If you’re an IT professional, chances are good you’ve been asked by management to integrate some form of AI into the company’s processes and procedures. Duolingo did just that recently, announcing last month that the company would become “AI first.”
But it’s not as simple as just adding an AI agent to your team’s workflow and calling it a day. You need to take care in making sure the tools you’re incorporating into existing systems are appropriate, secure, and cost-effective.
Start at the beginning. F5 Field CISO Chuck Herrin told IT Brew that the first step in the process is ensuring the AI tool you’ve chosen is one that will fit in with the organization’s priorities.
“You should try to start to separate the what from the how, figure out why you’re doing something, and then choose the right tool,” Herrin said. “And I think a lot of organizations are starting with the tool and then and then trying to fit it into some use case.”
Measuring effectiveness is another hurdle to overcome, Herrin added. You might be forgiven for thinking that tools are interchangeable. But there are problems related to integration and to data quality that can be “deceptively hard” to solve, according to Herrin.
“Managing model drift is kind of an ongoing task, and if you try to do all that stuff at the same time with a bunch of different use cases, then it gets really, really complicated, really quickly,” he said.
Check it out. Jim Routh, Saviynt’s chief trust officer, told IT Brew that he recommends a use case review process in order to measure the potential effectiveness of the AI technology in the organization. That means talking to people in the HR, cybersecurity, and data science teams, or at least getting a sense of what an effective AI-powered outcome looks like. A cross-functional team can figure out which use cases make the most sense—and can provide better insight than multiple perspectives on AI governance.
Top insights for IT pros
From cybersecurity and big data to cloud computing, IT Brew covers the latest trends shaping business tech in our 4x weekly newsletter, virtual events with industry experts, and digital guides.
“You say, ‘No, I’m going to just deploy controls for this use case, and that’s my starting point. And then I’ll take the second use case and then the third, and then the fourth,’” Routh said. “Each use case is reviewed by this cross functional governance team, and that’s a model that allows you to build policy and ultimately control standards bit by bit in a way that’s aligned with the business strategy for the enterprise.”
Making sure. Oftentimes teams need to explain to higher ups whether or not tools will work—and that can come in contrast to what executives might want to hear. Negotiating that complex back and forth can present problems. As Herrin put it, there’s a question about where AI might fit in your organization and how the products are useful, one that comes with risk of failure, or at least less benefit than was expected. Duolingo learned the hard way that AI doesn’t solve everything, backing away from its aforementioned AI goals after customer backlash.
“Proving out that business case is still the most important way to make sure that you’re actually putting your resources in a way that’s going to benefit the business—instead of just checking a box and saying, ‘Yeah, we did AI all the things,’” Herrin said.