Interview #92 Mangesh Pimpalkhare, SVP and GM, Splunk Platform at Cisco

Dr. Tony Hoang (0:04)

This episode is sponsored by Cisco Splunk.

Welcome to The Artificial Intelligence Podcast with your host Dr. Tony Huang. We're ranked in the top 2 percent of over 7 million podcasts on Spotify. And today we are in beautiful Denver, Colorado at the Cisco Splunk Conference with our friend Mangesh. Welcome to the show.

Mangesh Pimpalkhare (0:27)

Thanks, Dr. Huang. Really, pleasure to be here.

Dr. Tony Hoang (0:30)

Yeah. I mean, Denver is beautiful.

The conference is going great, and there's some awesome news that's coming from Splunk. So let's get to some awesome questions that I'm sure our audience would love to find out more about what you guys are cooking at Cisco.

Mangesh Pimpalkhare (0:48)

Absolutely.

Dr. Tony Hoang (0:49)

So every major platform right now claims to be the intelligence layer of the agentic enterprise. So when you have five vendors who own the same layer, who actually does own it?

Yeah.

Mangesh Pimpalkhare (1:06)

It's a great question. And if I think about it, like what does ownership mean ultimately, right?

There's many, many pieces to the puzzle. And if you look at the ultimate proprietary nature of enterprise data, not just data that's publicly available on the internet, customers ultimately own their data. But the question is, who provides them the end-to-end solutions? And the way we think about it is that if we can make use of some asymmetry that we have, add some solution pieces, and then add particular differentiated value, but make it not a take-it-or-leave-it proposition, but an interoperable solution, then that's the biggest value to the customer. So it's not a single-vendor ownership, but I think it's a little bit of who gains the customer's mind share, so that they can really unlock the power out of the data that they already own, but lighting up outcomes that they care about.

Dr. Tony Hoang (2:01)

Wonderful. And so right now, the industry has spent a decade preaching that data has been centralized to be useful.

Now, the message is, do you search it, where it lives, and then what broke?

Mangesh Pimpalkhare (2:18)

Yeah. I think it's not a question of just breaking. It's not as binary as that. Things come in waves, right? So I'm sure you've been involved in technology, and what makes sense at a particular phase of technology, actually tends to change in trends, right? And we saw this with client server software. We saw this with on-prem versus cloud, when there was a wave of workloads moving to the cloud.

And I would liken that to a similar paradigm here.

What made sense at a certain time, when data had certain volumes, was it made sense for groups that were working on certain problems to look at the data in their domain, and solve it as efficiently as they could. So that led to a partially centralized model. It was never fully centralized, right? You did not have machine data in the enterprise completely in one system, but it made at least sense for these partially centralized systems. What started breaking is massive changes in volume and complexity of data. And now with AI, it's getting even more connected. So I think it's a phase, and what makes a lot of sense is to ultimately connect all the data sources. But then the other fact is, it's not just about centralized versus decentralized. How do you actually extract the best value from all these interconnected signals without completely burning through your bills?

Dr. Tony Hoang (3:38)

This is one of our most talked about questions, which is machine data pricing has been the most complain about line item in enterprise software for years. What does it take for a category to walk away from its own billing model?

Mangesh Pimpalkhare (3:54)

That's another good question. So I'll address that directly with Splunk as an example, right? We, for a very long time, priced based on ingestion.

And this again, going back 15 years, made a lot of sense then because you're solving a localized problem. So almost all the data signals that you're looking at directly were helpful in answering that particular question. And so if there was value in answering the question, it was largely driven by data that came into Splunk. And so the metering was based on that data.

Back to the question about data volumes, as data started growing and started becoming even more diverse, customers really hated it, right? Because ultimately they care less about the managing of data, and much more about what are the questions they're able to answer and actions they're able to drive. And so we kind of changed this and went to infrastructure-based pricing where we started saying, hey, customers, you can lock in a certain amount of infrastructure.