WEBVTT

1
00:00:00.000 --> 00:00:01.376
I did know I wanted to be a founder.

2
00:00:01.459 --> 00:00:05.209
I actually thrive in situations where
I'm working on five things at a time.

3
00:00:07.296 --> 00:00:09.756
I'm Sam Hogan, I'm
a co-founder of Searchable.

4
00:00:09.839 --> 00:00:15.510
I became a founder right on the cusp
of when AI guided software became a thing.

5
00:00:15.593 --> 00:00:22.140
Searchable was built from the ground up
with a huge focus on agent infrastructure.

6
00:00:22.223 --> 00:00:27.226
Analytics can be super bulky,
so we utilize Vercel in a handful of ways.

7
00:00:27.309 --> 00:00:30.059
So obviously we have to have very
lean developer teams.

8
00:00:30.142 --> 00:00:34.425
So quick build times and fast deployments,
new customers request new features.

9
00:00:34.508 --> 00:00:37.608
Can we ship that in 30 minutes
and really blow someone's mind?

10
00:00:37.691 --> 00:00:40.312
What would really cost a lot
in man hours on managing

11
00:00:40.395 --> 00:00:42.545
and trying to build
something from scratch.

12
00:00:42.628 --> 00:00:45.322
The AI SDK's been super,
super helpful for us, right?

13
00:00:45.405 --> 00:00:49.568
As we launched the AI Gateway
actually came out too and being able

14
00:00:49.651 --> 00:00:54.244
to have, you know, really fast evaluation
periods on what models are performing

15
00:00:54.327 --> 00:00:58.539
extremely well, especially in AI search
and SEO has been really helpful for us.

16
00:00:58.622 --> 00:01:03.684
And the headache of rotating between
different SDKs or different API keys

17
00:01:03.767 --> 00:01:07.420
and having, you know, like building
a ton of different fallbacks.

18
00:01:07.503 --> 00:01:09.253
I think it does come down to speed.

19
00:01:09.336 --> 00:01:13.799
I mean, it's got to be 2 to 5x at least,
because feedback, when you have scale,

20
00:01:13.882 --> 00:01:15.582
you need to be able to fix things.

21
00:01:15.665 --> 00:01:21.161
You need to be able to identify bugs and
solve things in a manner that slow moving

22
00:01:21.244 --> 00:01:23.347
growth just doesn't provide.

23
00:01:23.430 --> 00:01:27.391
And there's going to be what we believe
to be a closed loop, in AI search,

24
00:01:27.474 --> 00:01:30.434
the autonomous aspect
of how someone can show

25
00:01:30.517 --> 00:01:33.062
up in ChatGPT and Google and Perplexity.

26
00:01:33.145 --> 00:01:36.945
I think the biggest thing, and it
was actually a mistake we made early days,

27
00:01:37.028 --> 00:01:41.067
is not tracking like
every possible data point.

28
00:01:41.150 --> 00:01:45.798
Like if you're running paid media or if
you're if you have an onboarding funnel,

29
00:01:45.881 --> 00:01:51.072
you need to know the literal user
path of every single person who enters

30
00:01:51.155 --> 00:01:53.955
your funnel and also
the identification of who they are.

31
00:01:54.038 --> 00:01:56.948
So with volume comes
the requirement for flexibility

32
00:01:57.031 --> 00:02:01.225
and being able to move quick and on your
feet. And Vercel has done that for us.
