Name: Sierra AI
Uploaded: 2025-02-14T06:38:36.055Z
Duration: 2 min 33 s
Description: AI platform offering personalized conversational agents to enhance customer experiences.

sierra-ai

AI platform offering personalized conversational agents to enhance customer experiences.

Sierra AI

Our next guests met while working at Google nearly 20 years ago. They're now teaming up to tackle artificial intelligence through a startup called Sierra. It enables companies to build AI agents that directly interact with customers. 

Join us right now in an exclusive interview is Clay Bore. He's the co-founder of Sierra. He previously led Google Labs, spent 18 years at that company, and then Brett Taylor's here, co-founder and CEO of Sierra. Brett is also the chairman of the board of OpenAI, former co-CEO of Salesforce, former CTO of Facebook. 

You got a lot of titles, like Twitter's former chairman. We could be here for a long time. And the co-creator, hold on, of Google Maps, folks. Okay, I almost—I was going to start cheering. Che Google Map, exactly. Now that gets me. Now I'm impressed. When my dad gets lost, he still calls me. I'm like, "Dad, I haven't worked there for 15 years." 

So let's talk about what Sierra's doing because you guys just introduced voice and are doing some very interesting things. We keep talking about sort of the application layer of what AI can actually do for businesses.

Folks and yours I think your your efforts are probably the first doing that in real time right now. Yeah it's really exciting. So I like to start just with consumers because I think consumer behavior drives all else. 

And uh I imagine when in 2007 when Steve Jobs announced the iPhone you probably all had Blackberries right. And did you say there's no way in heck I will ever type my emails on a touch screen? Still has the Blackberry 10 years later you were using an iPhone and you were Ty touch screen dis grunted. I have asked for a keyboard. 

We're in this new era just like we're in the era of touch screens. Probably 95% plus of human computer interaction are touch screens. You fast forward five or 10 years 95% of human computer interactions are going to be having a conversation. 

We're in the age of conversational customer experiences. So in 1995 if you wanted to exist digitally you needed a website. You know a c 10 years ago if you wanted to exist on that iPhone you needed a mobile app. Now in the age of conversational AI you need an AI agent. This is the way your company is going to have its share AI Agent Store with colleagues.

Branded customer experience that can actually have a conversation with you and it's going to do everything you can on your site. So right now I call up the phone number for you know the airline or this and that and I got to press one two three and I'm saying Operator Operator please just I don't want to go through the whole thing where they don't doing all of this agent agent agent thing. 

The average American spends something like 12 to 13 hours per year on hold. So even before getting to the touch tone menu, you're waiting on hold. So the question is how good is this technology today and how frequently are people saying agent agent agent? 

So here's the thing. The average price of a phone call when you call customer service is around $113. $13 $13 every time I call up Delta and it varies depending on who's answering the phone and how complex the case is. But if you're a business and the price of a phone call is $13, you don't want to do a lot of those things. 

Yeah, but you don't want to tick us off either and make us feel like we can't get through to you no matter what happens. That's exactly right so now in this age of conversational AI not only is it a delightful experience rather than press four for order returns you can actually just have a conversation. But it's also less expensive. So what I'm excited about is if you imagine a world where those phone calls are not only effective and delightful compared to what they are now but they're an order of magnitude less expensive. How many phone calls will you have with your customers? I mean this will be fine like will you go to a brand site and rather burying that phone number it'll be right on their homepage and say yeah call us anytime. 

So not only I think is this technology delightful you don't need to wait on hold you don't need to go through the the IVR phone trees. You know what they say when I when they ask me what I need and I tell them they always say the same thing I didn't get that can you ask it can you ask another way right I didn't get that. So obviously we need to refine this right but it's refinable. It can one of the really powerful things about the way we build agents for companies is we can imbue an AI agent.

Which is basically an autonomous piece of software that can reason, make decisions, and of course converse with a company's factual knowledge. So everything about its products, its history, its background, uh, its complex processes. 

And we're able to integrate it with the systems that they use to run the company itself. So not only can the agent understand what you're saying, but answer complex questions and actually get how much investment does it take on a business's side. Some companies have, you know, lots of user manuals, scores of data, other things they've put it in data sets that are actually readable, etc. 

And then there's a whole bunch of companies that either never put all that stuff on paper or should have or have it in sort of disparate areas. They've never actually turned it into, you know, readable data. How does that process work and what does that cost? 

So we have a really unique model. We're actually a full service shop. When we work with our customers, we partner with them. We actually deploy what we call an agent engineer to every single one of our customers, and we deliver them a fully.

Working agent. So if you go to SiriusXM's homepage, you'll see Harmony, which is their AI agent. If you know the shorts brands Chubbies, their AI is named Duncan Smothers, and he is everything you would hope Duncan Smothers would be. Um, yeah, we were actually just here in New York meeting with some customers. We were just talking to Carly, who leads the customer experience teams at Yahoo and AOL. This is a brand that is the reason why we got into tech in the first place. You know, they were one of the first websites on the internet, and now they're working with us to make their first AI agent. 

And so any company can do this. There is, you know, there is some cleanup necessary because if you think of your standard operating procedure, it exists in the ether. It exists in your head. You have to do something to formalize it, but it's not rocket science. This technology actually works today, which raises the question of how big of a moat do you have. There are other people who are doing the same thing. You know, it's a great question, and I think it is a really competitive time. There is no doubt. We always.

Like to say I think the AI boom sort of echoes uh the dot boom. We've talked a lot about this. If you think of the do com, like what brands do you think of? Like pets.com. 

Web I'm not sure I agree with that. If you think now, like look at the stock market, what are the top companies? Amazon, Google, enterprise software, you know my alma moer Salesforce, all born in the.com bubble to some degree. The most froy statements about the.com bubble were maybe true. Um, it doesn't mean each of the companies individually was great. We know that we are not entitled to success and we're focused on one thing, which is we want to make our customers successful. 

Here's my question though. Two things. One is the underlying technology. Are you using a frontier model like an open AI to sort of run this or using a llama or something an open source that's sort of uh running underneath all of this? And two, how much do you worry that actually if any of the frontier models get good enough, they can effectively talk about the mode? They'll be able to just do this on their own effectively underneath the hood of our AI.

Agents we've actually built a pretty thick platform that uses not just one model but on any action or response as many as five even 10 separate models for different purposes. 

So decision-making, reasoning, planning, action generation, and so on. 

And within that platform we're also able to express really complex instructions, if you will, for an AI that makes it possible for an AI not to hallucinate, as you might have heard, but to actually follow faithfully these complex processes that customer service teams have. 

And there's a lot of IP in that.

Note, that there may be slight differences in text and video. Watch video for the original content.

Sierra AI Video