[0:04] **Me:** Hey, good afternoon. It's a pleasure to make your acquaintance.

[0:07] **Me:** Good afternoon.

[0:11] **Voice 2:** Yeah, thanks for joining. We're looking forward to hearing what you've

[0:12] **Voice 2:** got.

[0:17] **Voice 3:** I know we don't have a ton of time today, I'll really jump right into things.

[0:22] **Voice 3:** As far as an agenda, I just wanted to start with a couple of hellos and then

[0:22] **Voice 2:** Mhmm.

[0:27] **Voice 3:** I don't actually have an agenda slide, so I should probably create one of those. I'll make a note

[0:27] **Voice 3:** to do that afterwards. But

[0:31] **Voice 3:** let's just say that we'll, look at our AI assisted whiteboard,

[1:06] **Voice 3:** talk very briefly about Rhyme's, history and and why we're we're quite unique. Then really dive in, spend the most of our time talking about areas you're looking to improve with your current voice solution, that you use a a nice Cognity for. I also have a a brief custom demo. I'd love to show you. And then afterwards, we can get some feedback and talk about potential next steps. Of course, along the way, please don't hesitate to stop me for any questions. Sound good?

[1:06] **Voice 2:** Yeah. That sounds fine. Just to be clear, what exactly are you hoping to get from this first conversation?

[1:08] **Me:** Really just an understanding of

[1:13] **Voice 3:** the reasons that you decided to take the meeting today and some of the challenges you're facing.

[1:16] **Me:** Well,

[1:18] **Voice 2:** get pitched a lot.

[1:21] **Voice 2:** Our priority is reliability and

[1:23] **Voice 2:** not creating risk.

[1:26] **Me:** So,

[1:26] **Voice 3:** reliability is a

[1:26] **Me:** pain point?

[1:28] **Me:** Sometimes.

[1:29] **Voice 2:** Yeah.

[1:32] **Voice 2:** Especially during peak call times.

[1:34] **Voice 2:** Though, to be honest,

[1:36] **Voice 2:** any change has to be justified.

[1:41] **Voice 2:** We don't swap

[1:41] **Me:** corn for the starting

[1:45] **Voice 3:** obviously, my name is Colin. I'm the founding account executive here

[1:45] **Voice 3:** Perhaps you guys could do a quick round of intros on your side?

[1:48] **Voice 2:** Oh, of course. I'm Dana.

[1:53] **Voice 2:** I'm over customer experience. I own the metrics around

[1:56] **Voice 2:** customer satisfaction and containment.

[1:57] **Voice 2:** I'm Marcus.

[2:08] **Voice 2:** IT and systems. I care about integration, security, and not breaking in what's already working. And I'm Priya from procurement.

[2:11] **Voice 2:** I come in when there's a clear business case,

[2:13] **Voice 2:** cost clarity, and

[2:14] **Me:** risk controls.

[2:17] **Me:** Super helpful.

[2:22] **Voice 3:** Just so you know, so we're using an AI

[2:24] **Voice 3:** whiteboard today, so it's gonna just help us organize a couple thoughts and really,

[2:29] **Voice 3:** make sure we track what's most important to you so we we lose track of anything. And I

[2:32] **Voice 3:** can still be present without taking a million notes. So let me jump into that and,

[2:33] **Voice 3:** let's go from there.

[2:38] **Voice 3:** Oh, you know, one last thing just so you have a bit of context on us. So we're at Rhyme. We're

[2:41] **Voice 3:** three years old, started by Lilly Brookton Ares.

[2:42] **Voice 3:** And

[2:47] **Voice 3:** we take a little bit of a different approach to this problem. You know, we we come from a linguistics background

[2:48] **Voice 3:** primarily as well as a

[2:53] **Voice 3:** having a lot of AI engineers. And when when we sat down to tackle it,

[2:57] **Voice 3:** we rather than using, you know, Actor generated training data or

[3:01] **Voice 3:** podcasts like everyone else did, we actually set up a recording studio.

[3:05] **Voice 3:** And we we started having, you know, legal folks actually speak

[3:10] **Voice 3:** for thousands of hours that we can capture natural human communication

[3:12] **Voice 3:** in a much more nuanced

[3:14] **Voice 3:** realistic manner than the

[3:16] **Voice 3:** models that are created with the traditional data.

[3:21] **Voice 3:** But the proof of all that is just, you know, looking even as a small

[3:23] **Voice 3:** we're now powering tens of millions of calls every month for

[3:26] **Voice 3:** healthcare, telecom, enterprise support,

[3:27] **Voice 3:** we're all over.

[3:32] **Voice 3:** Just even last month, did a study with 100,000 calls that we

[3:32] **Voice 3:** help sponsor.

[3:36] **Voice 3:** A third of those were with us and then two were from our competitors. And

[3:41] **Voice 3:** although no one actually loves receiving calls, this was a politics survey, so it's

[3:41] **Voice 3:** legitimate.

[3:45] **Voice 3:** We had the lowest hang up rate amongst any of them and the fastest completion rate.

[3:49] **Voice 3:** So people actually do enjoy speaking to our AI.

[3:56] **Me:** But anyway, you're you're you're primarily to talk about voice AI at Assurant.

[3:59] **Me:** It would be great to understand how you're generating speech today, what tools, and

[4:04] **Me:** you know, what sort of outcomes you're you're looking for if things work

[4:04] **Me:** and then maybe a little bit about where things are breaking.

[4:09] **Voice 2:** We use NICE with

[4:10] **Voice 2:** Cogniggy for orchestration.

[4:12] **Voice 2:** It works, but

[4:16] **Voice 2:** voices can feel, they're robotic.

[4:20] **Voice 2:** Yeah. And when customers are already frustrated about

[4:24] **Voice 2:** device protection claims or warranties, that doesn't help.

[4:28] **Voice 2:** Technically, we are stable, but

[4:31] **Voice 2:** latency spikes happen during peak

[4:33] **Voice 2:** and

[4:35] **Voice 2:** integration touches are non

[4:37] **Voice 2:** We don't

[4:39] **Voice 2:** have appetite for heavy lift unless

[4:41] **Voice 2:** return is obvious.

[4:46] **Me:** That makes a lot of sense. So you're talking about

[4:46] **Me:** the

[4:51] **Voice 3:** reliability and the robotic voices. Could you talk a bit more about the impact of those issues?

[4:53] **Voice 3:** Why do those Chatter.

[4:58] **Voice 2:** When voices feel robotic, customers tend to

[5:02] **Voice 2:** repeat themselves or ask for a human that hits

[5:04] **Voice 2:** containment and increases cost per call.

[5:07] **Voice 2:** It also impacts CSAT.

[5:09] **Voice 2:** Quite directly.

[5:14] **Me:** And I guess just to double picture,

[5:14] **Voice 3:** obviously, we're we're familiar with

[5:16] **Voice 3:** CESAP, but why does that really matter?

[5:21] **Voice 2:** Because lower satisfaction drives complaints

[5:21] **Voice 2:** and churn.

[5:24] **Voice 2:** And our clients notice.

[5:29] **Voice 2:** Also, spikes during peak create longer handle times, so

[5:30] **Voice 2:** that's

[5:31] **Voice 2:** more cost.

[5:34] **Voice 2:** Even small deltas add up at our scale.

[5:39] **Voice 2:** And that's where we get nervous.

[5:43] **Voice 2:** We need clear ROI, not just a better vibe.

[5:49] **Me:** That makes sense. So

[5:52] **Me:** would you say one of the the the best ways to measure

[5:53] **Me:** success would be a significant increase in CSAT?

[5:56] **Voice 2:** Yes.

[5:56] **Voice 2:** But

[6:00] **Voice 2:** also look for improved containment.

[6:01] **Voice 2:** Without

[6:05] **Voice 2:** increasing escalations and stable handle times.

[6:07] **Voice 2:** And from my side,

[6:11] **Voice 2:** clear cost predictability and

[6:11] **Me:** Who who's speaking you?

[6:13] **Voice 2:** Priya.

[6:14] **Voice 2:** Procurement.

[6:16] **Voice 2:** Just thought you na

[6:16] **Voice 2:** early.

[6:20] **Me:** And so what were you saying again?

[6:25] **Voice 2:** We need a clear cost predictability.

[6:27] **Voice 2:** And risk mitigation.

[6:33] **Me:** Cost predictability and risk mitigation.

[6:37] **Me:** So that that that's a a

[6:39] **Voice 3:** pain or a success criteria.

[6:42] **Voice 2:** That's success criteria.

[6:43] **Voice 2:** And

[6:46] **Voice 2:** to be clear, if this is a rip and replace

[6:48] **Voice 2:** that's a much higher bar.

[6:53] **Voice 2:** We need to understand how this fits

[6:55] **Voice 2:** without introducing operational risk.

[7:00] **Me:** Yeah. I mean, very simply put, you know, we'll we'll get into

[7:02] **Me:** some of the details of our solution later. I'm really focused on on understanding

[7:05] **Me:** your concerns right now, but

[7:05] **Voice 3:** absolutely,

[7:08] **Voice 3:** by virtue of being natively compatible with your orchestration layer,

[7:13] **Voice 3:** it's simply a matter of changing a couple API endpoints. There's very little configuration

[7:18] **Me:** required. Perhaps we'll do a a a few, you know, special awards that

[7:20] **Me:** might be unique to your business to get those

[7:24] **Me:** working with model, but we have a a

[7:24] **Voice 3:** few ways of doing that very effectively.

[7:26] **Voice 3:** And, yeah, that's it's

[7:26] **Voice 3:** it

[7:30] **Me:** the actual getting a POC up and running is is incredibly

[7:32] **Me:** fast and low cost.

[7:36] **Voice 3:** Okay.

[7:39] **Me:** Any other pain points

[7:41] **Voice 3:** Marcus, I haven't heard a lot from you yet.

[7:46] **Voice 2:** Main thing is we can't

[7:47] **Voice 2:** afford downtime or surprises.

[7:51] **Voice 2:** If anything touches PII, security review

[7:52] **Voice 2:** gets deep fast.

[7:57] **Voice 2:** And we don't wanna manage multiple new vendors for

[7:58] **Voice 2:** niche pieces.

[8:03] **Voice 3:** Yeah. So would would you say that having a a sock

[8:05] **Me:** two, Type two solution that's hosted on prem would would help a lot with your security concerns?

[8:15] **Voice 2:** Yes. That helps. But my

[8:17] **Voice 2:** bigger concern is still operational risk.

[8:18] **Voice 2:** If

[8:21] **Voice 2:** we were to switch TTS providers, that is.

[8:26] **Me:** And can can you speak a bit more about that concern?

[8:30] **Voice 2:** We have tight SLAs.

[8:34] **Voice 2:** Outages or latency spikes have contractual

[8:38] **Voice 2:** implications. So any change has

[8:39] **Voice 2:** to be basically invisible or

[8:41] **Voice 2:** demonstrably safer.

[8:44] **Voice 3:** That makes a lot of sense.

[8:49] **Voice 3:** I, yeah, would be happy to share some of our SLAs and,

[8:52] **Voice 3:** reliability stats with you later. But

[8:56] **Voice 3:** certainly, what I can say is that by having an on prem

[8:56] **Me:** solution with

[8:58] **Me:** extreme predictability when it comes to

[8:59] **Me:** concurrency

[9:04] **Me:** limits and GPU requirements. It's a very well run model that's

[9:07] **Me:** already hosting, you know, tens of millions of calls every month. So there there's just no issues there.

[9:12] **Voice 2:** Okay.

[9:19] **Me:** Any other concerns or we haven't talked about yet?

[9:22] **Voice 2:** Just that we get pinched

[9:23] **Voice 2:** constantly.

[9:25] **Voice 2:** So

[9:27] **Voice 2:** we need specificity.

[9:28] **Voice 2:** Not

[9:29] **Voice 2:** generic AI promises.

[9:31] **Voice 3:** Yeah. You know,

[9:37] **Me:** if there's no other concern so so far, we we've talked about the outages and latency spikes

[9:39] **Me:** and the SLA requirements, the need for predictability.

[9:43] **Me:** The operational risk, security review, and

[9:45] **Me:** just not being able to afford downtime.

[9:50] **Me:** If there's nothing else you you wanted to add, I I can, hop into a quick demo.

[9:52] **Voice 2:** Yeah.

[9:54] **Voice 2:** Okay.

[9:55] **Voice 2:** Freeze is

[9:55] **Voice 3:** good.

[9:59] **Voice 3:** Let's try it live. Wish me luck.

[10:02] **Voice 3:** Alright.

[10:07] **Voice 4:** Hello, and thanks for calling NorthStar Device Protection. This is a demo, and the call may be

[10:07] **Voice 4:** recorded.

[10:09] **Voice 4:** I'm NorthStar's virtual claims assistant.

[10:11] **Voice 4:** How can I help you today?

[10:13] **Me:** Oh my gosh. I just broke my phone.

[10:19] **Voice 2:** Oh, yeah. That's, that's pretty naturalistic.

[10:25] **Voice 4:** Thank you, Tal. With your broken phone, I just need to verify your identity.

[10:27] **Voice 4:** Can you please provide your ZIP code?

[10:30] **Voice 3:** Yeah. 94105.

[10:35] **Voice 4:** Thanks for that, Colin.

[10:38] **Voice 4:** I see your policy is active and covers accidental damage.

[10:40] **Voice 4:** When the damage to your phone happened,

[10:45] **Me:** It happened July 11.

[10:48] **Voice 2:** Thanks, Colin.

[10:53] **Voice 4:** Was the damage accidental, liquid related, mechanical, or something

[10:56] **Me:** Oh my god, I was getting ready for this really big presentation to these amazing people, but

[11:00] **Me:** I was so excited I knocked my phone off the table and it it cracked the screen.

[11:01] **Me:** Totally an accident.

[11:05] **Voice 4:** Got it. Accidental damage.

[11:08] **Voice 4:** Does the phone still power on and work for you?

[11:08] **Me:** It

[11:08] **Voice 3:** does.

[11:13] **Voice 4:** Great. And do you still have the device with you?

[11:14] **Voice 4:** I

[11:14] **Voice 3:** do.

[11:27] **Me:** Yep.

[11:35] **Voice 3:** What should I do?

[11:40] **Voice 4:** You'll need to keep the device. Yes.

[11:45] **Voice 4:** Just to confirm, your iPhone 16 Pro had an accidental drop on July 11.

[11:49] **Voice 4:** Screen cracked, but it still powers on, and you have the device with you.

[11:51] **Voice 2:** Yes. Is that

[11:51] **Voice 4:** correct?

[11:51] **Me:** Uh-huh?

[12:07] **Voice 4:** Your claim number is CLMDemo2049 The deductible for screen repair is $29 I can send a secure link to upload any documents needed to the phone number we have on file.

[12:09] **Voice 4:** Would you like me to send that now?

[12:13] **Voice 4:** Sure. Or would you prefer to talk to a device

[12:13] **Me:** claims specialist?

[12:13] **Voice 3:** I'll talk to a specialist as well.

[12:21] **Voice 4:** You're not connected to a device claim specialist who has all your claim details.

[12:23] **Voice 4:** They'll assist you further from here.

[12:25] **Me:** Hi, Colin.

[12:27] **Me:** This is the device claim specialist here.

[12:32] **Voice 5:** I understand you have claimed CLM demo 2049 for accidental

[12:33] **Voice 5:** screen damage.

[12:38] **Voice 5:** And I'm here to assist you with any questions have

[12:38] **Voice 6:** available. Mais est-ce

[12:38] **Me:** que vous pouvez le mettre en français

[12:40] **Voice 2:** Bien ça.

[12:42] **Voice 2:** Les pouvons continuer en français.

[12:47] **Voice 5:** Merci Colin.

[12:56] **Voice 4:** Si vous voulez parler directement à un conseiller en français, je peux vous transférer.

[12:59] **Me:** C'est c'est gentil, mais mais c'est ce que j'ai besoin. Bon journée.

[13:04] **Voice 4:** Très bien Colin. Merci d'avoir appelé le

[13:12] **Me:** to what you have now?

[13:16] **Voice 2:** Honestly,

[13:18] **Voice 2:** smoother than I expected.

[13:20] **Voice 2:** Less robotic.

[13:21] **Voice 2:** But

[13:24] **Voice 2:** I'm curious how it performs at scale under peak.

[13:27] **Voice 2:** From a CX lens, tone felt

[13:30] **Voice 2:** calmer, less robotic. That's good.

[13:35] **Voice 2:** I'd wanna see how customers actually respond in our

[13:35] **Voice 2:** context.

[13:37] **Voice 2:** And for me, it

[13:40] **Voice 2:** still comes down to commercials and risk.

[13:45] **Voice 2:** Sure.

[13:45] **Me:** Let me address Let address all of

[13:45] **Voice 3:** these.

[13:49] **Me:** Just so for context, this is something I made myself in just a few minutes. I'm I'm not

[13:54] **Me:** part of the engineering team. So, obviously, a production grade

[13:54] **Voice 3:** system would be a lot more sophisticated.

[13:59] **Me:** I used one of our our many voices named Vespera because

[14:02] **Me:** has more formal, professional, and reassuring tone. If someone just broke their phone, obviously,

[14:07] **Me:** that's quite important. That said, we even have different voices that

[14:08] **Me:** some customers use for folks from different parts of the country and

[14:12] **Me:** can do a lot a lot of flexibility there even based on gender.

[14:16] **Me:** So it's it's just a starting point. But

[14:18] **Me:** if you have some more questions, now might be a great time.

[14:22] **Voice 2:** Makes sense.

[14:27] **Voice 2:** Yeah. I'm I'm mostly thinking about real world performance and measure

[14:37] **Me:** Well, yeah, you know, as I say, we

[14:40] **Me:** were some of the largest brands in the country are already using our our systems.

[14:44] **Me:** We have some case studies that I'd be happy to to share with you.

[14:49] **Voice 3:** You know, did I talk about the fact that we did a 100,000 calls recently

[14:49] **Me:** and they all work

[14:55] **Voice 2:** Yes. That makes sense as a data point.

[14:59] **Voice 2:** I think the next step for us is probably to

[15:01] **Voice 2:** define what success would look like internally before getting too far.

[15:06] **Me:** Yeah, I mean, that's generally how we think about a POC

[15:09] **Me:** So, normally, we would arrange a a call to align on the scope and success criteria.

[15:14] **Me:** And then, start to,

[15:14] **Voice 3:** to pilot on one live call flow in in limited production.

[15:19] **Me:** Once that runs for for a couple weeks, we would reconvene to resume

[15:24] **Me:** review the results and really just measure success against

[15:27] **Me:** not just does it sound nice, but the things that really

[15:27] **Voice 2:** matter, like

[15:27] **Me:** call completion, C SCAT, or customer call sentiment.

[15:29] **Me:** Right.

[15:32] **Voice 2:** That sounds like a reasonable framing.

[15:35] **Voice 2:** We need to align internally first.

[15:39] **Voice 2:** I'm not making any commitments today, but

[15:41] **Voice 2:** sure we can consider it.

[15:46] **Me:** Sure. No, that makes a lot of sense. You know, we do have quite a few of these

[15:49] **Me:** on the books, though. It it would be great just to to carve out time in case

[15:52] **Me:** perhaps in, a week from now just to pencil that in.

[15:57] **Me:** Can we do that? And, that'll give you some time to align in turn

[15:57] **Me:** and see if that that date works or if we need to move things around.

[16:02] **Voice 2:** I think pencil

[16:04] **Voice 2:** thirty minutes is reasonable.

[16:07] **Voice 2:** But

[16:07] **Me:** Yeah. No. That that that would be great.

[16:07] **Voice 7:** Folks there.

[16:10] **Voice 3:** Absolutely. Who do you think are the decision makers that should be there?

[16:16] **Voice 2:** I'll start.

[16:19] **Voice 2:** Own CX metrics, so I'd be involved.

[16:23] **Voice 2:** And we bring in someone from our platform architecture team.

[16:27] **Voice 2:** And procurement if there's a real business case in Rochang.

[16:30] **Voice 2:** Yes, that works.

[16:31] **Voice 2:** Send a couple of times.

[16:34] **Voice 3:** I'll I'll absolutely do that.

[16:37] **Me:** You know, this was a a real pleasure.

[16:42] **Me:** If you need any more information, everything's here on the website.

[16:43] **Me:** And

[16:45] **Voice 6:** I

[16:45] **Me:** really appreciate you guys taking the time today.

[16:48] **Voice 2:** Sounds good.

[16:49] **Voice 2:** For walking through it.

[16:51] **Me:** Alright. Ency.

[16:53] **Voice 2:** Noted.