Start With the Work People Actually Do

Before introducing AI into a workflow, Terry believes organizations have to get underneath the official process and understand what is really happening.

The process map, he says, is often an aspirational version of reality. The real work may involve spreadsheets, shadow trackers, informal handoffs, and individual employees who have accumulated knowledge the formal system does not capture. That invisible work is precisely what leaders need to find.

“Look for things around where decisions actually get made, where handoffs fail, where the knowledge truly lives. If you just automate what’s documented, you’re basically automating something that doesn’t exist.”

Terry believes the opportunity for AI often sits in the gap between the formal system and what people have actually learned to do to make that system work.

Employees Should Be Involved from the Onset

When should employees become part of the AI design process? For Terry, the answer is immediate.

“From day one. That’s my always answer.”

He believes users should be involved before selecting a technology because they understand where the pain is—and what a successful outcome would look like.

“They are going to know where the pain is and what done truly looks like.”

That involvement also matters to trust. People are far more likely to believe in a system when they understand how it was built and have helped shape how it changes their work. It means allowing them to influence how the work changes and what their role becomes within it.

“The goal is co-design.”

If employees encounter a solution for the first time during testing, he says, organizations are no longer asking for design input. They are asking for a signature.

“If people are just seeing it for the first time at testing, then it’s not design input. It’s just validation.”

When people genuinely participate in building something, Terry sees a very different response to change. Resistance begins to fall because employees have a stake in making the new way of working succeed.

AI arguably makes that more important because the technology can influence decisions, outputs, and workflows at a deeper level than many earlier workplace tools. Terry believes human-centered design was always important, but AI has magnified its importance because of its potential to reshape how organizations think about productivity, output, and value. 

Workarounds are Design Feedback

Companies often look at adoption rates to determine whether a new system is working. Terry believes the more revealing evidence may be found in what employees do around the system.

If people continually recheck AI output, lose information, revert to an old tool, or create a new spreadsheet to compensate for what the system cannot do, those behaviors should not be dismissed as resistance. They are telling leaders something.

“Watch for where people are building their own workarounds, or where there’s an old tool that they keep going back to, or somebody makes a shadow tracker, or there’s a tracker for the trackers.”

Terry is particularly skeptical of treating widespread difficulty as a training issue.

“When a team of 100 struggles with the system, it’s a design problem, not a training problem.”

That shift in perspective is useful. Instead of asking why employees will not adopt the new way of working, leaders can ask what employees’ behavior is revealing about the design itself.

Clear Human Ownership

Terry sees people acting more like operators or orchestrators—staying involved in important decisions while AI accelerates analysis or handles work that can be appropriately automated. 

There needs to be a clear human owner. 

“AI needs to show its work, and people need to be able to look back and validate that.” 

And you can’t assign accountability to the technology. 

“If you can’t name who’s accountable for it without saying the model, then you need to shift where you’ve got that built.” 

Terry raises another issue that becomes particularly important as organizations automate more work: protecting the experiences through which people develop expertise.  

“You also have to protect the entry-level work, or you won’t have people to be experienced later on.” 

That complicates the simple idea that organizations should automate every task they can. Some work may be inefficient in isolation but still serve an important developmental purpose. If AI removes those experiences entirely, companies may eventually discover that they have also removed part of the path through which judgment gets built. 

Change Management Has to Change, Too

Terry is less interested in whether the team completed the change plan or updated the tracker than in whether the organization actually achieved the intended change.

“Did we move the needle on the real thing, not just the metrics around the thing?”

In other words, activity around transformation is not the same as transformation.

AI is arriving at a moment when the traditional cadence of organizational change is already being challenged.

Terry describes the older model as one in which teams investigate a problem, collect information, go away to analyze it, develop an answer, and then work sequentially through implementation. He does not think that model is keeping pace anymore.

“The traditional model of go investigate, get a bunch of information, take that back into your hole and look at it for a while and then come up with something amazing and then step by step work through it—it’s just not keeping up with the pace of business today.”

That applies not only to change management, he says, but also to design, optimization and other disciplines involved in transformation. His answer is not to discard change management. It is to make it much more adaptive.

“Every single step along the way, it’s got to shift. It’s got to be more nimble. It’s got to be agile.”

And again, Terry brings the discussion back to outcomes.

“Are we truly moving the needle in the way that we want? Not: did we get the change plan out? Did we make the spreadsheet? Are we changing hearts and minds, and is this really working?”

Looking Ahead: Three Priorities for Leaders

Find the invisible work before designing the AI.
The documented process is only a starting point. Leaders need to understand where employees improvise, where knowledge actually resides, where handoffs break down, and where people have created workarounds to keep the business moving. Those gaps may be more important to AI design than the formal workflow itself.

Treat employees as co-designers, not end-stage validators.
Bring users into the process before choosing the tool. Let them shape how the work changes and identify what successful work should look like. Testing a completed solution and asking people whether they like it is not human-centered design.

Measure the outcome, not the activity around it.
Log-ins, training completion, and adoption statistics can tell leaders something, but not whether the work has actually improved. More important measures are downstream: fewer errors, less rework, better decisions, easier work, and demonstrable movement on the business outcome the technology was meant to affect.

The Bottom Line

AI may be new technology, but Terry’s argument is fundamentally about how organizations design work. 

If leaders begin with the technology and try to fit people around it, they risk automating an idealized version of the process while missing the work that actually keeps the organization functioning. If employees are brought in only at the end, the organization loses both their knowledge and the chance to build trust through participation. 

The alternative is to start closer to the ground: understand the work as it really happens, involve the people who do it, keep human ownership of consequential decisions, and judge the result by whether the work itself gets better. 

And as AI continues to change, that process cannot be treated as a one-time transformation. Change management has to become part of how the organization continually learns, redesigns, and adjusts “while the engine is still running.”