Two years ago, almost every customer conversation about AI started with the same questions: Which model should we use? What can it do? Is it ready for the enterprise?
Today, those discussions have moved on. CIOs are far more interested in how to govern AI, integrate it with existing systems, prepare their workforce and make it part of everyday operations.
The challenge is no longer to prove that AI can deliver value. It's instead about how to embed AI into the business in a way that's secure, scalable and delivers measurable outcomes.
Most Organizations Still Don't Have an AI Baseline
The problem that many organizations are facing is that they are trying to embed and scale AI before they have a clear picture of how AI is already being used inside their own companies.
In reality, AI is often spreading from the bottom up. Employees are writing code with it, analyzing data, automating routine tasks and solving everyday problems long before leadership has a complete picture of what's happening. In fact, for a lot of leadership teams, the first real evidence of how much AI is already in use is the token usage bill that lands at the end of the month. That presents a challenge for enterprise IT leaders. You can't govern what you can't see.
There's also the harder truth that AI disrupts people's empires. In large organizations, influence is still often measured by headcount and AI can make a meaningful share of that headcount unnecessary. That's not always in a leader's interest to surface, which is part of why so much AI usage stays informal and unreported rather than being brought into the open.
Governance works best when it's built around the way people actually work, not the way we think they work. That starts with understanding what's already happening across the business.
Adoption Is Now the Real Challenge
Understanding where AI is already being used is only the starting point. The next step is to turn it into something people rely on every day.
For years, enterprise technology projects were considered successful if they went live. The platform was delivered, the migration was complete and everyone moved on to the next initiative. AI doesn't work like that. Rolling out the tech is only the beginning. The real test is whether people actually use and trust it and ultimately change the way they work because of it.
That's why I think the conversation has shifted so quickly. Technology is improving at an extraordinary pace. Capabilities that felt experimental six months ago are already becoming practical.
The model certainly isn't the bottleneck anymore — it's getting the organization to adapt with it that's proving difficult.
The most common reason a strong AI use case fails to reach production is human. Legacy processes are hard to change. People naturally stick with familiar ways of working. Priorities shift and leadership changes faster than transformation programs can run their course.
A CIO's tenure is often shorter than the multi-year journey an enterprise AI rollout actually requires. When a new leader arrives partway through, the instinct is too often to pause, rescope or restart and momentum that took a year to build disappears in a quarter. Those challenges didn't arrive with AI, AI has simply made them much harder to ignore.
It's also why I believe we're moving beyond the proof-of-concept phase. Over the past two years, organizations have invested heavily in pilots and experiments. Some have delivered genuine value, but many never moved beyond the demonstration stage. I was recently comparing notes with a partner of ours about how many proofs of concept they'd funded in the last year alone: roughly 3,000. The number that made it into production: zero.
The data backs up what I'm seeing in the room. MIT's NANDA initiative found that 95% of generative AI pilots deliver no measurable P&L return and Gartner expects over 40% of agentic AI projects to be canceled by the end of 2027 due to unclear ROI and governance gaps.
This is a pattern I hear echoed across the industry and it's exactly why the questions customers ask us now are more practical. How do we deploy AI across the enterprise? How do we build it into everyday operations? How do we measure the return?
Don't Add Another Layer of Complexity
One mistake I see organizations make is treating AI as something they can simply layer onto the technology they already have. For most enterprises, it just isn't that simple.
Years of legacy systems, applications and integrations have left many businesses with technology estates that are incredibly difficult to untangle, with the people who originally built them long gone. Every new platform connects to everything else, making the environment even more complex.
That's why AI shouldn't be about adding another layer of technology. Instead, it should be about removing friction and providing a governed, trusted platform for AI development. Used well, AI can also help unwind some of that legacy complexity itself — reading, documenting and modernizing old systems at a pace and scale that wasn't possible before, which in time should free up budget currently spent just keeping the lights on.
The best opportunities to apply AI effectively are usually the least glamorous. They're the structured, repetitive tasks that consume time without adding much value. That's where AI can make an immediate difference.
At the same time, organizations need to keep simplifying the technology underneath it. AI is far more effective when it has access to reliable data and systems that are easier to connect, maintain and evolve.
Before thinking about what's next, it's worth asking: what's making the business harder to run today? For many organizations, that's where the biggest opportunity still lies.
AI Is Changing Workforce Expectations
There's another shift that enterprise leaders can't afford to ignore.
The next generation of employees is AI natives. To them, it's not a specialist capability. It's simply how work gets done.
This changes expectations for enterprise IT. CIOs won't just be responsible for deploying AI securely. They'll also need to make sure it's available where people need it and easy enough to become part of everyday work.
I've seen this play out inside our own business. We had a licensed software tool for scheduling engineers and modeling program costs and margins. A group of our younger engineers asked why we were still paying for it when they could just build it themselves. I gave them three weeks. They built it, and now Qody, as we call it internally, runs that part of our business. We switched the old license off.
Organizations that get this right are far more likely to see adoption happen on its own because employees can see the value for themselves. Those that don't risk frustrating some of their most capable people, particularly those who expect AI to be part of the way they work rather than an exception to it.
What Comes Next
If the conversations I'm having with customers are any indication, enterprise AI has already entered a different phase.
The early excitement hasn't disappeared, but it's become much more grounded. Organizations are asking fewer questions about the technology itself and many more about how to make AI work in the context of their own business. That's a sign that expectations are changing and, in my view, that's a positive thing.
We're moving away from treating AI as a standalone initiative and toward treating it like any other business capability. Once that happens, AI stops being something separate from the business. It instead becomes part of how decisions are made, how work gets done and how value is created.
To me, that's the most significant shift of all.