AI Interaction Is Replacing Pure Intelligence | Ashley Kramer

Ashley Kramer (0:00)

I'll start with how ElevenLabs is using ElevenLabs' own technology. We just launched, two weeks ago, an AI sales development representative. The name is Dom, an AISDR, that helps them purchase our solutions if they want to get to a fast resolution, leverage our APIs better, or quickly get booked to go talk to a human. Next step is handing off to an AI AE, if we want to keep them along the AI path.

The AI AE has already closed two deals. We have an AI lawyer that helps get to the next step of contract negotiations, so you can translate that, really, to efficiency and productivity into companies as much as serving external customers.

Theo (0:42)

All right, we are back, we are live with Ashley Kramer, who is the Chief Revenue Officer at ElevenLabs, which had a really insane week this week between Eleven v4, their latest voice model, which is really good, and a new $300 million employee tender at a $22 billion valuation. So congrats, Ashley, welcome to MTS.

Ashley Kramer (1:04)

Thank you for having me, Theo. I'm excited for the conversation.

Theo (1:06)

Me too. So tell us about Eleven v4. What is different about this model? Why is it- Yes.

Ashley Kramer (1:13)

Oh, sorry. So yeah, we are super excited earlier this week. We launched v4, which is our most emotive model to date. It has sharper context, a brighter emotional range, dynamic multi-speaker conversations. And just as importantly, we now support over 90 languages across the world, spoken by over 5.5 billion people. So we are really, really excited to get this into the hands of people and customers.

Theo (1:41)

Yeah, it sounds awesome. So what were some of the specific changes that were made between Eleven v3 and v4 that allowed v4 to be so much better?

Ashley Kramer (1:51)

We actually did an architectural change, and so we were able to infuse a lot more of those capabilities. We've all had those experiences where we're trying to interact with AI, but it sounds like a robot talking back or there's long latency pauses. Our Turbo v4 eliminates that and gives that real live back and forth conversational type experience that you would expect to really want to interact with AI versus feel like you're talking to a robot. So the team did great work as far as making sure that the architecture was in place, we had the right datasets, and we alpha tested with customers, and we're really excited that we were able to launch it this week.

Theo (2:36)

How big of a trade-off is there between speed and expressivity? To what extent is the model being more emotive, more able to express emotions and feelings?

To what extent does that cause penalties for how quickly it can respond?

Ashley Kramer (2:54)

I think it depends on the experience you're trying to provide. So if you're having an experience where there's deep context that's needed, obviously we leverage large language models to get that context.

Sometimes there needs to be a slight delay, but not a big one because that will lose the customer, which is why we have V4 Turbo to have more of that direct back and forth. The emotion and the tagging that we do is part of that, is a big part of the experience. For instance, if you are a patient calling to make an appointment with a hospital because you have a heart issue, you want that sympathetic, that empathetic and probably urgency behind it type experience, versus if you're a Gen Z calling a retailer and you want to get new sports equipment, obviously, they have their own way of speaking in their own language. And so that's what we're really focused on is giving that experience. And where that translates for enterprises is you can stay true to your own voice, brand tone across any channel that you're using to support customers or your internal team.

Theo (4:02)

Totally.

So what are some of the new use cases that are enabled by 11v4 being so much more expressive and so much better? Like what are you really able to do now? What like thresholds have been crossed?

Ashley Kramer (4:15)

Yeah, so we're still doing some benchmarking in the thresholds. We believe we're going to perform very well, but the new use cases will be all across how the interaction is happening. Of course, there's the new supported languages, which has been really big. Mandarin, Japanese, Brazilian, Portuguese are big ones that will enable us to really help those markets in a more effective way.

So that's on the language side, but more on the use case side. It's back to my point about being able to be specific to the situation that's happening, the model consents by the context of the person talking to it or interacting with it, what kind of response it needs, what kind of emotion it needs behind it. Then our future vision that we're working on is being able to capture that so we can actually have that as part of the customer experience. Maybe if you are a retailer, you know that the last time the customer called, even if it was speaking to an agent, they were pretty upset. They didn't have a great experience, so maybe this time when they call, start by apologizing first and understanding that because we're carrying that context, whether it's in flat text or a human conversation or an AI conversation across the entire Customer 360