Interview #103 Somya Kapoor, CEO of IFS Loops

SPEAKER_1 (0:01)

You know you don't have to wait to Shop Black Friday at Lowe's, right? Save up to $150 now on select power tools, equipment, and accessories. Got gifts to get? We've got plenty under $50.

Plus, My Lowe's Rewards Members get same day delivery for free. Shop Black Friday deals before the rush at Lowe's. Valid through 1118, $25 minimum in loyalty membership acquired. Order eligible items by 2 p.m. Subject to weight and size restrictions and availability. Terms at lowes.com/shippingterms.

Tony Hoang (0:35)

This episode is sponsored by IFS.

Welcome to The Artificial Intelligence Podcast, ranked in the top 2% of over 7 million podcasts on Spotify. I'm Dr. Tony Huang, and today we are in beautiful Orlando, Florida, at IFS Unleashed with Somya. Welcome.

Somya Kapoor (0:57)

Thank you, Tony. Great to be here.

Tony Hoang (0:59)

Can you give the viewers a little bit of your background?

Somya Kapoor (1:02)

Yeah, so Tony, I come from SAF, been in the enterprise space for the last 30 years, kind of dating myself. I was at SAP ServiceNow, went to the startup land, and then eventually got bought by IFS last year. So been in the enterprise space for the last 30 years, building enterprise software.

Tony Hoang (1:21)

Wow, exciting. We have some very interesting questions I'd like to dig into about your work at IFS. So most AI agents' projects fail.

What's the most important way that they die?

Somya Kapoor (1:37)

So most enterprise projects were failing.

The very fact is because of the change management. The technology is 30 to 40 percent. It is definitely there. It works. The other aspect of it is change management, right? It's not about taking an agentic solution and just redoing the way you did historically. It's kind of rethinking the business process from the lens of now. You have these agents that you have to remove approval processes. You're no longer restricted in the data residing in different systems. How do you kind of bring the escalation parts, the approval parts? How do you bring the organization together so they don't feel threatened by these agents kind of coming in? How do you keep seeding? Because historically, I mean, automation is nothing new in the ERP space or enterprise space. It's been there for the last 30 years. What is different from that to now is that you don't have to have endless developers. You have to spend eight months deployment cycles. You can actually do it in real time. And that's the biggest difference. But for that, you can't take your old processes and kind of adapt it to the agentic world. You have to rethink the process and then bring people along with them, kind of bringing the change management aspect of it as well.

Tony Hoang (2:53)

Gotcha. And you've said your 11-year-old could build an agent.

So if building is the easy part now, what's the hard part that no one budgets for?

Somya Kapoor (3:04)

So building is the easy part, right? That's why there's a lot of demo ware in the market. People are able to quickly demo an agent. Putting in production requires guardrails. Putting in production requires the memory management aspect of it. Putting in production requires you're not burning all the tokens out on one deployment of the agent. Structuring, designing, security elements. How do you scale it across regions? Where the data is getting? What data do you see the LLM? How do you orchestrate the LLM, right?

It's all that context that's around using these large language model is where the work kind of goes into moving your agents from demo ware to production readiness.

Tony Hoang (3:45)

Gotcha. And I've seen like a lot of demo ware is right now is passing for products. What one question exposes an agent that will never survive production?

Somya Kapoor (3:57)

You know, it's scaling of that agent in an enterprise, especially with the customers that IFS have, right? You can design an agent to respond to one vendor. How are you going to design that agent to respond to multiple vendors when you have to replenish that part across regions, across time zones, right? I think that's where it kind of fails. It's very easy to build a demo ware for one vendor as opposed to scaling it across.

Tony Hoang (4:26)

Gotcha. And you put the split at 40% technology, 60% change management. So in your opinion, why does the industry keep funding the 40 and starving the 60?

Somya Kapoor (4:37)

I think it's a mindset shift. You know, there is a lot of agent fatigue in the market. I constantly have conversations with the CIOs. They're like, how is your agent different? What is it?