out how the reasoning gets applied, consistently, with the rules attached. That matters most when a decision touches someone's job, pay, access, or opportunity.
What could this mean for an employee using an AI career tool?
More clarity, and a lot less mystery. An employee shouldn't get a black-box "you're a 72 percent match" with nothing behind it.
They should see the real picture: here are the skills you already have, here are the ones this role needs, this requirement is non-negotiable, this experience counts as equivalent, and here's the learning that closes the gap. If you're not ready today, the tool should show you how to get ready.
The bigger deal is that employees are working from the same governed knowledge as HR and leadership. How does this company actually move people forward? What makes someone eligible, what's the process, how do I raise my hand? That's usually the most guarded, word-of-mouth information in a company. When it lives in canon, everyone can see it. That's the difference between AI that screens people out and AI that helps them move forward.
What are most organizations getting wrong about AI agents?
They think a smarter model will fix messy knowledge. It won't. If your policies contradict each other, your job descriptions are stale, or the real rules live in five people's heads, a more powerful model just reaches those problems faster and states them more confidently.
Most of what companies call an AI problem is a knowledge problem. The model was never the missing piece, a governed source of truth was. Before you hand an agent more autonomy, you have to be able to say what's true, who owns it, how it stays current, and which decisions still belong to people. That's the work canon and SIGN are for: getting your knowledge into one governed place before you point agents at it. Skip that step and the agent will happily automate your worst inconsistencies.
What should leaders do before moving AI agents into production?
Don't start with "how many agents can we deploy." Start with one process.
Pick something real, define its inputs and outputs, map the information and rules that actually govern it, and write that down as canon in SIGN. Then make that one governed source available to your people and your agents together, and measure two things: did productivity go up, and did your AI spend go down.
Better yet, pick an area that's already burning a lot of AI cost. When knowledge is scattered across long documents, agents have to be fed enormous amounts of context on every request, and you pay for all of it. When the rules and relationships are written directly in canon, an agent needs far less context to reach the same answer. We see roughly a 50 to 60 percent cut in the tokens agents spend on the knowledge layer. But cost is just the easiest thing to measure. The real prize is that the answers get consistent and you can trace every one back to a rule. Prove it on one process, then scale it.
What will separate the organizations that use AI well from those that do not?
Governed use of enterprise knowledge, across their people and their agents alike.
The organizations that win won't be the ones with the most agents or the flashiest model. They'll be the ones where everyone and everything works from the same source of truth, under the same rules. They'll know what skills their people have, what their roles require, how the work is changing, which rules govern a decision, and who has the authority to make it.
For the workforce, that's what Skills Intelligence gives you, and it's what drives the decisions that actually matter: internal mobility, where to invest in AI, where L&D spend pays off, skills-based pay, talent readiness. SIGN and canon are the layer underneath, and that same layer extends to the rest of the business. That's the whole idea behind Career Highways: help people grow, help organizations thrive, and make sure the technology opens more pathways, not fewer.
Sara Montes de Oca is the Editor in Chief of TechEchelon. Previously a correspondent and producer in Washington, D.C., covering business, finance, and politics.
More from Sara →Written for readers who already know the basics — markets, AI, and the policy decisions that shape both.