TL;DR: Organizations across industries are rushing to deploy AI tools that promise dramatic productivity gains. The risk is rarely that the technology fails. It is that it works, and in doing so quietly erodes the human capabilities organizations need to evaluate what it produces. As Ryan Simmons, Vice President of Human Resources at Colorado College, argues, the real challenge is designing work so people don’t outsource their thinking to automation and lose the capabilities needed to question its answers.

Ryan Simmons has a simple way of explaining one of the most important risks in AI adoption.

“If your goal is to lift weights, it makes tons of sense to bring a forklift to the gym. You just have that forklift lift weights all day long. You’re going to get so much weight lifted. But if your goal is to get strong, that tool actually undermines what you want.”

The metaphor cuts to the heart of something many organizations are beginning to confront. The same tools that accelerate work can quietly erode the human capabilities that make work meaningful and trustworthy. That tension is the central challenge of AI transformation, and it is one that rarely appears in conversations about software deployment.

Ryan is Vice President of Human Resources at Colorado College, an institution that has been educating students for more than a century and a half. Before that, he spent seventeen years at the Children’s Hospital of Colorado, an environment where professional development pathways had been built carefully over many years. That experience gives him an unusually clear view of what happens when technology changes faster than the systems around it.

“You can install software overnight. But the human adaptation to that change takes the amount of time it takes a human to change.”

That distinction shapes everything else in Ryan’s thinking about AI, and it is the right starting point for any organization navigating this moment.

Why AI Transformation Is Not Primarily a Technology Problem

When organizations begin exploring artificial intelligence, the conversation typically starts with the tools. Leaders want to know which systems are most capable, where they can be applied, and how quickly they might generate measurable gains. The instinct is understandable. Over the past several decades, most waves of enterprise technology have followed a familiar pattern: new systems are introduced, processes become more efficient, and organizations adapt around those improvements.

Artificial intelligence appears to fit naturally into that pattern. Early use cases reinforce the idea that it is primarily a productivity tool: drafting documents, summarizing reports, and searching internal knowledge bases without manually navigating folders. For organizations under pressure to move faster or reduce costs, the appeal is obvious.

But that framing assumes the underlying structure of work remains largely intact. In reality, AI often does more than simply speed up existing processes. Once automation enters a workflow, it begins to alter which activities still matter and which capabilities employees are expected to develop. Those shifts are easy to miss at first because the early signals look positive: work moves faster, bottlenecks shrink, and teams produce more output with less friction. The deeper implications take longer to surface.

Ryan sees this clearly in the context of higher education, where the pressures are particularly visible. Colleges are confronting rising costs, shifting demographics, and growing scrutiny about their value. Under those conditions, the promise of automation becomes especially attractive, but Ryan believes the familiar language of efficiency may no longer be the right frame. “Do more with less may not even be the right frame. It may need to be ‘do different with different.’”

“Do more with less” assumes the work itself remains fundamentally the same. “Do different with different” implies that once tools change what is possible, organizations may need to reconsider the work entirely: what activities still deserve human attention, which should be automated, and what capabilities employees must continue to develop.

The distinction matters because it changes the nature of the decision. If the goal is simply efficiency, leaders ask which tasks can be automated and which processes can run faster. If the goal is to “do different with different,” they must ask harder questions: which work is actually still worth doing, which capabilities must the organization continue to build, and what happens to the people whose roles are reshaped in the process. Some of what organizations have been doing may simply no longer be necessary. Doing more of it faster is not progress. At that point, the challenge of AI adoption moves beyond the technology itself and becomes a question of work design.

When Faster Output Comes at the Cost of Capability

The most important risk in AI adoption is not that the tools will fail. It is that they will work so well that organizations stop developing the human capabilities required to evaluate their outputs.

Many tasks that professionals perform early in their careers appear routine on the surface: drafting documents, conducting research, analyzing data, and assembling arguments from different sources. Those activities are not merely steps toward producing an output. They are the mechanism through which individuals develop the judgment required to evaluate complex situations later in their careers. Through repetition, people learn how to assess the quality of information, distinguish credible arguments from weak ones, and identify the subtle details that change how a decision should be made.

When AI produces the output instantly, the result may still look correct. What disappears is the practice that builds the ability to judge whether that result should be trusted. Ryan illustrates this point through writing. “If you never learn to be a good writer, you’re never going to learn to second-guess the writing that comes from AI.”

Technology may increase output while quietly weakening the human capability required to evaluate that output. In environments where quality depends on professional judgment, such as education, healthcare, finance, and government, that risk is difficult to ignore.

There is also a subtler mechanism at work: the nature of trust. Ryan draws a useful comparison with spreadsheets. When an Excel formula generates an answer that makes no sense, most people’s immediate reaction is to check their formula. They assume the error is theirs. With AI, the dynamic flips. Because the tool communicates in natural language and sounds confident, people are more inclined to accept what it says. The answer arrives in a form that feels authoritative, and the internal check, the instinctive “did I make a mistake?”, does not automatically trigger.

“Because it talks like a human, I think we sometimes just go, ‘Oh well, it’s smarter than I am. I’ll just trust it.’ And that’s just using the tool the wrong way.”

That instinct toward deference is the real risk. It is not that employees are careless. It is that the tool’s conversational surface actively suppresses the skeptical reflex that good professional judgment depends on. And if organizations do not design work specifically to counteract this tendency, passive trust becomes the default.

This is also a structural risk for talent pipelines. Most professions rely on a gradual progression of responsibility. Early-career employees begin with tasks that require attention to detail. Over time they move into roles that demand independent judgment. Those early tasks function as an apprenticeship: the work may seem routine, but it provides the training ground through which expertise develops.

If automation removes too many of those early stages, the progression becomes harder to sustain. As Ryan puts it, “You’re going to be missing rungs from your ladder.”

Organizations may not notice immediately. Initially, experienced professionals remain in place, and efficiency appears to improve. The effects appear later, when the pipeline of talent behind them is thinner than expected. New employees know how to operate the tools but have had fewer opportunities to develop the expertise that allows them to question the results those tools produce. That is why decisions about automation are also decisions about how capability develops inside an organization. Remove too many of the learning steps, and the long-term strength of the workforce begins to weaken.

How Automation Reshapes the Value of Skills

Every major wave of automation reshapes which skills become economically valuable. When machines begin performing certain tasks, the capabilities associated with those tasks become less scarce. At the same time, other skills, typically those involving judgment, interpretation, or coordination, become more valuable.

Ryan sees this dynamic already emerging, and frames it as a question of which layer of work the automation is reaching. “If you’re in a job where your lower-level skills are being automated, then probably over time your compensation might increase because your productivity is increasing. You’re going to be using higher-level skills.”

In that scenario, automation removes routine tasks and frees employees to focus on more complex activities. Work evolves upward. But the same technology can produce the opposite outcome when it replaces the higher-level elements of a role. “If it’s the higher-level skills that are being automated, then it’s the opposite case. There’ll be more people competing for the same jobs, which will drive the price of that job down.”

Automation does not simply eliminate work. It redistributes value across different kinds of work. And critically, neither the employee nor the employer always gets to choose which scenario applies: it depends on what the tool can actually do. If a system is capable of automating the judgment-intensive parts of a role, that is what will be automated, regardless of what anyone would have preferred.

This introduces a deeper question about identity. Ryan raises what happens when someone who has built their professional identity around a particular skill set sees that skill being automated away. What if the work that remains is precisely the work that person never found meaningful? What if it is someone whose entire career has been defined by their technical expertise, now asked to perform coordination or documentation tasks because their core craft has been absorbed by a machine?

Ryan makes this concrete in the context of IT. If a developer has built their professional identity around coding, and tools like those emerging from AI labs can now write and review code in a fraction of the time, the identity disruption is immediate and personal. The question is not just “what do I do now?” but “who am I if I’m not the person who does this?”

This means that workforce adaptation is not only a skills challenge, but an identity challenge. People who experience automation primarily as loss are far more likely to resist than to experiment. The organizations that navigate this well are those that help people find meaning and contribution in what remains, rather than simply presenting automation as progress and expecting adjustment to follow.

And there is still the developmental problem Ryan identified. If the lower rungs of the ladder disappear, people cannot climb to the higher-level work that automation supposedly frees them for. You cannot ask someone to exercise judgment they were never given the chance to develop. Automation decisions are therefore not only efficiency decisions. They reshape skill value, compensation pressure, future role design, and the pipeline through which the next generation of expertise is built.

Designing Work So AI Strengthens Human Judgment

If the risks of automation lie in how work is structured, the response must begin there as well. AI does not determine how people use it. The surrounding system, including how teams collaborate, responsibilities are distributed, and decisions are evaluated, shapes behavior far more than the tool itself.

Ryan began applying this principle directly inside his HR organization. When turnover created an opening to rebuild how a group of coordinators worked, he did not simply rehire for the same roles. Instead, the team redesigned the physical and collaborative structure from the ground up: desks were rearranged so team members faced each other, hiring decisions were made collectively, and responsibilities were distributed through cross-training rather than rigid specialization. The goal was to create an environment where questions naturally became collaborative discussions rather than individual transactions.

Then came the AI component. The team had a shared drive containing procedures and policies, but new staff rarely knew where to find the right document. Rather than pointing people at a folder structure, the team built an AI agent capable of querying that repository and returning a proposed answer to any policy question. The tool reduced the friction of retrieval considerably.

But Ryan was deliberate about how it would be framed. “They don’t trust it, and I don’t want them to trust it.”

The reasoning is practical, not philosophical. AI outputs hallucinate. Documents become outdated. The data set is never complete. In a policy context, acting on a wrong or stale answer has real consequences for the employee on the other side of the interaction. So the AI was positioned explicitly as a research assistant, not an authority. Its job was to surface a candidate answer quickly. The team’s job was to evaluate that answer together.

“Looking around in a folder for a piece of information does not add value. But saying, ‘I’ve got this information. Does it make sense and will it help this employee?’ That adds value.”

What Ryan describes is not just a workflow change. It is a deliberate redirection of where human effort goes. The low-value work of searching and retrieving is handed to the machine. The high-value work of interpreting, verifying, and applying is kept with the team. And because that interpretive work happens through conversation rather than in isolation, it compounds: coordinators develop judgment by watching each other reason through uncertainty, not just by encountering problems on their own.

Ryan notes something important about how that critical thinking actually formed in practice. He had given the team a training session explaining that AI outputs could be wrong and should not simply be trusted. That training was not, by itself, what changed their behavior. “I don’t think it’s training that somebody did six months ago that’s causing them to think differently. I think it’s more their experience trying to get the work done.”

They used the tool. They encountered an answer that made no sense. They talked to a colleague about it. They saw it was wrong. And from that point forward, they were incrementally less likely to accept the next answer at face value. The skepticism was built through practice, not instruction. This is an important lesson for organizations trying to cultivate healthy AI habits across their workforce: orientation sessions and policy statements establish the expectation, but it is the design of the work itself that determines whether that expectation becomes ingrained behavior.

The result is visible. Ryan describes sitting just outside his team’s workspace and listening to the quality of problem-solving that now happens in the room. It is a different kind of conversation from what he heard before: more deliberate, more peer-to-peer, more focused on meaning and consequence than on retrieval and compliance. The technology did not produce that. The system around the technology did.

Why This Requires HR and Technology Leaders to Work Together

Ryan sees this moment as something genuinely new: “This is the most human-technological combination situation we’ve ever been in.”

What he means is that the boundary between a technology decision and a human decision has become almost impossible to draw. Choosing which AI tools to adopt is not a neutral procurement exercise. It determines which capabilities the organization will stop developing, which roles will need to be redesigned, and which employees will experience their professional identity as suddenly uncertain. Those are HR questions, embedded inside what would historically have been an IT decision.

“We would be remiss to say, ‘The IT guys go off and figure out the technology impact; the HR guys go off and figure out the human impact, and we’ll meet in the middle someday.’”

The interdependency runs in both directions. It is not only that HR needs to be involved in technology decisions. It is also that IT teams are themselves living through the same human adaptation challenges that everyone else is. A developer who has spent a career mastering a craft, and who now sees AI systems capable of performing core parts of that craft, is experiencing the same identity disruption that organizations are trying to help their broader workforces navigate. Treating them purely as technology implementers, rather than as people undergoing change, misses the point.

“They’re humans too. We can just say, ‘You need to learn this tool. How are you learning it? What is that teaching you?’”

Those questions, asked of the people implementing the technology, generate firsthand experience of what it actually feels like to adapt. That experience creates better empathy for what the rest of the organization is going through, and more realistic expectations about how long genuine adaptation takes.

Ryan’s broader point is that transformation cannot be designed by one function and delivered to the others. The decisions about which tools to select, how to implement them, how to redesign roles around them, and how to help people find their footing in the new configuration all intersect in ways that require ongoing joint work. Technology determines what becomes possible. Human systems determine what actually happens. Getting those two things aligned requires conversation that starts early, before commitments are made and before people feel like change is being done to them rather than with them.

The Slower Work That Determines What AI Actually Produces

AI can generate answers rapidly, but the value of those answers still depends on the people interpreting them. When employees treat automated outputs as unquestioned authority, organizations risk transferring judgment to systems that were never designed to carry it. Over time, subtle consequences become visible. Work appears more efficient while the processes that once developed professional expertise quietly disappear. Early-career experiences fade. People learn to operate tools without developing the ability to question them.

Ryan’s experience points to a different outcome. When organizations design work so that AI accelerates access to information while humans remain responsible for interpretation, the technology can strengthen the capabilities of the workforce rather than erode them. The coordinators on his team are developing higher-order problem-solving skills than their predecessors did, not despite the AI tool, but because of how the work around it was designed.

That distinction matters most when it is hardest to see. In the early stages of AI adoption, the efficiency gains are real and the capability erosion is invisible. The pipeline looks fine because experienced professionals are still in place. The habits look fine because the tool is new and people are still figuring out how to use it. The damage, if it comes, shows up years later, when organizations discover that the people behind the experts have fewer rungs to stand on.

Installing software may happen quickly. Adapting the organization around it takes much longer. In a moment when new tools seem to arrive almost overnight, that slower work of redesigning human systems may ultimately determine whether artificial intelligence strengthens organizations, or quietly erodes the expertise they depend on most.

Practical Takeaways: Designing AI Work So People Keep Thinking

As AI tools spread across everyday work, the leadership challenge is no longer just adoption. It is deciding how work should be redesigned so technology improves speed without weakening judgment, capability, or accountability.

1. Why do many AI transformation efforts struggle even when the technology works well?

Because the technology changes faster than the human system around it. Ryan Simmons (VP of HR, Colorado College) draws a distinction between deploying a tool and helping people adapt their judgment, habits, and sense of role around it. Leaders can implement a tool quickly, but they cannot accelerate human adaptation at the same speed. If employees have access to AI but are still unclear how their work, accountability, or value creation should change, the transformation is not complete.

2. What does “designing work around AI” actually mean for HR leaders?

It means deciding which tasks should be automated and which must remain human because they build expertise. Ryan Simmons (VP of HR, Colorado College) argues the more important question is not whether output is faster, but whether work is being redesigned so people spend less time retrieving information and more time interpreting it. For HR leaders, the implication is structural: work design determines whether AI strengthens capability or gradually weakens it.

3. What early warning signs indicate that AI adoption may be weakening employee capability?

One clear signal is when employees begin treating AI output as something to accept rather than something to test. Ryan Simmons (VP of HR, Colorado College) notes that because AI communicates like a human and sounds confident, people are more inclined to defer to it than they would to a formula in a spreadsheet. Teams can become more dependent while becoming less discerning. Pay attention when peer challenge declines, outputs are checked less often, and speed starts to matter more than whether the answer actually makes sense.

4. How does automation change which skills become more valuable inside an organization?

Automation changes value depending on which layer of work it replaces. Ryan Simmons (VP of HR, Colorado College) explains that if lower-level tasks are automated, employees may spend more time on judgment and problem-solving, and compensation may follow. If higher-level work is automated, the opposite risk appears: more people competing for fewer differentiated roles. Automation decisions are not only efficiency decisions. They reshape skill value, compensation pressure, and future role design.

5. What long-term talent risks arise if AI removes too many early-career learning steps?

Organizations may lose the developmental ladder through which expertise is built. Ryan Simmons (VP of HR, Colorado College) warns that when foundational tasks disappear, future capability weakens because employees never build the underlying skills that prepare them for more complex judgment later. Ryan frames it plainly: organizations will be left with missing rungs on their ladder. For senior HR leaders, this is a succession and workforce planning issue, not simply a training question.

6. Why is “do more with less” often the wrong framing for AI-driven transformation?

Because it assumes the work itself remains unchanged. Ryan Simmons (VP of HR, Colorado College) argues the better frame is “do different with different.” Some work may need to be deprioritized, redesigned, or stopped entirely once new tools change what is possible. The strategic leadership implication is that executives must decide which work still deserves human attention and which no longer does, rather than simply asking how to produce more of the same output with fewer resources.

7. How should HR and IT collaborate in an AI transformation?

HR and IT cannot work separately and expect the organization to integrate the change later. Ryan Simmons (VP of HR, Colorado College) says that treating AI tool selection as a pure IT decision ignores the human consequences embedded in that choice. AI affects capability building, workflow design, and employee experience all at once. In addition, IT teams are undergoing the same identity and adaptation challenges as the rest of the workforce. Treating them only as implementers, rather than as participants in the change, means missing a significant source of organizational insight.

8. What role does work design play in ensuring AI strengthens judgment rather than replacing it?

Work design determines whether AI becomes a shortcut or a support tool. Ryan Simmons (VP of HR, Colorado College) illustrates this through his team’s use of an AI policy assistant: the tool retrieves answers quickly, but coordinators are responsible for evaluating whether those answers are current, accurate, and applicable. That structure keeps human effort focused on interpretation rather than retrieval. The key structural question is, where does human effort go once retrieval is automated? If the answer is into higher-order interpretation and peer discussion, the system is working.

9. Why is training alone not enough to build healthy AI habits in a team?

Because critical thinking toward AI is built through practice, not instruction. Ryan Simmons (VP of HR, Colorado College) observed this directly with his coordinators: the training session he ran did not change their behavior. What changed it was the experience of encountering a wrong answer, discussing it with a peer, and becoming incrementally less likely to trust the next output at face value. Ryan’s conclusion is practical: build the conditions for that experience to happen repeatedly, rather than relying on a one-time awareness session.

10. What practical indicator should leaders watch to ensure AI adoption is strengthening capability over time?

Whether AI use is pushing employees into higher-order problem-solving rather than passive answer consumption. Ryan Simmons (VP of HR, Colorado College) points to his own team as a reference: coordinators are now doing more deliberate, peer-led reasoning than their predecessors did, not despite the AI tool but because of how the work around it was designed. Watch for more judgment-based discussion and better use of human discretion. If speed improves while critical thinking declines, the organization is becoming more efficient but less capable.

To hear how the full conversation played out, listen to Ryan’s podcast episode.

If you’d like to be informed about more articles like these and community updates, subscribe to our newsletter.

For more perspectives on how leaders are navigating challenges like this, follow Level Up HR on LinkedIn.