Work and Workforce: Build It on Purpose, or Lose It by Default

The systems that built your best people won’t survive AI unless you protect them deliberately

Part 2 of the AI Absorption Framework series

---

Look at the people in your organization whose judgment you trust most, the ones you hand the hard calls to. Trace their careers back and you’ll usually find the same path. They started as analysts or associates doing work that felt routine. They sat in meetings taking notes, wrote the first draft nobody saw, and received feedback that stung. Over hundreds of those reps, they learned to tell good work from bad, read a room, and push back on a senior person’s bad idea without torching the relationship.

That’s the apprenticeship model: slow, imperfect, and still the most reliable way we’ve ever found to turn a novice into an expert.

Right now we’re in a Goldilocks moment. Most organizations are full of people who developed that way, with enough expertise to carry them through the next decade. But many of the activities that built that expertise are exactly the ones we’re handing to AI. The question isn’t how much of the work AI can do. It’s where the next generation of judgment comes from when the work that built it is gone.

Organizations are spending enormous energy figuring out how AI can perform work. They’re spending far less figuring out how people become capable of performing it in the first place. That’s the tension underneath this pillar.

Short-Term Gain for Long-Term Pain

Work & Workforce is the hardest pillar to get right. It’s where AI absorption either creates lasting value or weakens the very capabilities it depends on.

The temptation is to treat AI implementation as a decomposition exercise: break work into tasks, automate the routine ones, stop backfilling junior roles, and harvest the efficiency. It looks like value immediately: more output, fewer people, a cleaner budget. But it trades a number you can see this year for a capability you won’t miss until you reach for it and it’s gone.

The challenge runs deeper than workforce reduction. The same logic reshapes how work is designed, how expertise is developed, how people experience growth, and how organizations handle the people whose work changes most. Those four questions sit at the heart of the pillar.

Work Design & Task Alignment: can you take the work apart without breaking what makes it valuable?

The first move in most AI redesigns is to break a role into tasks and automate the routine ones. For a lot of work, it’s the right move. But not every role can be understood as a collection of tasks. In many, the value comes from one person integrating information, context, and decisions across those tasks.

An account manager, for example, may spend time in meetings, responding to emails, preparing updates, and coordinating across teams. On paper, each activity looks highly automatable. But much of the real value comes from understanding the customer, sensing problems before they’re raised, and building trust. Break the role into a tidy list of tasks and you can automate every step while destroying the thing that made it effective.

People aren’t fractional, either. Some of the highest-value work comes from people who have worked together long enough to build trust, shared context, and a common way of operating. Those relationships generate value that never appears on a process map. You can redesign the workflow, but you can’t assume you’ll recreate that accumulated context by assembling strangers around a newly automated process.

A test before you decompose a role: if this work were split across several people, would the output still hold up, or does the quality depend on one person seeing how the pieces inform each other? When it holds up, automate freely. When it doesn’t, the value is in the integration. The goal isn’t to carve the role into smaller pieces. It’s to remove the routine work so people can spend more time applying the judgment, context, and coordination that create the value in the first place.

And once work is restructured around AI, the effects don’t stop at productivity. They shape how people develop expertise, how careers progress, and how organizations build capability over time.

Workforce Capability & Skills: if AI does the work that trains people, how do people get trained?

The skills that matter most as AI takes on more of the output –  judgment, discernment, taste, and knowing when a confident-sounding answer is wrong – aren’t skills you hire off the shelf. They’re built over years through experience and working alongside other people. A junior person usually doesn’t walk in with them.

Here’s the assumption I see taking hold: because AI can do the junior work, organizations decide they need fewer juniors and less investment in developing them. AI doesn’t close the path that builds expertise. It convinces organizations they need less of it. The logic is simple: if AI does the junior-level work, why hire junior people? Organizations hire one level up and assume someone else will produce the talent pipeline. The entry bar quietly moves higher.

The problem shows up when every company makes that assumption at once. If no one trains juniors, no one produces the experienced people everyone now expects to hire. You can’t all skip the first rung and still expect a full ladder.

It’s already starting. Stanford’s Canaries in the Coal Mine study found that since late 2022, employment of early-career workers (22-25) in the most AI-exposed jobs fell about 11%, while their peers in the least-exposed jobs grew about 10%. Experienced workers showed no such gap. The bottom rung is being pulled up, and no one has a good answer for where the next generation of experienced people will come from.

So what do you do instead of cutting the entry-level roles that have always built expertise? Preserve the experiences that develop judgment. The developmental value was never only in producing the first draft. It was in producing it, evaluating it, improving it, and understanding why one approach beat another. AI can accelerate the work, but it doesn’t remove the need for those learning reps.

Put junior employees on the AI’s output. Ask them to challenge it, find what doesn’t pass the sniff test, explain why, and improve it. Keep enough hands-on work that they still understand what it takes to create the thing, not just review it. The goal isn’t to protect outdated tasks. It’s to preserve the experiences that build judgment.

What AI won’t create on its own is connection. Feedback, coaching, observation, and watching experienced colleagues navigate hard decisions still matter. Those experiences have to be designed intentionally.

The apprenticeship doesn’t have to die with AI. But it survives on purpose, not by accident.

Workforce Adaptability & Mobility: how do people know if they’re getting somewhere?

Even if organizations preserve the experiences that build expertise, a second challenge emerges: how people experience growth.

For decades, career progression was easy to understand. Learn the job, take on more responsibility, earn a promotion, move up. AI doesn’t eliminate the human desire for progress, but it scrambles many of the signals people have always used to measure it.

One popular response is the skills-based organization: an internal marketplace where people move between opportunities as business needs change and traditional career paths matter less. The idea is appealing. The problem is that it gets people wrong.

Most people don’t identify themselves by their skills. They identify by their role, profession, or title. They don’t see themselves as a collection of capabilities for rent. They want mastery. They want a clear area of expertise. They want to be known as the person others turn to for a particular problem.

Levels never only coordinated work. They also provided momentum. They gave people visible evidence that they were growing, taking on more responsibility, making a bigger contribution. Remove those signals in the name of agility and you don’t automatically create a more adaptable organization. You risk creating talented people who no longer feel they’re getting anywhere.

The challenge isn’t flattening the organization into a marketplace of skills. Nor is it bolting on a parallel expert track as a substitute for management, which often becomes nothing more than a second ladder people compare against the first. The challenge is helping people see a future for themselves when the path forward is no longer as obvious as it once was.

In a world where judgment matters more than output, advancement can’t be measured only by title, tenure, or headcount managed. Organizations may need to recognize progress differently: through the decisions someone shapes, the complexity they navigate, their influence on outcomes, and the degree to which others rely on their judgment.

A title was always a proxy for “this person matters more now.” If that signal weakens, organizations have to find other ways to recognize expertise, contribution, and growth. That recognition has to be visible, and it has to answer the question every employee eventually asks: am I progressing?

Few organizations have cracked this. The ones that do won’t necessarily have the cleverest org chart. They’ll be the ones that give people a genuine sense of growth and momentum, even when advancement no longer follows a familiar script.

Reskilling & Redeployment: reskill into what, and redeploy to where?

The previous section was about the people you keep and grow. This one is about the people whose work AI may fundamentally change, a harder conversation and one many leaders would rather avoid.

“We’ll reskill and redeploy our people” is often the first promise organizations make during a transformation. It sounds responsible, optimistic, and people-centered. Sometimes it’s true. But too often it’s treated as a strategy before anyone has checked whether there are meaningful opportunities for people to move into.

Reskilling only works if there’s something real to reskill toward, and redeployment only works if there’s somewhere to go. In an AI transition, the number of destination roles may be smaller than the number of people displaced. Sometimes those roles don’t exist yet at all. You can’t retrain someone into a role no one can name.

Announcing a reskilling program isn’t the hard part. The hard part is the math most leaders skip. What work will AI absorb? What new roles will emerge? How many seats actually exist? If it’s 50 people and 30 seats, you don’t have a reskilling plan. You have a reduction you haven’t named. Knowing that early is the difference between a humane transition and a chaotic one.

The real challenge isn’t training. It’s honesty. People handle difficult news far better than uncertainty dressed up as optimism. If there are genuine opportunities, invest in them aggressively. If there are only a few, be transparent about that too. Credibility comes from telling the truth about the size of the destination, not from overpromising the journey.

Where real destinations exist, they’ll often be in the work that remains distinctly human: judgment, relationships, trust, and navigating ambiguity. Build toward those capabilities intentionally, and build the learning muscle to keep doing it, because the targets will keep moving.

And where there honestly isn’t a path for someone, say so. Dressing up a reduction as redeployment doesn’t build trust. It destroys it, not only with the people leaving, but with everyone who stays and watches how the transition unfolds.

Build it on purpose, or lose it by default

Go back to the people you trust most with the hard calls. Every one of them is the product of a system: years of experience, feedback, coaching, and leaders who invested in their development. Most organizations funded the career paths and the training programs. What they never had to do was consciously protect the underlying experiences that made those investments work, because the work that built people was simply there.

That’s the real work here, and it isn’t a technology decision. Every choice about what to automate is also a choice about what people will learn and where they’ll create value. The hard part is deciding what human value is worth protecting, the judgment, relationships, and hard-won expertise your organization depends on, and then building it deliberately, because it will no longer build itself.

The organizations that come through this well won’t be the ones that automated the fastest. They’ll be the ones that understood what created value in the first place, and were deliberate about what they couldn’t afford to lose.

Next in this series: Operating Model, how structure, decision rights, and roles have to change for a human-and-AI organization to actually work. If you’d rather talk through your own organization, reach me at lisa.taylor@ taylorsteinconsulting.com.2