What it takes to level up your org’s AI maturity

Over the past 12 months I’ve engaged with hundreds of CIOs to understand where they are in their AI transformation journeys. Through this process, interesting patterns have emerged.

First, companies generally fall along three levels of the AI maturity curve. A small number of organizations are at the “101 level,” implementing AI copilots, seeing some adoption, and telling their board they’re “doing AI.”

At the wide 201 level, organizations have an outcomes-based strategy, have secured budget and board support, and are zeroing in on use cases that move the business forward, not simply as pilots but with scale in mind.

At the top are those few organizations that are ahead of the pack, enjoying first-mover advantage. They are driving business-impacting AI transformation. And they have a CEO who’s leading the way with an AI-first agenda.

Recent studies confirm that this top-of-the-diamond group is the outlier. AI software firm Metal found that 85% of Fortune 500 companies now mention AI in annual 10-K filings, but for most, AI remains an operational tool, not a revenue driver. A PwC survey of CEOs found only 12% say AI has delivered both cost and revenue benefits.

To help IT leaders advance their organizations along the AI maturity curve, I sat down with two industry practitioners who are helping companies chart their AI transformation and innovation journeys, drawing from their years of experience leading real, practical, business-impacting AI transformation.

Afshean Talasaz, former SVP of strategic projects and innovation and chief technology and data officer at Colonial Pipeline, and Zar Toolan, former chief data and AI officer for Edward Jones, are part of a new Practitioners for Practitioners (P4P) network, which brings experienced technology executives together with CIOs to provide guidance and practical insights, based on their lived experiences, successes, and lessons learned.

Here, Talasaz and Toolan shed light on the nuances that matter, the blind spots many miss, and the questions IT leaders need to ask to move AI from being an operational tool to having a tangible impact on their businesses. Our conversation has been edited for length and clarity.

Dan Roberts: How do CIOs need to shift their mindsets to drive AI transformation?

Afshean Talasaz: Leaders need to have an innovator-operator mindset, which I don’t think was a hard requirement in the past. The innovator mindset thinks about how to evolve and adapt to external pressures — changes in business demands and rapid changes in technology capabilities. The operator mindset creates clarity and stability internally to execute consistently. In the past, you could have split those apart. Today’s business climate requires bringing those two things together.

Zar Toolan: Another dimension of it is how CIOs are thinking for the future. Think of the classic run, grow, transform approach. There’s a lot of run-the-business stuff that you have to keep doing. How can AI help make it operationally more efficient and effective? As for growth, how can you leverage data in new ways to inform new markets, create new ways of growing the business? And then the third is in the transform space. All three should be happening simultaneously, but the transform space is where the new data, the new architecture, and the new way of thinking needs to be anchored. Part of the mindset shift is to get to that third phase: a constant state of transformation, reinvention, and reimagining, while also running the business that earned you the right to transform in the first place.

What’s your advice to CIOs who struggle to do this in practice?

Toolan: I do a lot of the “from-to so-that” modeling: from where we are today, to where we want to go so that the business can do X, Y, Z strategic business outcomes.

Here’s an example from Edward Jones. In 2023 into 2024, we looked at the current states of various parts of our business that had a data element, an insights and analytics element, and a process element. From that, we could articulate the from statements at that time. Some of those statements included “hard truth” elements such as: We do not embrace data as a valuable asset; we’re unreliable and ineffective searching across multiple knowledge hubs; we have limited data and fragmented content curation; we do not have a consistent taxonomy for our data. On the analytics side: Our insights are periodic; they describe what happened and why, not what’s going to happen and when. And then on the risk side: Our processes are distributed; they’re not integrated for efficient information flow.

The to ties to outcomes, which we identified by shifting those froms into four areas: trust, business value, time to value, and mindsets and skillsets around data and AI. The to statements became: Continuous focus on data quality that builds foundational trust; data and AI will generate meaningful business value; focused investments and behaviors will reduce time to value; and adopting a data-driven, knowledge-powered mindset enables everyone to engage.

Those initial from-tos became the objectives in 2025 and the scorecard for our C-suite. All the objectives and key results for data and AI were driven around those four major priorities with 360-degree alignment and accountability. Enterprise AI transformation is a team sport. It can’t be a “push” model from the CIO, but rather needs to be a “pull” model that includes alignment across the entire C-suite.

You’re advising companies across the country and in different industries. What are the winners doing differently?

Toolan: When you look at the relative investments in overall infrastructure, winners and losers are clear. The winners put their money where their mouth is in terms of aligning AI investments with long-term business strategy. Maybe they don’t have 12-month outcomes yet, but they’re going to have them at 18 to 24 months as those transformation investments take hold.

There is a certain amount of positioning, especially in public companies, of “What have you done for me lately? You spent all this money; what am I getting?” Shareholders and boards of directors expect to see ROI. And it’s tough to look below the surface to say, “Well, it’s there; you just can’t see it yet.” The winners are showing a balance of both short-term wins in the marketplace while also making longer-term data and AI-infrastructure bets for the future.

Talasaz: To add to Zar’s point, the ones who are winning with AI recognize that there is risk but assess what they need to do to enable the business to proceed with focus and intentionality.

Part of what helps de-risk AI use is preparation. Strategy, planning, assessing, and road mapping — the question is: Is the right amount of preparation being done before you go too far down the road? Because preparation is a de-risking mechanism. You prepare to increase the probability of success.

But what ends up happening is teams want to “get their feet wet” and feel their way through as a substitute for preparation. To be clear, you should try the technology out with a couple of POCs, learn what it is and isn’t good at, at a small and manageable scale. But getting your feet wet isn’t preparation for scale. If I don’t know how to swim, stepping into the ocean isn’t going to prepare me to go swimming out far from shore. It can be a good step, but it doesn’t translate into capability.

It’s one thing to put the money behind your AI strategy, but how do you make sure you’re putting your money behind the right things?

Toolan: You have to identify the business objectives you’re seeking. I’ll give you an example from Jones, within our CRM data. We switched from a legacy, homegrown, two-plus-decades-old system called JCMS, Jones Contact Management System, which was a treasure trove of rich information. Unfortunately, it was a treasure trove of rich information that you couldn’t personalize at scale. We’re talking mainframes and DB2, and once it was there, it was there —forget about putting AI on top of it.

As we shifted from JCMS to Salesforce, we had to do the hard work of data cleanup. Nobody wants to do data cleanup. But as for business value, Edward Jones has over a half a million conversations every single week across North America. Think about the robustness and richness of that information. What if we could put AI on that data to understand in real-time what is happening with our clients, what is happening with the markets? In the end, it took an 18-month process to get there in terms of cleaning, organizing, and modernizing the data infrastructure to be able to put scalable AI on top of it.

And it was a slog. I had to remind branches why we were doing this. We gave them a data scorecard, and we had data support teams that would work with them. We had to get to a critical mass, and this goes to the mindset shift. We created a trailblazer group of the top 200 or so branch teams that we put in first, and we were like, “This is not going to be a great experience for you at first, but you are blazing the trail, which will help you go so much deeper with your clients and helping them achieve their life’s goals.”

We had to light the trail as we were going, when branch teams would tell us, “I have this family that’s coming in, and before, I would have had to go through 18 steps in the old standard operating procedures, and I may or may not have even come to the right answer. But now, you’re actually bringing up insights that I didn’t even think to ask, because now the data is curated in a certain way, and when I search it, I get better responses.” Then what you’re starting to do is you’re building a whole new mechanism by which you learn from 20,000 branch teams. This is the classic definition of crowdsourcing a problem. Where you’re pulling the data, you’re thinking about themes, and then you’re creating archetypes that, guess what, you can put AI against. But until you understand what those true use cases are, and how you’re transforming those systems, it’s hard to put AI on stuff that you can’t even get access to.

That experience brings up the importance of change management. What are some of the key challenges around change and how can IT leaders address that, especially at scale?

Talasaz: Changing behaviors and changing technology are both hard. It’s important that technology leaders stop selling themselves, their teams, and the technical challenges short.

Getting the business change, people change, and technology change right goes back to the innovator-operator mindset and how you have to do both. This stuff is complicated, it is complex, and it does get difficult. There’s a lot of moving parts. You need detailed outcome statements, and then you need to break things down to the capabilities that create those outcomes. In Zar’s example, the capabilities were everything from how you put in data, to cleaning data, to how the branches used it. A dozen things were probably happening underneath to create that, so you have to know what those are, and then underneath those, you have to know what the technologies are, and then you can talk about the skills and actions.

This is really detailed and nuanced stuff. And the one thing that is oftentimes missing is that when people think about scale, the most important thing to scale, for me, is detail. You can’t scale without getting the details right. In these abstract conversations, we talk about these big macro things, when really the things that matter are the details.

You can’t just say, I’m going to let my vendor or someone else handle those details. The C-level executive has to get in and understand those details because they impact the strategy. The details of both the business and technical execution and the strategies and the large goals that are created at the more abstract level are directly related and they feed each other. You need to understand how those things are working together.

In a lot of cases, I don’t think people know what those details are, and that’s why they’re scaling the wrong thing. How do you get to the details that matter most?

Toolan: If you break all this information down to its sub-component parts, you can create archetypes and personas: Who are the people that are using my stuff, and for what purpose? Then you can create themes and trends: Based on those archetypes and those people, what are they actually doing? That’s what you can use to inform what you do about AI as a business and how you put measurable results around your AI strategy, which then takes you back to the personas, which takes you back to the data.

Afshean’s comment is brilliantly simple: You need the detail to scale. And to think about these future data streams as future revenue streams, you have to follow the ball: How do you have a time-series approach, or a longitudinal approach, to your business? And pulling in data and information is nonlinear; it has to be orchestrated in theoretically infinite dimensions. There are so many ways you can transpose and use inferences in your data for those personas, for those themes that drive those future outcomes, and ultimately get to hyper-personalization at scale. But without the foresight and the ability to put your data there, you can’t do it. The future of the agentic enterprise is going to depend on this use of massive data in pathways we are only beginning to imagine today. CIOs, and CFOs for that matter, need to double down on this use of their “sovereign enterprise data” as both a strategic asset and a competitive moat.

How important is it to have a tech-forward, AI-first CEO making this a top priority?

Talasaz: A highly supportive CEO is not an attribute of success; it’s a requirement. You can get single-base hits without it, but you’re not getting a home run. You can do some ground-up stuff that has positive impact and can drive value, but the kind of change that’s required to really integrate and scale this in your business is the kind of change that requires the chief executive support.

We talk a lot about VUCA as an external factor, but there’s a risk of companies creating internal volatility, uncertainty, complexity, and ambiguity as well. What do leaders need to do to minimize that risk?

Talasaz: When you’re going through any change, you are at risk of creating VUCA internally. Because when things are changing fast around you, it can be easy to create chaos internally in how you choose to respond to it. Once internal VUCA happens, though, you’re in a really tough spot. It’s not that complexity and ambiguity don’t increase during times of change. They do. But what you’re trying to do is take active steps to minimize it internally as much as is reasonable and deliberately trying to avoid increasing it.

How do I reduce uncertainty? Take the time to go through the story and narrative in detail to help people understand. Volatility? Be as consistent as you can in your messaging and your planning. Preparation reduces VUCA. Leaders have to really own reducing internal VUCA, because that’s one of the key tenets of being successful.

Data and AI and Digital is all about consistent execution. It’s about raising the floor, not just the roof. Enterprises that are successful are going to raise the floor, and the only way you raise the floor is through being consistent with execution and delivery. Not perfect, just consistent.

VUCA is increasing in the world, and it can be difficult for people to deal with that much uncertainty. You can’t fully get rid of VUCA; ambiguity is a natural thing as you go up in responsibility. But doing your best to reduce that is really important, because what you don’t want is the two internally and externally going up the same slope.

See also:

• Building IT leaders for an AI-driven future
• What CIOs must get right for the AI era
• Rethinking IT leadership to unlock the agility of ‘teamship’