WEBVTT

NOTE Sentence-level transcript of https://www.youtube.com/watch?v=pyvRID_CZZU

NOTE One cue per sentence. Cue ids are the line anchors on /transcripts/pyvRID_CZZU.html. A cue ends where the next begins, or 2 s after its last word.

s1
00:00:00.000 --> 00:00:01.309
Cool.

s2
00:00:01.309 --> 00:00:03.309
[music]

s3
00:00:13.920 --> 00:00:14.639
All righty.

s4
00:00:14.639 --> 00:00:15.679
Can everyone hear me?

s5
00:00:15.679 --> 00:00:16.160
Okay.

s6
00:00:16.160 --> 00:00:17.760
All All levels coming through good.

s7
00:00:17.760 --> 00:00:18.160
All right.

s8
00:00:18.160 --> 00:00:18.640
Sweet.

s9
00:00:18.640 --> 00:00:22.400
Well, uh, how's everyone doing?

s10
00:00:23.439 --> 00:00:25.439
How's everyone doing?

s11
00:00:25.439 --> 00:00:25.840
Better.

s12
00:00:25.840 --> 00:00:26.640
Okay, there we go.

s13
00:00:26.640 --> 00:00:27.119
Cool.

s14
00:00:27.119 --> 00:00:30.080
Well, I hope uh the conference has been good for everybody.

s15
00:00:30.080 --> 00:00:33.760
Appreciate you all being here to to listen to a chat from uh from me.

s16
00:00:33.760 --> 00:00:37.840
Um for anyone who hasn't met me at the at the booth yet, uh my name is Matt Lawler.

s17
00:00:37.840 --> 00:00:40.480
I'm a forward deployed engineer at Assembly AI.

s18
00:00:40.480 --> 00:00:47.920
Uh mostly focused on bringing in inbound customers, um helping win their business, u making sure that they have a good time building with our API.

s19
00:00:47.920 --> 00:00:52.239
Um so, how how many other like forward deployed engineers are out in the audience right now?

s20
00:00:52.239 --> 00:00:54.239
Any of us?

s21
00:00:54.239 --> 00:00:55.600
Okay, a couple.

s22
00:00:55.600 --> 00:00:56.079
Cool.

s23
00:00:56.079 --> 00:00:58.640
Well, um, still coming through.

s24
00:00:58.640 --> 00:00:59.760
Okay.

s25
00:00:59.760 --> 00:01:00.320
All right.

s26
00:01:00.320 --> 00:01:01.199
Um, yeah.

s27
00:01:01.199 --> 00:01:12.560
So, I wanted to make sure I I gave you guys an overview of kind of some of the go to market engineering stuff that we've been doing at Assembly AI that maybe you can apply to your own FTE team or maybe you can apply to your company um, as something that you can do as an easy win

s28
00:01:12.560 --> 00:01:13.520
for your customers.

s29
00:01:13.520 --> 00:01:20.080
Um, so for a little bit of overview of like assembly AI in case you haven't heard of us, um, we provide voice AI infrastructure.

s30
00:01:20.080 --> 00:01:23.680
Um, so we train our own foundation models for speech to text primarily.

s31
00:01:23.680 --> 00:01:27.040
Um, but we're starting to branch into other parts of the voice agent space.

s32
00:01:27.040 --> 00:01:34.320
Um, so if anyone's ever had like a Fireflies note-taker join your meeting or something like that, you've also you've already seen a transcript from Assembly AI.

s33
00:01:34.320 --> 00:01:43.680
Um, so we train our own foundational models not only for like async use cases like meeting recordings and things like that, but also for real time uh real-time like voice agents.

s34
00:01:43.680 --> 00:01:53.280
Um, so if you've ever used like a a phone customer support agent when you called in somewhere and it wasn't a human or like drive-through ordering, they're more than likely building with Assembly AI's models.

s35
00:01:53.280 --> 00:01:58.719
Um, and for anyone who's building a voice agent now, you can use us for the entire stack, not just speech to text.

s36
00:01:58.719 --> 00:02:04.880
Um, so you can use our voice agent API and we'll string together the speech to text, the LLM, and the text to speech as well.

s37
00:02:04.880 --> 00:02:12.720
Um, as our competitors know, as we know, as anyone in the voice AI space knows, demand is absolutely insane um, for voice AI right now.

s38
00:02:12.720 --> 00:02:16.959
Um, so assembly at our current moment, we see around a thousand API signups every single day.

s39
00:02:16.959 --> 00:02:27.440
Um, which is somewhat of the theme of why we were focusing on automation and wanting to build kind of a new support bot for our team because at least until like two days ago, I was the only

s40
00:02:27.440 --> 00:02:29.520
uh onboarding engineer at Assembly AI.

s41
00:02:29.520 --> 00:02:35.519
And so you can imagine, as much as I'd want to talk to a thousand new signups every single day, that's just not feasible.

s42
00:02:35.519 --> 00:02:39.200
Um, so we had to build something to to solve that problem.

s43
00:02:39.200 --> 00:02:43.599
And then for anyone here who doesn't know like what a forward deployed engineer is, it's a very catch-all title.

s44
00:02:43.599 --> 00:02:48.319
It varies a lot at every different company, but effectively we want to get deployed with our end customers.

s45
00:02:48.319 --> 00:02:49.840
We want to understand their use case.

s46
00:02:49.840 --> 00:02:52.720
We want to make sure that we are building exactly what they need.

s47
00:02:52.720 --> 00:02:56.239
Um, so we frequently commit code directly to their repositories.

s48
00:02:56.239 --> 00:02:59.599
We usually meet with them several times a month, if not a week.

s49
00:02:59.599 --> 00:03:03.120
Um, and we really try to understand their use case and get to know them very personally.

s50
00:03:03.120 --> 00:03:13.040
Um so like at least on the inbound side for anyone who comes into assembly who wants to build with us we are the people um that are their main point of contact for anything they need on the technical commercial side

s51
00:03:13.040 --> 00:03:20.800
anything that they might need they get from us their FTE um and we're responsible for winning their business um and building that trust with them.

s52
00:03:20.959 --> 00:03:22.480
So it's a great services model.

s53
00:03:22.480 --> 00:03:33.040
It's great uh if we can actually get that kind of relationship with every customer, but again with like a thousand API signups every day, I I can't be one to many serving that many customers and getting on that many calls.

s54
00:03:33.040 --> 00:03:35.920
Um and so ultimately this doesn't really scale super well.

s55
00:03:35.920 --> 00:03:43.200
Uh we want to make sure that we're freeing ourselves up to be leading uh humanled deals as much as possible, but we can't do that with every deal.

s56
00:03:43.200 --> 00:03:45.920
Not every customer win can be like handbuilt by us.

s57
00:03:45.920 --> 00:03:50.799
Um and we can't just continue to hire more and more FTEEs if we're going to try and serve more and more customers.

s58
00:03:50.799 --> 00:04:02.239
Um, so ultimately like for anyone who's an FTE in here, if you're looking to become an FTE, if you already have an FTE team at your company, I would encourage you to take that knowledge that you have as an FTE and try to automate yourself out of a job.

s59
00:04:02.239 --> 00:04:03.920
I think that should be your role every single day.

s60
00:04:03.920 --> 00:04:12.560
If you want to serve your customers better, if you want to work with them more, um, and if you want to deliver a better customer experience at scale, you can't be the bottleneck to having a good customer experience.

s61
00:04:12.560 --> 00:04:15.360
And we learned that firsthand.

s62
00:04:15.439 --> 00:04:18.720
So again, we were drowning in like inbound um, volume.

s63
00:04:18.720 --> 00:04:23.360
Um, you can see like we basically had a thousand API signups up there earlier.

s64
00:04:23.360 --> 00:04:28.080
Um, and we had the natural instinct that a lot of teams do where we were like, we just need to deflect more conversations.

s65
00:04:28.080 --> 00:04:29.360
I just want to talk to fewer people.

s66
00:04:29.360 --> 00:04:32.400
I want to make sure I'm freeing myself up to work on more important stuff.

s67
00:04:32.400 --> 00:04:36.800
Um, so we basically bought this off-the-shelf support bot and we're like, let's point it at the docs.

s68
00:04:36.800 --> 00:04:38.479
It'll answer all the docs questions for us.

s69
00:04:38.479 --> 00:04:39.120
It's going to be great.

s70
00:04:39.120 --> 00:04:41.919
We'll never have to answer a simple question from customers again.

s71
00:04:41.919 --> 00:04:44.320
And it only resolved about 10% of the conversations.

s72
00:04:44.320 --> 00:04:45.520
Um, so it helped.

s73
00:04:45.520 --> 00:04:50.560
Um, you can imagine if we had like a thousand conversations a day, a hundred of those were now entered by a bot.

s74
00:04:50.560 --> 00:04:53.199
But then our team still had to field 900 tickets a day.

s75
00:04:53.199 --> 00:04:55.440
So it wasn't exactly very scalable.

s76
00:04:55.440 --> 00:04:59.680
Um, so 10 10% uh end to end resolution rate wasn't good enough.

s77
00:04:59.680 --> 00:05:03.120
And ultimately we couldn't iterate fast enough to be able to fix this.

s78
00:05:03.120 --> 00:05:05.040
We didn't have access to the system prompt.

s79
00:05:05.040 --> 00:05:06.800
We didn't have access to the tools.

s80
00:05:06.800 --> 00:05:08.880
We didn't have access to the rag infrastructure.

s81
00:05:08.880 --> 00:05:14.080
So anytime we needed to change something, we'd go to this vendor and they'd say, "Well, it's on the road map."

s82
00:05:14.080 --> 00:05:15.600
and that didn't quite work for us.

s83
00:05:15.600 --> 00:05:18.160
So, we wanted to build our own.

s84
00:05:18.160 --> 00:05:23.039
So, we built uh another member of our team rather than hiring one and we named him Joey.

s85
00:05:23.039 --> 00:05:31.280
Um so, anyone now who reaches out to us, whether it's on our chat widget on our site, whether you reach out to contact sales, whether you reach out to our support alias,

s86
00:05:31.280 --> 00:05:33.280
you're going to meet Joey instead of a human.

s87
00:05:33.280 --> 00:05:37.680
Um he's the first line of defense now for every single conversation that happens in bounded assembly.

s88
00:05:37.680 --> 00:05:41.600
And he's built to be the most infinitely scalable FTE on our team.

s89
00:05:41.600 --> 00:05:44.160
Um, so he remembers past conversations with you.

s90
00:05:44.160 --> 00:05:52.080
Um, we track all of his metrics via Pylon because he's still a member of our support team and so he has access to Pylon to manage conversations and remember who you are.

s91
00:05:52.080 --> 00:05:54.400
Um, and he's built on top of the cloud agent SDK.

s92
00:05:54.400 --> 00:05:58.400
And so the cool things that he's able to do is he can manage his own infrastructure.

s93
00:05:58.400 --> 00:05:59.840
He has a file system.

s94
00:05:59.840 --> 00:06:01.919
He can write and debug code for you.

s95
00:06:01.919 --> 00:06:03.039
He can call tools.

s96
00:06:03.039 --> 00:06:04.800
He can give himself new abilities.

s97
00:06:04.800 --> 00:06:13.759
Um, and so he's actually able to operate more like an FTE and less like just a standard chatbot that you'd be talking with, you know, on on most major sites nowadays.

s98
00:06:14.479 --> 00:06:21.120
So, um, if you want to know a little bit of like how to build your own Joey, um, the architecture kind of has like four major parts.

s99
00:06:21.120 --> 00:06:25.280
Um, so as everyone might know from working with Claude, Markdown is your friend.

s100
00:06:25.280 --> 00:06:29.520
You should be giving Markdown files to Claude so that he understands everything about your business.

s101
00:06:29.520 --> 00:06:33.759
And so we wanted him to have access to all of our documentation and be trained up on it all the time.

s102
00:06:33.759 --> 00:06:37.600
So, we checked all of our docs out as local markdown and gave it to Joey.

s103
00:06:37.600 --> 00:06:44.800
Um, and so anytime that we push an update to our documentation system, he's able to get all of those files and he has them locally in his file system.

s104
00:06:44.800 --> 00:06:54.639
So, actually what's really cool is if our entire doc site was down right now, you could go to Joey and still get live answers on all the newest stuff that we've shipped without our team having to actually get involved.

s105
00:06:54.639 --> 00:06:56.479
um for retrieval so we can give you a good answer.

s106
00:06:56.479 --> 00:07:04.800
We partner with Voyage so we use their embeddings um so that he gets the most relevant documents up front so he doesn't have to spend a lot of time searching and he can get you a faster answer.

s107
00:07:04.800 --> 00:07:16.880
And then since he has all these docs as local files if he gets confused or he needs to string together multiple resources to create a response he can agentically search the file system that he lives on so that he can go and find those resources for you.

s108
00:07:16.880 --> 00:07:23.199
Um so he can kind of go into this like deep thinking mode where he's able to search across all these resources and then we shipped it on railway.

s109
00:07:23.199 --> 00:07:24.560
We didn't want to deal with infrastructure.

s110
00:07:24.560 --> 00:07:25.840
We didn't want to spin up an EC2.

s111
00:07:25.840 --> 00:07:27.360
We didn't want to have to change instance types.

s112
00:07:27.360 --> 00:07:29.599
So, we deployed it all on Railway.

s113
00:07:29.599 --> 00:07:31.280
Their service is pretty incredible.

s114
00:07:31.280 --> 00:07:35.440
Basically, we can just go ahead and see if he has a bad conversation with someone in Slack.

s115
00:07:35.440 --> 00:07:40.240
We can write a PR real quick and we can deploy it and within 30 seconds there's a new live version of Joey.

s116
00:07:40.240 --> 00:07:42.240
So, we're able to iterate really, really quickly.

s117
00:07:42.240 --> 00:07:53.199
Uh we've actually had times where we've actually monitored a live conversation, caught a bug, deployed a fix, and then it's been proof for the rest of the session, and that customer doesn't even have to know that we we shipped that fix behind the scenes.

s118
00:07:53.199 --> 00:07:56.560
Um so Joey is able to to basically get updated whenever we want.

s119
00:07:56.560 --> 00:08:01.520
Um so we actually have full control over everything that that he does.

s120
00:08:01.840 --> 00:08:04.720
Um and so the results were pretty remarkable from us building this.

s121
00:08:04.720 --> 00:08:09.120
And it's another reason why and wanted to encourage everyone who's here to try and build your own Joey.

s122
00:08:09.120 --> 00:08:16.879
So, we went from 10% to 80% endto-end resolution rate in just the first week of deploying this this build with a pretty naive implementation.

s123
00:08:16.879 --> 00:08:22.400
And we did that all for around $700 a month in both token and inference uh or infrastructure costs.

s124
00:08:22.400 --> 00:08:27.360
So, we resolved 80 80% of our our tickets end to end with no human in the loop.

s125
00:08:27.360 --> 00:08:29.599
And some people like to game these metrics.

s126
00:08:29.599 --> 00:08:34.159
They'll say like, oh, we resolved 80% but only on like a subset of types of tickets.

s127
00:08:34.159 --> 00:08:35.279
Not the case here.

s128
00:08:35.279 --> 00:08:38.800
you cannot reach a human directly unless you go through Joey and you tell him to escalate.

s129
00:08:38.800 --> 00:08:42.800
So he actually does handle 100% of these inbound tickets for us now.

s130
00:08:42.800 --> 00:08:45.839
Um and he only escalates 20% of them to a human for the right reasons.

s131
00:08:45.839 --> 00:08:47.120
Maybe it's a rate change.

s132
00:08:47.120 --> 00:08:48.399
Maybe it's a data opt out.

s133
00:08:48.399 --> 00:08:54.399
Maybe you need to sign an agreement with us or you have needs that only an FTE or someone from legal needs to be involved in.

s134
00:08:54.399 --> 00:08:57.760
These are things that actually still need a human but eventually we can automate.

s135
00:08:57.760 --> 00:09:04.240
Um and so the best part is whatever Joey can't do yet gives us a very clear list of what we need to do to continue to scale our team.

s136
00:09:04.240 --> 00:09:09.680
Um, if he's not able to send them a BAA, well, okay, maybe we should give him a link that he can reference as a resource.

s137
00:09:09.680 --> 00:09:14.240
Um, if he's not able to talk to customers about pricing, well, you can negotiate with Joey.

s138
00:09:14.240 --> 00:09:18.720
Um, if you go on the site and you test him out afterwards, you can tell him exactly how many hours you have.

s139
00:09:18.720 --> 00:09:22.399
Um, and he'll actually quote you a rate and then you can negotiate with him.

s140
00:09:22.399 --> 00:09:29.600
Um, so we tried to automate a lot of these other things that he was still escalating and there's still there's still plenty more to come.

s141
00:09:29.600 --> 00:09:34.959
But the real thing that, you know, we wanted to test out by building Joey was we wanted to dog food our own product, too.

s142
00:09:34.959 --> 00:09:38.399
We didn't want to just like build something that deflects more stuff from our customers.

s143
00:09:38.399 --> 00:09:43.680
We didn't want to build something that just um is kind of this general support bot that anyone can build.

s144
00:09:43.680 --> 00:09:46.480
We work in voice and our customers build voice agents.

s145
00:09:46.480 --> 00:09:49.680
So, I wanted to build a voice agent and so I wanted to give a voice to Joey.

s146
00:09:49.680 --> 00:09:55.120
So, now rather than just chatting with him via text, you can actually chat with him using your voice as well.

s147
00:09:55.120 --> 00:09:57.839
Um, so we have a voice agent API as I mentioned earlier.

s148
00:09:57.839 --> 00:09:59.920
Um, and we integrated this into Joey this week.

s149
00:09:59.920 --> 00:10:04.720
So now if you go onto our site, you can switch to voice mode and you can talk to Joey using just your voice.

s150
00:10:04.720 --> 00:10:15.519
Um, because I ultimately think for any FTE here, no matter how much you've been deployed with your customers, no matter how embedded you are with them, the best way to understand what your customers are building is to actually build their same product yourself.

s151
00:10:15.519 --> 00:10:19.120
And that's exactly what we wanted to do with this.

s152
00:10:19.200 --> 00:10:22.000
So we put our voice agent API into Joey.

s153
00:10:22.000 --> 00:10:30.160
Um, it's all speech in, speech out, one websocket connection where we string together our speechto text model, an LLM and text to speech, all optimized for voice.

s154
00:10:30.160 --> 00:10:31.760
It's real-time and low latency.

s155
00:10:31.760 --> 00:10:36.240
It handles all the pauses, interruption handling, barge, all of that kind of stuff for you.

s156
00:10:36.240 --> 00:10:39.279
U, we bring our own voices, so you don't have to find another provider.

s157
00:10:39.279 --> 00:10:42.640
And it's all the same tools that he already had if you were communicating with him via text.

s158
00:10:42.640 --> 00:10:45.120
It's just now you can actually talk to him like you were a human.

s159
00:10:45.120 --> 00:10:47.600
Um, and hopefully later this week we can hook them up to a phone number.

s160
00:10:47.600 --> 00:10:54.240
So now we'll have an Assembly AI phone number so you can get real-time support from Joey whenever you need it.

s161
00:10:54.480 --> 00:10:58.800
So let's go ahead and talk to Joey a little bit and we can walk through some of the architecture live.

s162
00:10:58.800 --> 00:11:08.880
Um so for anyone who wants to know a little bit more about how this was actually built um we basically have all of the the docs checked out locally in Markdown.

s163
00:11:08.880 --> 00:11:14.399
Um so every single time someone puts an update out to our docs, these files also get locally checked out into this repo.

s164
00:11:14.399 --> 00:11:24.000
So they stay perfectly in sync with any new features that we're shipping, our change log, our pricing, all of these major uh pages from our website get converted to markdown and get given to Joey.

s165
00:11:24.000 --> 00:11:33.440
Um so he sees all of these in markdown format and he's able to see what the URL is so he can site his sources to the customer in case they want to double check his work or if they just want more information.

s166
00:11:33.440 --> 00:11:37.600
Um at the same time, um all of it is provided as like Claude MD instructions.

s167
00:11:37.600 --> 00:11:44.000
I think it's like a 30,000 line Claude MD of all these guard rails and and advice on how he should be operating with customers.

s168
00:11:44.000 --> 00:11:54.560
Um, and this is what we majorly update whenever we see that he had a bad experience with a customer or gave a wrong answer, we ship an update to this cloud MD and now for every conversation that he has going forward, he's able to be better.

s169
00:11:54.560 --> 00:11:56.880
Um, and it's all contained as a Docker file.

s170
00:11:56.880 --> 00:12:00.480
Um, and so it's really easy to go and just containerize and deploy him on Railway.

s171
00:12:00.480 --> 00:12:07.839
Um, and so again, every time we ship an update to this repo, a new version's already live on Railway and no one had to manually monitor the deployment.

s172
00:12:07.839 --> 00:12:11.040
Um, so let's go ahead and uh I'll fire up Joey here.

s173
00:12:11.040 --> 00:12:16.240
we can ask him a pretty typical question so you can see what our voice agent API is like.

s174
00:12:18.399 --> 00:12:25.040
So I'm going to simulate this locally because I I know probably many of you are going to test Joey after this and I don't want to add to the queue for our actual support team.

s175
00:12:25.040 --> 00:12:28.639
Um so I'll I'll go ahead and just do a local simulation of it.

s176
00:12:28.639 --> 00:12:32.959
But typically, yeah, you would go into here, you'd meet Joey, you'd want to have a typical question.

s177
00:12:32.959 --> 00:12:37.200
You might talk to him via text, but let's go ahead and talk to him with voice.

s178
00:12:37.200 --> 00:12:39.680
Hi, this is Joey at Assembly AI.

s179
00:12:39.680 --> 00:12:41.360
How can I help you today?

s180
00:12:41.360 --> 00:12:47.200
Hey Joey, I'm looking to build an ambient medical scribe with Assembly AI, but before I can test, I need to assign a BAA.

s181
00:12:47.200 --> 00:12:50.160
Is that something that you can do?

s182
00:12:51.680 --> 00:12:52.720
One sec.

s183
00:12:52.720 --> 00:12:55.680
Let me check that for you.

s184
00:13:06.639 --> 00:13:09.839
Yes, you can execute the BAA yourself right away.

s185
00:13:09.839 --> 00:13:17.360
Just complete our standard HIPPA business associate agreement, but note that you'll need a paid account with a card on file first.

s186
00:13:17.360 --> 00:13:21.360
And signing the BAA automatically opts you out of model training.

s187
00:13:21.360 --> 00:13:23.839
The relevant docs are linked in the chat.

s188
00:13:23.839 --> 00:13:24.240
Cool.

s189
00:13:24.240 --> 00:13:24.800
Thanks, Joey.

s190
00:13:24.800 --> 00:13:26.720
Well, I got to I got to get back to my demo.

s191
00:13:26.720 --> 00:13:30.720
I'd love to chat more, but uh thanks for all the help.

s192
00:13:32.000 --> 00:13:33.600
You are very welcome.

s193
00:13:33.600 --> 00:13:37.519
Good luck with your demo and feel free to reach out if you need anything else later.

s194
00:13:37.519 --> 00:13:39.360
Have a great day.

s195
00:13:39.360 --> 00:13:40.724
Isn't he so nice?

s196
00:13:40.724 --> 00:13:41.600
[laughter]

s197
00:13:41.600 --> 00:13:43.200
I enjoy talking to Joey.

s198
00:13:43.200 --> 00:13:45.440
I think our customers will enjoy talking to Joey.

s199
00:13:45.440 --> 00:13:49.600
But this is a great example of like two years ago when I first joined assembly on our support team.

s200
00:13:49.600 --> 00:13:52.079
This is a conversation that would have to be handled by a human.

s201
00:13:52.079 --> 00:13:59.279
This is something that if you roached out in in chat and a customer uh or a support engineer wasn't able to uh answer you immediately.

s202
00:13:59.279 --> 00:14:00.000
They'd have to wait.

s203
00:14:00.000 --> 00:14:01.360
They'd have to get a response from us.

s204
00:14:01.360 --> 00:14:02.880
It was a canned response.

s205
00:14:02.880 --> 00:14:04.800
Um it was just not a great customer experience.

s206
00:14:04.800 --> 00:14:08.800
And now you have an FTE that you can talk to 24/7 for these basic questions.

s207
00:14:08.800 --> 00:14:10.240
Um and you can do it through voice.

s208
00:14:10.240 --> 00:14:12.399
So you can actually test our product at the same time.

s209
00:14:12.399 --> 00:14:18.480
Um so there's there's kind of a meta layer there of where if you're talking to Joy about how to build a voice agent API, you're already using it.

s210
00:14:18.480 --> 00:14:23.279
Um, and you're already seeing exactly what your product experience could look like.

s211
00:14:25.519 --> 00:14:29.279
So, if there's anything that you should get out of this talk, it's not just that you should build your own Joey.

s212
00:14:29.279 --> 00:14:34.399
Um, but I think that it's like if you're if you're an FTE at your company, you need to be dog fooding your own product all the time.

s213
00:14:34.399 --> 00:14:37.440
It's not enough to just be teaching your customers how to get started with it.

s214
00:14:37.440 --> 00:14:48.079
any new feature that you launch, any new product that you offer, you need to be teaching your customers, um, with all the knowledge that you've built from actually using and building and trying to ship the same product.

s215
00:14:48.079 --> 00:14:49.760
Um, so I had to run through all the same walls.

s216
00:14:49.760 --> 00:14:51.279
I had to figure out how to deal with latency.

s217
00:14:51.279 --> 00:14:55.440
I had to figure out how to deal with turn taking or bargain or interruption handling.

s218
00:14:55.440 --> 00:15:00.720
Um, and I think I have a lot better empathy of how to like work with customers that you want to use our voice agent API.

s219
00:15:00.720 --> 00:15:02.560
Um, because now I've actually used it.

s220
00:15:02.560 --> 00:15:03.680
I've used the same API.

s221
00:15:03.680 --> 00:15:10.240
It's the exact one that my customers build on and I can give them better advice because I've actually shipped something with it.

s222
00:15:10.320 --> 00:15:12.240
So go build go build something.

s223
00:15:12.240 --> 00:15:19.360
Go build whether it's a your own Joey, go build whether it's a voice agent, especially if it's a voice agent, build it on Assembly AI.

s224
00:15:19.360 --> 00:15:25.440
Um, but if you're if you're actually working in an FTE capacity, try to build the exact product that your customers are building and do it on your own.

s225
00:15:25.440 --> 00:15:27.440
Don't just wait for them to ask you for help.

s226
00:15:27.440 --> 00:15:29.279
Don't just wait to remove all their blockers.

s227
00:15:29.279 --> 00:15:31.519
You should be going out and building the exact same thing.

s228
00:15:31.519 --> 00:15:35.920
And at the same time, you might actually be able to make your customer experience better simultaneously.

s229
00:15:35.920 --> 00:15:42.639
So, um, if anyone has like questions on how to build Joey, if you want to check them out at the booth afterwards, our booth is just right over here.

s230
00:15:42.639 --> 00:15:51.680
Um, if anyone has feedback on Joey since this just shipped this week, if you have a weird conversation with him, I'd love to know about it so we can keep making fixes and I can show you how we deploy that live.

s231
00:15:51.680 --> 00:15:56.480
U, but overall, really excited to uh, be here at the conference and talk to more people about voice agents.

s232
00:15:56.480 --> 00:16:00.560
So, thanks for uh, coming and listen to this talk.

s233
00:16:00.896 --> 00:16:02.896
[applause]

s234
00:16:16.616 --> 00:16:18.616
[music]