AI Adoption is an Operational Transformation
Building AI into your operating model is what separates real transformation from shelfware.
They buy the platform, flip on access, run a training session, and wait for the organization to get more efficient. That's not how it works. AI is an operational capability. It has to fit within the organization's larger operational strategy. Its value depends on how it is incorporated into the way work happens.
I made this argument earlier this year in Inserting AI into Operational Processes: organizations need to build AI into workflows, not bolt AI tools onto the tech stack. The more transformation work I do, the more I think this is the whole ballgame. AI doesn't replace the operating system of the business. It becomes part of it, or it becomes shelfware.
I watched this play out with an organization I was working with on performance management. I'd just finished a work management implementation and was connecting it to their broader operating model when leadership started talking about consolidating platforms to cut costs. One leader didn't like the system, hadn't really used it, and decided to bring in an AI strategist to build the organization's performance management structure instead. I got pulled off the project.
A few weeks later, the organization was supposed to demonstrate this new AI-powered approach to managing performance. There was nothing there.
They tried pulling information into the AI and didn't have enough structured input to work with. They tried getting people to enter information in real time, and that fell apart, too.
The AI wasn't the problem.
Nobody had built anything for it to operate against. There was no structure, no standardized input, no reporting environment, and no clear logic for what the organization was trying to accomplish. The strategist had done some scaffolding, but the underlying mess was bigger than anyone had accounted for.
That's the failure mode I keep running into. Organizations expect AI to intuit their business logic because they bought a seat license.
It doesn't matter whether you're working with ChatGPT, Claude, or anything else. You still have to build the environment around the work. Something has to define where the work lives, what information feeds the system, how it's standardized across contributors, what gets measured, and what the output is supposed to look like.
Point AI at a spreadsheet, and the spreadsheet needs structure. Point it at a work management system, and the fields, ownership, workflows, and reporting need structure. Ask it for real analysis, and the underlying data has to support that analysis.
None of that shows up because you purchased a platform.
Before I deploy anything, I want hard answers to some basic questions. What are we trying to accomplish? What are we trying to produce? How will we know it worked? What's the risk? Is the team ready? What are the actual use cases?
Those questions get skipped because everyone is racing to get something into production. That's how you end up with an organization that has technically deployed AI but has no real adoption strategy.
McKinsey's 2025 research backs this up. Eighty-eight percent of organizations reported using AI in at least one business function, but only 7 percent said AI had been fully scaled across the organization. McKinsey also found that workflow redesign was one of the organizational changes most strongly associated with capturing value from generative AI.
Deloitte's 2026 research lands in the same place. Eighty-four percent of organizations have not redesigned jobs or workflows around AI, even as access continues to expand. Using AI is not the same thing as changing how work gets done.
Team readiness gets underestimated just as badly.
People can be excited about AI and scared of it at the same time. Some think it's coming for their job. Some don't trust it. Some have absorbed bad information about what it can and can't do. Some will experiment in front of the whole team. Others won't touch it unless no one's watching.
I've had engagements where step one wasn't training on the tool. It was giving people a room to say what they were actually worried about, almost like a group therapy session. Then came the one-on-ones and surveys that helped us understand where people really stood.
That's change management. It's not something that happens after the technology lands. It's part of the implementation.
Training works the same way. It has to be progressive. People need to understand the structure, work inside it, get role-specific coaching, and have reinforcement while the system is live. You don't get behavior change because you handed someone a login.
I'm not interested in using AI because we can. The first question is where it buys us the most leverage.
Where are people burning hours on work that doesn't require judgment? Where are steps repeating? Where is information being moved manually between systems? Where is reporting eating time that should be going toward analysis?
This is where time studies earn their keep.
If someone's spending two hours a week feeding a tracker or building the same report by hand, that's a real use case. If we can remove that work, the recovered time can go toward execution or toward giving people their life back. It doesn't automatically need to become more work.
And not everything should be automated.
Automation belongs on the parts of a process that are repetitive, administrative, or rules-based. Judgment, approval, authority, and consequential decisions stay with people. The organization has to be deliberate about where responsibility sits.
Every AI-touched process needs an owner. Someone has to understand how the process works, what the AI is doing inside it, what the expected outcome is, and what happens when it breaks.
I wouldn't let AI maintain AI.
A human still owns the machine and the process around it. This is consistent with the NIST AI Risk Management Framework and the OECD AI Principles, both of which emphasize human oversight, accountability, and ongoing risk management throughout the AI lifecycle.
The deeper AI gets embedded, the more the operating model underneath it matters.
Who owns the data? Who maintains the process? Who validates the output? Who can approve an action? What happens when the AI is wrong? What happens when the process changes? What should it have access to, and what should it never touch?
These are operational questions.
Standardization matters for the same reason. If five people enter the same field five different ways, the system won't interpret it consistently. If every team defines a KPI differently, automated reporting has no reliable foundation.
The AI can be capable. The environment underneath it still has to be built.
None of this should be measured by how many people have access, how many prompts they use, or how many tools the organization has purchased.
The question is whether the organization is operating better because AI was introduced.
Are decisions faster? Is reporting more reliable? Is administrative work going down? Are errors decreasing? Are bottlenecks clearing? Is the quality of the work improving? Are people spending more of their time on work that actually requires their judgment?
Those are the measures that matter.
Deloitte makes the same distinction between adoption and transformation. People can use AI without changing how they work. Transformation means the jobs and workflows actually change.
That's why I don't think this belongs to the technology function alone.
Technology owns a piece of it. Operations owns a significant piece. Business process owners own a significant piece. Leadership owns the decisions and the risk. And the people doing the work have to be in the room because they understand what actually happens inside the process.
AI adoption crosses all of those lines. Treating it as an IT initiative is how you end up with the empty performance meeting I described above.
I'm not anti-AI. I think the upside is enormous.
But potential isn't implementation.
The organizations that get real value from AI will be the ones willing to do the operational work around it. They'll define real use cases, build the environment, standardize inputs, redesign workflows where it makes sense, prepare their people, establish ownership and governance, and measure outcomes.
And they'll give it time.
That means having a strong operational plan, strong use cases, a real communication plan, and change management that runs through the implementation instead of stopping at the kickoff.
AI doesn't eliminate the need for any of that. It makes it more important.
The technology will keep changing. The operating principles don't have to.
AI belongs inside the operating system of the organization, wherever it creates real leverage. It was never supposed to become the operating system itself.
That's the difference between adopting AI and transforming how an organization works.
