Enterprise AI Maturity Depends More On Data, People, And Process Than On The Models

Enterprise AI programs often get compared on the size of their model portfolio, the certifications their teams hold, and the tokens they consume. Nearly nine in ten organizations now use AI regularly, but most still haven't begun scaling it across the enterprise. The gap tends to open in the work around the models, where data quality, planning, talent, and cost discipline decide what AI delivers. AI maturity, meaning an organization's ability to convert AI into measurable business value, depends on getting that work right.

Anubhav Mishra is Senior AVP, Data Engineering and Analytics at EXL, where he manages a data engineering and analytics portfolio for healthcare clients in the United States. His career spans more than two decades in data and AI, including leadership roles at Infosys, a data and analytics startup he founded, Encore Capital Group, and HSBC. Mishra designed his own framework for AI maturity and uses it in his day-to-day work.

"Data is the foundation, AI is the accelerator, and leadership is the differentiator. Unless leadership is strong enough to drive all of these parameters, there will be no success. It doesn't matter how many models or how many tokens you have. You need a very strong leadership layer," says Mishra. He compares the AI market to a Formula One race, where every team has a capable car and the win goes to the driver who is both fast and safe. Enterprises compete on model counts and certifications, while the outcome depends on who's steering.

Clarity, capability, competitiveness

Mishra's framework pairs three Cs with six Ps. Clarity, capability, and competitiveness each run through priority, planning, people, process, performance, and profitability, for 18 parameters in total. Clarity covers what the organization wants to achieve and the roadmap, talent, processes, and tracking needed to get there. Profitability stays on the list because every organization needs to earn a return on what it spends. "The very first thing is the priority, what you want to achieve. That is the most important aspect. Without that, you can't do anything," Mishra explains.

Capability asks whether the organization's data platforms and models, large or small, can deliver the value it needs. Accuracy may fall short on early attempts, so teams keep testing, redesigning workflows around business needs, and evaluating models against a target value. The talent identified during planning runs those iterations until the output is right. "Nobody is perfect, we're all learning. Similarly, models are also learning on a day-to-day basis," he notes.

Competitiveness measures where the organization stands against its market, and a product priced far above what competitors charge won't find buyers. The final parameter asks what value an AI investment adds to the business. Mishra sees the industry moving away from return on investment toward what he calls "value of investment," since enterprise AI is still early in its maturity. "Achievement will only come when we're mature enough. Right now, we're emerging and learning every day. That's why the value we're adding to the ecosystem is more important," he says.

Data before the model

Many companies treat AI as a race and expect a new model to transform the business within a week. Mishra would rather teams move step by step, starting with a clear objective and an honest check on whether their data can support it. The work can include revisiting the data and running several rounds of cleaning before any model work begins. "If you take the stairs, you can go slowly, but you will definitely get there," he explains.

Data quality is where he'd direct most of the effort. Once the data is fit for the model, teams can choose the right algorithm and then match the right people, infrastructure, and platform to it at minimal cost. The upfront homework pays off because the models themselves remain very expensive. "My recommendation to all professionals is to focus heavily on data quality. If data is good, your efforts will be 50% less," Mishra notes.

Mishra compares data to the foundation of a building. Someone still has to make sure the construction follows the rules and regulations and uses the right materials, and in an enterprise, that job falls to leadership. "Garbage in, garbage out, but leaders are important so that they can avoid this garbage in. If you don't control the garbage in, there will definitely be garbage out," he says.

Ready for what's next

Mishra's framework also depends on people who can adapt. He expects leaders to keep teams learning continuously through defined roles, planned training, and stakeholders who add value. Performance needs tracking and continuous feedback, so leaders can take corrective action when something falls behind. "We never know what we'll need tomorrow. Two years back, a lot of the focus was on generative AI. Today, more of the focus is on agentic AI. As leaders, we need to make sure our talent is ready to pick up new challenges," Mishra explains.

As AI becomes more autonomous, Mishra holds the people who build and use models responsible for how they're applied. He wants models developed with ethics, culture, and responsibility built into their training, and he expects governance and regulation around AI use to grow. "Responsible AI is the need of the hour. Today, we have the role of Ethical Hacker. In the future, I'm pretty sure we'll have the role of Ethical AI Engineer, people who control its misuse," he concludes.