Getting the most out of integrating AI—without falling prey to ‘AI sprawl’

There’s more than enough AI FOMO to go around in the business world right now.

C-suite executives fret about getting left behind as more and more companies automate their systems and processes. But jumping in head-first on AI, even when IT pros are being directed to do so, can lead to redundant applications, inefficient use cases, and the dreaded “AI sprawl.”

So how do you follow the broad directives from above to incorporate AI and increase efficiency without sacrificing the carefully optimized IT systems already in place? Working IT professionals told Morning Brew that getting AI integration right relies on careful preparation, planning, and supervision.

Secure the foundation

Before adding any AI or automation to IT functions, it’s important to audit current processes and personnel, Dhiraj Sankala, chief AI and technology officer at healthcare subrogation company Intellivo, told Morning Brew. That’s because incorporating AI can be an opportunity to scale the intelligence that an organization already has, in addition to upping efficiency.

“Start with your people,” Sankala said. “What kind of individuals do we have? What are the skill sets? How do we augment their intelligence, human intelligence, with AI? Same thing with the processes.”

Once IT teams have decided what it is they’re hoping to get out of AI, Sankala said the next step is to evaluate potential AI tools to see if they’re a good fit for the company and its goals. To do so, he told Morning Brew, it’s best to look at things through architectural, risk analysis, and financial lenses.

“We have to wear multiple hats to be able to really start thinking about, how do we introduce AI?” Sankala said. “You wear your architect hat: You really think about design, how this is going to basically fit into your current stack…You have to also think like a risk officer. What happens if tomorrow you have these AI tools in the system…but at some point in time we want to have our own agents?”

As for the financial “hat,” Sankala’s rule of thumb is a “70/30 principle”—70% of the IT budget should be devoted to the company’s current tech stack, and the remaining 30% can be used to invest in integrating AI into the stack.

“We want to really benefit from AI,” Sankala said. “We want to use AI for scaling intelligence, but we also [want to] develop the AI capabilities.”

Identify use cases

The first step in reaping AI’s benefits involves a bit of research: What do the people working with it believe they can get out of it?

Jason Kierna, VP of IT at school bus solutions company First Student, told Morning Brew that a big part of deploying the company’s AI platform, Halo, was meeting with the customers who would be using it and getting their feedback. When doing “field research” with schools and bus drivers, Kierna said his team focused on understanding what features they would want to see in Halo.

“We formed some user councils, and we involved them in all of our software development, from three years ago to today, as key stakeholders in that process,” Kierna said. “They were the ones that put together those feature lists: ‘These are all the things that we could use, and these are all the things that we would love to see to help our customers and to keep those students safe.’”

And Kierna said his team returns to those user councils to gut-check how Halo is performing and discuss current AI options.

“Understanding what’s going on in the field, and then involving people in the process and bringing them along,” Kierna said, “it really set us up for success.”

Streamlining tools

At billing and payments software company Zuora, SVP and CIO Karthik Chakkarapani told Morning Brew that success for his team follows a simple principle: Less is more. After consolidating and retiring the company’s SaaS applications, Chakkarapani said he’s being judicious about which AI tools he integrates into the company’s tech stack.

“We shouldn’t end up in an AI sprawl. It is a key lesson. We don’t want to be having 1,000 agents now and less SaaS apps,” Chakkarapani said. “We need to make sure your work is orchestrated in the right way, where there is less friction, better experience, and it’s also intent-driven work—nobody has to navigate to multiple applications and multiple agents.”

To avoid sprawl, Chakkarapani said he actually uses AI “as an analyst” to reduce redundancy and software spend.

“We were always under the expectation that we need to reduce our cost. Because, to be fair, we do have overlapping products,” he said. “The key part of AI is we need to amplify the existing potential, reduce or eliminate the friction as you do the work, [and unlock] new productivity gains.”