Interview #94 Kamal Hathi, SVP and GM, Splunk at Cisco

SPEAKER_1 (0:00)

College football is back, so Hilton called on me, the superstition concierge, to make your fan rituals a reality. Need a room to match your lucky number? We got you. Wanna make sure our team doesn't wash your lucky jersey? Ooh, that smells lucky.

Hilton's unmatched hospitality can keep up with any superstition. Even a marching band wake up call at 555 in 55 seconds. Hit it! When you need a team that will do whatever it takes on game day, it matters where you stay. Hilton, for the stay.

SPEAKER_2 (0:32)

When you've got a job to do, Lowe's knows how to help you keep it moving. Milo's Pro Rewards members save more with member volume discounts on eligible orders through a quote of $2,000 or more. Plus, save time with job site delivery on select purchases. Not a member? Join for free today.

Exclusions, more terms and restrictions apply. Can't be combined with any other discount, including but not limited to contract and or special pricing. Loyalty programs subject to terms and conditions. Details at lowes.com/terms. Subject to change.

Tony Hoang (1:04)

This episode is sponsored by Cisco Splunk.

Welcome to The Artificial Intelligence Podcast, ranked in the top 2% of over 7 million podcasts on Spotify. My name is Dr. Tony Huang, and today we're with Kamal. Pleasure to meet you. My pleasure. Could you introduce yourself?

Kamal Hathi (1:21)

Yes. I'm Kamal Hathi. I'm the general manager for Splunk at Cisco, and we are here today at Splunk.conf in Denver.

Tony Hoang (1:33)

So the line of this era is that agents are digital coworkers. What is the first thing that enterprises will regret delegating to them?

Kamal Hathi (1:44)

Let's be careful about starting with regret. I think agents can be very, very helpful, can lead to massive productivity gains. I think the thing to watch out for is assuming that instructions given to AI, given to agents, are going to be as you intended. Intent is not sufficient. You need to be specific, and you need to have evaluations, metrics that measure their specificity, so you are getting exactly what you want from the agent versus what the agent thinks you want.

And it sounds like a subtle thing, but it's important, because sometimes what happens is, AI and agentic solutions are more or less ruthless in being goal seekers. They will do anything that it takes to meet a goal. And if your goal isn't clear, everything is fair game, including things you might regret, which could be things like you might say, hey, I want to optimize this space on my computer by 3 percent, and I'll make sure that you keep my most important files. But what you didn't say is don't delete everything else.

You might end up with 97 percent displace free because they decided to delete everything. Because your instructions were that, and it was exactly right what you were doing. So this is the thing where you have to be careful. And so I don't think it's a matter of regret, it's a matter of being very, very explicit both in instructions and in measurements and control.

Tony Hoang (3:05)

Gotcha.

And then you also called trust in AI almost existential this morning. So what are vendors confusing no for trust?

Kamal Hathi (3:18)

Yeah, I think, again, it's very similar to what I said before.

Trust is about being sure what your AI is going to do.

And what vendors or even consumers confuse is having the notion of instruction, even instruction that says don't do anything bad as being trustworthy. It's about verification. It's about measurement. It's about visibility of what the data you're collecting. It requires a full-blown framework and focus in order to make sure that you have trust. Trust doesn't come automatically. It's a difference between you hire a human being and they want to work for you, and you tell them, do something, and you can trust them. It's just trust me. You got certain values. You got the certain human tendencies we take for granted. And I think as I said before, with agentic and AI solutions, these are non-deterministic that way. And so I think it can get confusing if you start thinking about trust, but don't think about the verification, the visibility, the controls, the overall end-to-end system upfront. Doing it afterwards is too late.

Tony Hoang (4:30)

Gotcha.

And as data grows, inference and cost is also going to grow. So the industry is moving into a model where inference costs will be billed back to the customer. Or do you see like a future where like SaaS providers are going to to bill that into the cost of running the product?

SPEAKER_1 (4:50)

College football is back. So Hilton called on me, the superstition concierge to make your fan rituals a reality. Need a room to match your lucky number? We got you. Want to make sure our team doesn't wash your lucky jersey? Oh, that smells lucky.