EP 78: Do You Know What Agents You Have Running? Transcend's Rowan Stewart on AI Agent Governance

Sam Day (0:00)

You watched companies rush AI products to market without governance in place. What did that actually cause them when it caught up?

Rowan Stewart (0:09)

Yeah, I think when you're building really fast, you're under so much pressure to just get something out the door and prove that it's valuable, right? Especially with a very hype-heavy technology like AI.

Sam Day (0:22)

What's a governance mistake you see across highly regulated and consumer-facing companies like? Do you see a pattern that you want to share with my audience?

Rowan Stewart (0:32)

Yeah, for sure. So there are a number of things that kind of pop into mind. I think one pattern that I see is we're going to try every tool that we come across.

Sam Day (0:44)

How do you tell that person that move fast and govern your data aren't actually intention, rather they can go side by side, they can work together?

Rowan Stewart (0:53)

The analogy I might use is imagine you're on a switchback road in the mountains.

Sam Day (0:58)

If a company could only fix one governance gap before their next AI launch, what would be that and what is your advice to them?

Rowan Stewart (1:07)

Say, do you know what agents do you even have running?

SPEAKER_3 (1:13)

Today, we're joined by Rowan Stewart, an AI and Agent Governance Leader at Transcend. Rowan works at the intersection of AI, governance and responsible innovation, helping organizations scale AI with confidence.

In this episode, we'll dive into AI agents, governance and what businesses need to consider as AI continues to evolve. He is the company's co-founder and co-host of AssembleAI, Sam Day.

SPEAKER_4 (1:43)

Welcome to AssembleAI. Subscribe and follow us on YouTube, Spotify, Apple Podcasts, iHeart Radio and Substack.

Sam Day (1:52)

Welcome to Inside AssembleAI.

I'm your co-host, Sam Day. Right now, most companies are racing to ship AI products as fast as possible. And governance is the thing that gets promised later. Our guest today has seen exactly how the story ends. At Boston Consultants Group, she watched companies rush AI products to market without the governance structure to back them up and pay for it downstream. That's part of why she joined Transcend, where she now helps companies like Robinhood, Brex, and Groupon govern and activate their data responsibly. Please welcome our today's episode guest, Rowan Stewart, AI & Data Safety Product Leader at Transcend. Rowan, I'm very happy to have you here. Thanks for joining us. Our listeners are excited to get to hear your perspectives.

Rowan Stewart (2:46)

Thank you for having me.

Sam Day (2:48)

Rowan, let's start with the first question, which is kind of related to the cost of skipping governance. So at Boston Consulting Group, you watched companies rush AI products to market without governance in place. What did that actually caused him when it caught up? And what did you learn from that experience?

Rowan Stewart (3:09)

Yeah, I think when you're building really fast, you're under so much pressure to just get something out the door and prove that it's valuable, right? Especially with a very hype heavy technology like AI, you're trying to show the board members or your boss or your boss's boss that you know how to use this stuff and you know how to turn a profit on it.

And so it usually ends up kind of one of two ways. One of them is called, you know, something that we call that BCG pilot purgatory. So you build and you build and you do some kind of cool proof of concept in your sandbox and then you get right up to the edge of production and it's dead. It's just completely stalled because somebody, you know, on the compliance side or the legal side said, whoa, whoa, whoa, whoa, whoa, whoa, you can't actually do that. You know, you can't release this, you know, without more testing or it's accessing data that it shouldn't. And then you're in a really tough spot because you've invested all of this time and money into something that you can now not actually see the return on investment for. And so we see so many enterprises with these kind of little pilot graveyards where they're kind of sort of still kicking, but slowly they're being wound down because they can't prove the ROI, but they can't show that there's going to be enough ROI because they haven't proved it yet in order to bother going through all the compliance. The inverse of that is the folks who kind of rushed through and pushed something out into the market and accidentally found that there was a serious gap.

You hear all of these kind of different horror stories ripped from the headlines, so I won't run through them. And obviously, I'm under a number of NDAs not to reveal too much. But I can say that if you release AI, especially customer-facing AI, without doing your proper checks, you can end up with much more than regulatory fines. That is the least scary part. The compliance part is not important. What is very important is, I mean, it's important, but what's even more important potentially, is the kind of bad press that you get, the loss of trust from your customers when they realize you've been using their data in a way that they did not consent to, or potentially that your AI was giving responses that you did not intend and could not control. So I think that's kind of the bifurcation. You end up with either a lot of wasted money in pilots that never really show a return, or you end up with a ton of stuff out there that ends up with a huge embarrassment in PR and a huge loss of customer, investor, stakeholder trust.