Akamai’s CIO pilots AI, but isn’t often sold on full adoption, due to worries about costs and technology maturity

Kate Prouty, the chief information officer at Akamai Technologies, says workers at the cybersecurity and cloud computing company have developed a voracious appetite for artificial intelligence. They firmly believe that those with sharper AI skills will have an advantage.

“The demand for AI is out of control,” says Prouty. “It feels like a tsunami. Everyone feels like they need AI.”

While such enthusiasm may make it easier for Akamai to see strong adoption rates when new AI tools are deployed, it can also come with organizational challenges for leaders like Prouty, who oversees the company’s global IT organization and is responsible for assessing, testing, and deploying new AI tools to be used by over 9,000 employees globally.

Following the beginning of the generative AI boom in late 2022, Akamai initially embraced a “thousand flowers bloom” strategy for AI. The company quickly set up internal infrastructure that would give employees a safe “sandbox” to test use cases. While none of these AI applications were intended for full production, Akamai saw the value in encouraging experimentation. But only up to a point. 

“It didn’t make sense to me as a CIO that I would have people out in the Akamai ecosystem just developing AI and developing copilots,” explains Prouty, a 26-year veteran at Akamai. 

Now, Prouty prefers a more centralized approach and has adopted the thesis that most generative AI use cases that Akamai will adopt will come from the company’s vendors, including Cisco, Salesforce, and Google. Her team spends a lot of time with vendors, including a wide variety of AI startups that pitch Akamai, to better understand what their technologies can deliver, their future innovation roadmaps, and the changes that Akamai may need to make internally so it can best tap into these AI tools and generate meaningful productivity gains. 

When Akamai does opt to move forward with an AI feature, it does so in a highly measured rollout. “We’re sort of looking at as many of our vendors as we can in a very small pilot way, to understand what it is they’re offering and how it’s going to benefit us,” says Prouty, noting that she still has concerns about the maturity of AI technology and about the “murky” cost structures she’s seen from some vendors.

Each time an AI pilot is launched, Akamai creates a team chat channel in Webex so that people can share what’s working—or not working—when trying new AI capabilities.

Github Copilot has been rolled out under a “controlled release” for software engineers and in some cases, projects that would take weeks can be achieved in hours. But in other cases, the code written by the AI assistant doesn’t make sense and more work is needed to fix the errors. “There’s a learning curve,” says Prouty.

There is also some internal appetite to test offerings from other AI coding assistants, including Cursor and Anthropic’s Claude. But before Prouty signs off on that, she wants to really home in on measurable productivity benefits.

“I am still seeing that the technology is not quite there,” says Prouty. Workers still hit a lot of roadblocks when adopting these new AI tools and when that’s conveyed to vendors, they’re quick to say, “‘Oh yeah, that’s coming in the next release,’” she adds.

For some limited cases, Akamai sees a competitive advantage in building in-house AI tools. For example, the company partnered with French AI startup Dataiku to build a chatbot for the sales team, which taps into a blend of OpenAI’s LLMs and internal data from Akamai. The sales team is able to use this tool to pull a mix of private and public information about customers before making a pitch.

And while 2025 was christened as “the year of agents,” Akamai remains firmly on the sidelines when it comes to testing agentic AI. “I just don’t know if the technology is there yet,” says Prouty.

But even with the IT department exerting greater control of Akamai’s AI strategy, Prouty says she encourages an open-door policy when it comes to fielding new AI ideas.

“Let’s encourage, not discourage,” says Prouty. “Bring us your use cases. Let’s help you do that in a way that’s secure. But we’re going to put some cost controls around it.”

John Kell

Send thoughts or suggestions to CIO Intelligence here.

This story was originally featured on Fortune.com