By Calvin Hennick
CIOs must create a cohesive strategy for managing enterprise AI applications and data, which requires establishing a set of validated use cases, drafting policies and frameworks to govern use of AI tools, and centralizing oversight of the technology, says Nutanix CIO Rami Mazid.
Sprawl: It’s a problem nearly as old as the IT industry itself. Introduce a new technology, and sooner rather than later, it will spiral out of control—leading to budget overruns, governance issues, and security vulnerabilities. It happened in the 1990s with datacenter hardware. It happened in the 2010s with public cloud resources. And today, it’s happening with AI.
Although organizations only began adopting AI tools at scale a couple of years ago, these solutions are already reaching the point of “uncontrolled proliferation,” said Rami Mazid, chief information officer of Nutanix.
“They’re spread all over. And unfortunately, they are resulting in inefficiencies, redundancies, and security concerns. You’re exposing your data to unknown environments.”
Mazid noted that AI solutions are particularly susceptible to sprawl because they are so readily available to users, who might decide to inadvertently upload sensitive corporate data to free versions of tools like Claude or ChatGPT. Also, he noted, organizations are “rushing” to implement AI across their organizations to avoid being left behind, without taking the time to define appropriate use cases or analyze potential return on investment.
“When AI is not well thought out, and not integrated effectively with your other tools, you’re going to see a lot of wasted time and money,” Mazid said.
“And when you’re deploying in silos, you’re asking for trouble. CIOs really need to be one step ahead, and figure out how they’re going to attain the value and capabilities that AI solutions have to offer.”
What Causes Sprawl?
Mazid compared AI to cybersecurity, in that there’s no “silver bullet” solution available on the market. While organizations may be able to standardize on a single vendor for applications like video collaboration or customer resource management, they must currently build out both their cybersecurity and AI environments with a patchwork of disconnected solutions — a perfect scenario for sprawl.
“You’ve got too many tools to really address too many things, but there’s not one single tool that becomes your go-to,” Mazid said.
“In the AI space, the risks are escalating at an alarming rate. The accessibility of these tools — both at the enterprise and personal level — makes it easier than ever for anyone to deploy AI without proper safeguards. That’s what truly keeps me up at night.”
Mazid gave the example of sales and marketing. The two departments have similar needs regarding customer insights, and yet these teams often adopt separate AI solutions. The same pattern plays out, he said, across other related departments like research and development and engineering, with even sub-teams deploying their own suites of disconnected AI tools.
“You will see two different AI solutions that are doing very similar things, and yet they do not interoperate,” he said.
“You now have to pay for two different enterprise licenses, where you could potentially have one tool that meets the demand for both sales and marketing.”
New Risks to Sensitive Data
In addition to increasing licensing costs, Mazid said, this sort of sprawl can put an unnecessary burden on IT infrastructure. Even more concerning: Sprawl often leads to thorny problems related to data governance, compliance and security.
The fact that end users are the ones bringing new AI applications into the enterprise, Mazid said, means that these tools frequently go unvetted for privacy, security and compliance concerns.
“You might expose some sensitive data without even knowing it, and also create potential violations of privacy policies,” he said.
“A lot of people really don’t understand how to deal with compliance, and how to deal with data regulations like GDPR. It’s not their responsibility, and it’s not what they’re thinking about. They’re just looking for a tool to help them analyze their data, or whatever it is they need to do.”
Curating a More Mature AI Environment
To stop AI sprawl in its tracks, Mazid said, CIOs must put an end to the ad hoc adoption practices of the past couple of years, and create a more cohesive AI strategy for their organizations. This means establishing a set of validated use cases, drafting policies and frameworks to govern use of AI tools, and centralizing oversight of the technology.
At Nutanix, Mazid is establishing a cross-functional committee to take an organization-wide view of AI systems. He said it’s important for IT leaders to take a comprehensive inventory of all AI applications running in their organizations. This presents an opportunity to analyze benefits and risks, and also ensure that IT leaders have visibility into all AI tools. Over time, this process will allow organizations to standardize and unify their AI environments.
“We need to fully understand the use case, how these tools solve these problems and how we are going to enable them,” Mazid said.
“Then, the next step is to develop a strategic plan, and put in place policies and frameworks and best practices and use cases. These need to be aligned with overall business objectives.“
“If you don’t do that, you’re just going to be chasing your tail,” Mazid added. “You’re going to have sprawl forever.”
Learn about Nutanix Enterprise AI capabilities and how the company created its own GenAI applications to improve productivity across different business functions in this series of articles:
