Transcription
My name is Mitch Lieberman, VP of product. So for 15 minutes, a little bit of an anacronism. The word agent means person. Okay. So we need to think about what that means in the context of a contact center.
So I'd like to talk to you today a little bit about is what's the problem we're trying to solve? Why 11 Labs and how we chose 11 Labs, the platform that we're building, the demonstration, and some results.
The problem statement, finding people, hiring candidates, onboarding new hires, engaging tenur agents needs change. Agent turnover in a contact center is between 30 and 50%. So do the math. Onboarding is expensive and coaching isn't consistent. So we need some help. We need to build competency, right? It's very people intensive. We have to focus on soft skills, compliance, hard skills. And finally, classroom is not enough. PowerPoints don't do the trick. So we need voice-based engagement.
So we started looking around to see how we can solve this in a conversational AI way. We wanted to roleplay in simulation. We have an interactive AI coach. We want to create immersive experiences.
Okay. So why 11 Labs? We started our journey almost to the day one year ago. User experience latency that was our top priority. Hint, it's low and what we heard this morning is getting better. Voice quality also voice selection.
So what's our approach? We need pace of innovation from our partners. We need a platform that we can depend on that has flexibility and balance cost and how quickly we partner for speed and we build for differentiation.
So when we considered the problem space, we started in the middle because that was where the best product market fit existed. We have to train. The word is speed to proficiency. How quickly can we bring customer service or contact center agents up to speed because of the churn? How do we help them feel more comfortable and confident? And then we started branching out towards coaching, voice-based coaching and hiring and a candidate experience.
Okay. So, we ready for the demonstration? Are we ready for the demonstration?
[cheering] [applause]
>> There we go. So, what I have here is I created a scenario. The scenario is called my phone is locked. It might be a little bit simple, but it's very straightforward and it kind of illustrates the points. We set up this scenario and we deliver it to an 11 labs agent. be creating an agent that's persistent so everybody can use the same scenario and we ask we develop skills that the contact center associate should work on when they're working through this scenario.
So I'm going to go through this and let's see how I do. I'm going to accept the call.
>> Hi, this is Mitch. How can I help you?
>> I can't get into my phone. It's asking for a pin and I can't remember setting one up. Can you help me? I need to call my sister.
>> I'm sorry to hear that. My name is Mitch. Who am I speaking with?
>> I'm Janet Smith.
>> Okay. And so you set up a new phone. Did you buy it in the last few days? And do you remember if it's an Android device or if it's an iPhone device?
>> Yes, I bought it yesterday. It's not an iPhone. It's one of those camera ones.
>> Okay, that's interesting. Um, when you set it up, did the process ask you to use your finger to set up a an a biometric or a finger fingerprint identification?
>> Yes, I remember using my index finger.
>> Do you remember where you used it on the phone? Can you put it back on that same spot?
>> Oh, you mean the home button? Let me try. Oh, it opened. Thank goodness.
>> So, roleplay simulation is different. So, thank you very much for help. Glad we could help. So, do you need anything else for me today?
>> No, that's all. Thanks for your help.
>> Appreciate it. I'm going to say goodbye now. Goodbye.
>> Goodbye.
>> That was aggressive. Okay. So, what we do now is we have a team of data scientists that have built evaluation criteria. We divide each skill into a unique system prompt and we use models that we control to create an evaluation and we play the evaluation back to the contact center associate or the one in training so that we can then review with them how it went and sometimes it takes a little bit longer.
So what happens here is this is something that we knew and this is how fast we can move with 11 labs. This particular box was is a new creation we created a week ago. Last night, I was not happy with how quickly it popped it up. So, sitting around a table, we had our head of design, our engineering manager, and our head of engineering who reviewed the code on his phone so we could push it to production and we could make it live.
So, I'm going to select the default agent and I'm going to start another interaction. Hi Mitch, I'm here to share some more thoughts on your just completed roleplay simulation so we can further refine and improve your skills. Are there areas you would like to start with? Let's dive into your recent call. I noticed you did a great job guiding Janet to use her fingerprint to unlock her phone, which directly resolved her immediate problem. That's a fantastic demonstration of your problem solving skills. I don't think I value our customers very well, though. What do you think?
>> That's a fair point, Mitch. It sounds like you're reflecting on how often you get to apply certain skills in this specific call. Your ability to ask clarifying questions about the device setup was key to helping Janet. You asked about when she bought the phone and if it was an iPhone, which helped narrow down the possibilities. This shows good foundational active listening.
>> What do you What do you think my manager is going to think of this evaluation?
>> I'm sorry, Mitch, but I can't provide an opinion on salary issues or what the managers should or should not have done regarding compensation. My role is to help you improve your
>> Thank you very much for sharing your thoughts. So, this was the middle of the experience. This was the training experience scenario and AI coach. I'd like to show one more point.
We talked a little bit about something that we put together which is what we call a screening application for candidate experience. I mentioned in the larger picture that we have a candidate experience. So this is meant to facilitate and by the way we're being very very cautious with this is that we're facilitating an onboarding experience so we can do simulated roleplay. humans are in the loop. We don't feel like we're at the point in time we can perfectly grade a candidate experience. So, we're working on it and we're pushing we're pushing as fast as we can.
So, a candidate will go through a very similar process. We'll go through a similar role play and I will then be able to give this and so what I'm grabbing here is you see up on the left side is the 11 labs copy of the audio which I can play back. We have the transcription here which you can see that this is not a very long one but the a very so we can we can so the recruiter can play back to begin the interview. Hello. That one was taking bad audio. Apologies for that. And over here we actually have the evaluation criteria where we're very particular that we want people to select the score. We are not giving a score. We're giving an observability using a model. Okay, which we find to be very important within the process. So we can look at the evaluations and go through the process.
What a what a candidate might see is something like this. I did pre-populate this. So we have an interview registration. They sign up. They agree through can we have this is the interview instructions and I can then start another simulation so that I'm able to give a realistic picture to the recruiting team on whether this candidate fits the profile of somebody that they're looking for or maybe just to align them with a particular department or particular call type. So I can accept and continue and it's a slightly new experience timed out which is actually very much part of the process. So can leave the site and go in here create what's called the new batch. Copy the link. Go here. Do we want we don't we want people to take this in a controlled environment? and continue to the interview.
The other team, the other thing the team thought would be fun is that there's a randomization in here so that I'm never quite sure which one's going to come up.
>> Hi, I'm ready to begin the interview.
>> Hello, my name is David Alexander Smith. I'm calling because I requested a replacement credit card two weeks ago and I was told it would arrive within 5 to seven business days, but I haven't received it yet. The tracking information hasn't been updated either. I'm a bit frustrated about this delay.
>> Very sorry to hear that. I don't think we need to go through the entirety of this, but if we can click go on to the results screen, I think it's it's it's really important to understand that we've been able to move very quickly with 11 Labs.
So, we were able to we made the selection January 1st of this year. We had GA product on May 1st. We since May 1st, we've had 35,000 training scenarios run through the system to all the way through to full evaluation. We're deploying to contact centers with over 5,000 customer service agents for upskilling and we're starting to launch coaching. the coaching the the coaching application and the screenre the application you just saw there were built over the last three weeks and they're pushed into alpha so success is how quickly we can move 11 labs has been a great partner we've pushed them and actually the joke I made at the beginning about agents was an interesting conversation with their engineering team is we agent is a reserved word in the 11 labs platform right our head of engineering said we need the word and 11 labs we need the word. So we figured it out. They joined up with us. We collaborate on Slack and they helped us solve the problem.
The results we're seeing is 20% reduction in time required to bring agents up to being proficient. So if a training takes 10 weeks and you can reduce it by two weeks and multiply that by 30,000 people a year, that's numbers. So it's absolute ROI production in time for script adherence. How how did they apply? So we can use the knowledge base to pre-populate these agents with what people are supposed to say and reduction in attrition. They're leaving less. They feel more comfortable and confident in their ability to requ to take customer calls and their ability to be prepared. And interestingly, they like work new contacts and our agents enjoy working with an AI because they don't feel like somebody's judging them at the other end of the phone. we've separated their ability to learn and have an immersive experience from their ability to be judged. And so the the the verbatims we're getting on that have been very positive.
So just wanted to do say thank you for your attention and really appreciate your time and we're outside if you want to talk more details, but I really look forward to hearing the rest of the speakers and have an amazing rest of the conference.
[applause]