0:00
Bye. Bye. Hey, I'm Brian Teller. I work in DevOps
0:11
and SRE, and I run Teller's Tech. Ship It Weekly
0:14
is where I filter the noise and focus on what
0:17
actually matters when you're the one running
0:19
infrastructure and owning reliability. Most weeks,
0:23
it's a quick news recap. In between those, I
0:26
drop interview episodes with folks who are actually
0:28
building in the space. Today is one of those
0:31
interviews. I'm joined by Austin Reed from Horizon
0:34
.dev. He helps small and mid -sized businesses
0:37
save serious time with AI and automation. And
0:41
this isn't theoretical. We talk about what actually
0:44
works when you're automating real workflows with
0:47
real revenue behind them. not just demos. We
0:50
get into the biggest misconceptions, why projects
0:53
fail more from communication than code, where
0:56
human in the loop makes sense, and what tools
0:59
his team is using behind the scenes. From GPT
1:02
and Claude to Cursor, CICD, and code review bots.
1:13
Today, I'm joined by Austin Reed. Austin Reed,
1:16
your automation expert with Horizon .dev. He
1:19
helps SMBs save serious time with AI and automation.
1:23
And we're going to talk about what actually works
1:25
in the real world, not just demos. Austin, nice
1:28
to meet you. Thanks for coming on. Thank you,
1:29
Brian. I appreciate it. So I'm curious, what
1:32
do you do? day -to -day in your job? Man, a little
1:36
bit of everything. I do a lot less programming
1:38
nowadays than I did when we started. When I started,
1:41
it was just me and my ex -business partner, Mattel.
1:44
We were just keyboard warriors, always programming,
1:47
me and him. And I would do sales and accounting,
1:49
whatever. But nowadays, I mostly manage the team,
1:53
lead the team. I hire people. I help the team
1:57
with project structure and architecture. I help
1:59
debug. more advanced problems. Sometimes I jump
2:02
into code, but it's not that often. But I'd say
2:05
the majority of my time is with client relations
2:07
and on sales calls. What is the real unlock or
2:12
benefit for AI and automation for small to medium
2:15
businesses right now, in your opinion? Well,
2:17
I guess it would depend on what area of the business
2:21
you're talking about unlocking, right? Every
2:23
business is unique with its own problems. Some
2:26
businesses, you know, they would really benefit
2:28
by having some sort of sales automation, sales
2:31
AI enablement, right? Helping them generate more
2:34
leads or helping them to where they don't drop
2:37
leads or helping them with statistics on that
2:39
side of things. Other businesses, they have that
2:41
unlock and they would benefit more from fulfillment
2:43
side of things, making sure everything is tight,
2:46
everything's efficient, working properly. There's
2:49
not any errors going on. AI can be used. across,
2:53
I'd say, big portions of a lot of businesses.
2:56
Most businesses, I mean, they essentially function
2:57
the same. They take in some sort of customer,
3:00
they service the customer in some way, or they
3:03
give them some sort of product, right? And then
3:05
there's the, after you turn in the project, like
3:08
transition period, right? Where you get like
3:10
referrals on the back end and stuff. And so adding
3:13
AI in any of those core areas, I think really,
3:17
really helps. 100%. Yeah, that makes sense. So
3:19
what would you say is like the biggest misconception
3:21
that companies have when they say, we want automation,
3:25
but they don't necessarily know what that means,
3:26
right? They want automation. They want to move
3:29
quicker, maybe, but they don't necessarily know
3:31
what automation is or what that actually entails.
3:34
Yeah, I mean, I think there's two. One, on one
3:37
side of things, you have people who think that
3:39
AI is going to cause more problems than it's
3:42
going to solve, or they don't trust the quality
3:44
of AI. They don't have... a firm belief in the
3:48
powers of what it can do and on the other side
3:50
of things they're definitely too gung -ho to
3:52
get things done i i find the biggest issue is
3:57
usually around some sort of content generation
4:00
so they want to automate their instagram they
4:03
want to automate their video creation things
4:04
like that and so they think that ai can just
4:07
do everything and another thing is they also
4:10
think that because there's ai and everything's
4:12
easy we have things like cursor and stuff now
4:14
that It's really easy to build these things.
4:17
And so like they think it's going to cost like
4:20
a couple hundred bucks or a thousand bucks to
4:22
get a system that's robust enough to like manage
4:25
the business. It's like, hey, you know, your
4:27
business is working like. $100 ,000, $200 ,000
4:30
worth of sales a month. And you honestly think
4:33
that like a $300 product is going to be able
4:36
to sustain that. That doesn't really make a whole
4:38
lot of sense, right? So you have a lot of these
4:40
problems. It's just underestimating and overestimating
4:44
their capabilities, depending on what side you're
4:46
on. And then definitely overestimating how cheap
4:50
it is. Yeah, no, that's fair. Okay, so I know,
4:54
like we talked about this a little bit, depending
4:55
on the company, there's a lot of different...
4:58
automation wins. But what, based on your experience
5:01
for companies, what is the most popular or common
5:05
win for automation? Okay. Yeah. So it's going
5:09
to be one of two things. Usually it's either
5:11
on the customer service side of things, or it's
5:13
going to be on the sales side of things. So specifically
5:16
for B2B businesses. it'd probably be more on
5:18
the sales side of things like i said automating
5:21
lead flow increasing the amount of leads you're
5:23
reaching out to enabling your bdrs sdrs aes to
5:27
be able to do their jobs even better that drives
5:29
a massive amount of impact in sales organizations
5:32
whereas just two years ago like sdrs bdrs they
5:36
were sending 50 100 emails a day they were handwriting
5:38
the emails right now Then they moved on to doing
5:42
sequencers. And so then all of a sudden they
5:44
were sending thousands of emails, but they were
5:45
all templated with spin tax, whatever. And now
5:49
it's all AI generated. And I know businesses
5:52
that are sending millions of emails a month,
5:54
which is absolutely insane, right? On the other
5:56
side of things, you have the customer service.
5:59
And it really kind of depends, but like you could
6:01
imagine your mom and pop dental shop, like. They're
6:04
not always there to answer the phone. It would
6:05
be nice to be able to call on like Saturday at
6:08
six when people are busy or maybe people aren't
6:10
working and be able to schedule an appointment
6:12
without having somebody there needing to take
6:16
the call, right? And so that also translates
6:18
to a chatbot or an email responder. It's essentially
6:22
the same thing. It's just the medium of which
6:25
you communicate changes. But the AI is still
6:27
doing the same thing. It's taking schedules,
6:29
doing follow -ups, taking information, you know,
6:32
filling out forms and stuff so that the customer
6:35
service side of things is on point. Is there
6:37
any part of a business right now that you specifically
6:41
try to not automate or you stay away from automating?
6:44
Yeah, I'd say one, finances. And I don't think
6:49
it's because it's bad to automate finances. I
6:52
think it's just because I'm not the best finance
6:55
person. So because I don't understand it at a
6:58
very intimate level, I tend to stay away from
7:00
it. And I don't mean like automating like your
7:02
accounts receivables and whatever. No, I mean
7:05
like actually automating like statistics and
7:08
things like that with accounting. Like I don't,
7:10
I'm not an expert at that, right? The other thing
7:13
I stay away from. Firstly, is anything medical
7:16
or HIPAA related? I don't want to touch, right?
7:18
Because yeah, same thing with government. I don't
7:21
want to touch it. But in terms of business, I'd
7:24
say mostly just finances and the things that
7:27
really require a personal touch or somebody overlooking
7:31
it. So we'll do something called men in the loop,
7:34
which is like, maybe we'll do an automation,
7:36
but then a human has to approve it. So the human
7:39
edits it, approves it, says it's good, right?
7:42
So for very bespoke services, you know, it's
7:45
always good to have that little bit of human
7:47
touch, right? So you can have some automations,
7:48
but the automations, they don't do the job. They
7:51
enable the person to do the job better. So it
7:54
sounds like you're using AI as a tool. It's not
7:56
replacing someone too. I mean, that's kind of
7:58
an important point. Like if you're an expert
8:00
in something, you can use it to help further
8:02
along. development or use as a tool, but you're
8:05
not replacing the individual. It's not because
8:08
you don't understand finance. It could potentially
8:09
hallucinate, right? So you don't want to just
8:13
recklessly use a tool without knowing what the
8:15
output could be. Right, exactly. And so I guess
8:18
the reason why I'm really good at automating
8:19
sales and customer service and different fulfillment
8:22
things is because I intimately understand a lot
8:24
of those problems because, you know, I have those
8:26
problems within my own business. I'm not saying
8:28
I don't have finance problems in my business.
8:30
I totally do. But I'm like not looking at that
8:33
as well. I'm more focused on the day -to -day
8:35
operations, right? So yeah, no, I totally agree.
8:38
Not replacing the individual is key. I mean,
8:41
even like with my assistant, like every once
8:44
in a while, I'll be like, hey, what do you spend
8:46
most of your time on every single day? And she'll
8:48
tell me, I'll be like, hey, well, let's take
8:50
this off your plate, but let's add this new thing.
8:53
You know, now that you have more time, now that
8:55
you have more leverage, let's make the... return
8:58
on time even greater for what you're working
9:01
on. Now that makes sense. Is there anything that
9:04
you've gone to automate with AI that's backfired
9:07
that you're willing to talk about? I mean, I'm
9:09
going to say not really in the sense of it backfired
9:13
because it wasn't a good automation. And the
9:16
reason why is because before we dive deep in
9:19
automation, we had already been a programming
9:22
agency for a while. So we already knew the capabilities
9:25
of what we we'd already been through the crap
9:28
right like we've already made a lot of mistakes
9:31
and learn from them and whatever but i would
9:33
say the ones that have failed most of the time
9:36
they don't fail because of the job they fail
9:39
because of communication or improper vetting
9:43
of the client or just clashing personalities
9:46
is it ever where they don't maybe give the right
9:49
spec on what they actually need or they're not
9:52
even sure of what the output is that they need
9:54
i mean that happens a lot we've gotten better
9:57
at identifying that problem so by first of all
10:02
don't automate something that hasn't happened
10:04
yet you know When we first started out with automations,
10:07
there was a lot of people that were like, oh,
10:08
we're going to have a lot of demand in February,
10:11
so we need to build this system so that when
10:12
the demand comes, we'll be ready to go. And then
10:16
February comes and there's no demand. It's like,
10:18
okay, so you just spent $10 ,000 on a system.
10:20
That doesn't help you. It doesn't make sense,
10:23
right? And so I always try to make sure that
10:26
there's actually a need there. And I think also
10:28
the way we build kind of weeds some of that out.
10:31
So we always like a lot of business owners. common
10:34
they're like oh we want this and this and this
10:36
and this and that it's like a huge menu order
10:38
of like a restaurant menu of what they want right
10:41
and that's cool but we're like okay that's cool
10:45
but how about we what's the most impactful quick
10:48
win we can do for you right and so we'll start
10:50
with that that one impactful thing and then we
10:54
give it to them as quickly as possible so we
10:56
work a lot with cicd framework i don't know if
10:59
you know what that is but basically we'll develop
11:01
a small prototype as quickly as we can as best
11:04
as we can and we'll give it to them as quickly
11:07
as they as we can in order to get feedback on
11:10
on how that is because maybe like you know we
11:12
build an automation but then it doesn't quite
11:14
work the way that they wanted or maybe the owner
11:18
has this great idea but then when the team members
11:21
get their hands on it like their reality is a
11:23
little bit different because the owner isn't
11:26
doesn't have his head in the trenches, so he
11:28
doesn't know exactly what their day -to -day
11:30
looks like. So we always try to get... MVPs and
11:33
prototypes in the business owner's hands as quickly
11:36
as possible so that they're able to play around
11:39
with it and test. And then as we stack on more
11:42
features, you know, we get feedback like, hey,
11:45
we like this. We don't like that. We'd like to
11:47
change that. And it would be cool if we could
11:48
add this. We're like, OK. And so we can start
11:50
adding those little tweaks early on instead of
11:53
having a project where it's like, OK, cool, we'll
11:56
get it done. And then two months later, we turn
11:58
something in. That's not what they wanted. Right.
12:00
What's the tech that you're generally using behind
12:02
the scenes to build these automations? You're
12:05
talking about like the code itself or like the
12:08
tools we give our developers? The models, the
12:10
code, APIs. Everything? AWS, yeah. Is it more
12:15
like AWS? Are you using like Gemini models? Are
12:18
you using Claude models? Yeah. Yeah. We use a
12:23
lot of GPT and Cloud. So I find Cloud is a lot
12:26
better when it comes to sales and anything writing
12:29
related, you know, like blog posts or customer
12:32
service. Like Cloud is just sounds more human
12:34
for some reason. I don't know why. GPT is a lot
12:37
better at analytical things. GPT Codex 5 .2 that
12:42
just came out is really good. We've been using
12:44
that with Cursor to build. apps. So that's been
12:47
helping out our development team a lot. We've
12:49
also been experimenting with having AI review
12:52
pull requests that programmers are implementing
12:55
into code bases to see if it's good or not. And
12:58
it's been working quite well. That's been quite
13:00
interesting. We've been thinking about implementing
13:03
even a bot that whenever somebody adds an issue
13:06
to GitHub, it'll maybe do a code to see what
13:09
it looks like. And then the developer just reviews
13:12
the code and edits it. We know that AWS has a
13:15
a service that does that. We haven't fully implemented
13:17
that part yet, but we're definitely experimenting
13:20
with it now. So, I mean, there's definitely a
13:23
lot of changes in the industry. And then in terms
13:26
of like servers, we don't use AWS or Google as
13:29
much unless the client already has it. We actually
13:31
prefer to use Volter, which is digital, same
13:34
thing, right? I mean, same server stack, same
13:37
technology, just a different name. And a lot
13:39
cheaper, I would say. We use a lot of Kubernetes,
13:42
a lot of Docker, Docker Swarms, Django, Node
13:46
.js, NNN, yeah. Has there been any technical
13:50
challenges that you've run into with building
13:54
out these automations? One thing is getting credentials
13:57
from the customer. I know that sounds simple.
13:59
Yeah, I've run into the same issue. Yeah, it
14:03
sounds like a simple thing. It's more of a customer
14:06
service problem than it is like a programming
14:08
problem. But I mean, a lot of these people, they
14:10
don't even know what their passwords are. So
14:12
it's like trying to get, or like you made a custom
14:16
app and you need to get into their Google console
14:18
to add whatever. And it's like, that's... That's
14:21
difficult to do with some of these people who
14:24
aren't as tech savvy, right? Anything else that
14:26
I find makes it particularly difficult? Scope
14:29
creep is totally a thing. Definitely have to
14:31
push back on that. People get overly zealous
14:33
about things. That's why we always try to, we
14:37
record all our meetings. We always try to put
14:39
project plans right. So this is what you agree
14:41
that we build, you know, so make sure everything's
14:44
very transparent. But even with that, scope creep
14:46
comes in sometimes. And then other than that.
14:50
I mean, at first we didn't, there were some clients
14:54
where we had trouble fully understanding what
14:56
their vision was, but I think that was more an
14:59
experience than it was actual, actually a problem
15:02
because now we don't see that as much because
15:04
we have a lot more things in place, like making
15:06
sure like there's already something to automate,
15:08
giving them the prototype first so that we can,
15:11
right. So we flushed out a lot of those issues.
15:13
How do you handle? the this is scary or I don't
15:17
trust AI reactions pushback that you may get.
15:20
Well, I mean, if they came in from a cold email,
15:23
I'll be like, well, who do you think wrote the
15:24
email? No, that's a hard objection. And for a
15:30
lot of people, that's pretty hard to know because
15:33
they just have some sort of mental block there.
15:35
And, you know, I'm not here to convince people.
15:38
If they don't want the automation or if it doesn't
15:41
serve them, I mean, it will serve them. if they
15:44
truly don't want it like i'm not gonna sit there
15:47
and try to convince somebody otherwise like i'd
15:49
much rather go to the guy who's like super pumped
15:51
and excited about getting some new things done
15:54
in the industry that hasn't been done before
15:56
you know like that's much more exciting but i
15:59
guess where we see some of that pushback is like
16:01
in a business where it's like an old father and
16:04
he's passing it on to the son and so the son
16:05
wants to implement these new things but the father's
16:08
like no like the old way works why would we change
16:11
it You know, we see some of that and they're
16:15
usually a quick win. Something super small and
16:18
kind of insignificant is the best way because
16:20
you can talk to them all you want. But when you
16:22
actually show them something, it really works.
16:25
So even just showing them a case study, but like
16:27
really running through and be like, hey, you
16:29
know, this. Similar business, different industry
16:32
and problem or whatever. Like, hey, look what
16:35
we did there. This is what it looks like for
16:37
them. We can do this for you. And they're like,
16:40
oh, really? Is that possible? I'm like, how about
16:41
we start small? We'll do something really small
16:44
and quick, you know, quick win. It won't cost
16:46
you very much. It won't take a lot of time. And
16:49
you can just get your feet wet and see if you
16:50
like it. Where do you think AI is going? If you
16:53
had your crystal ball, where do you think like
16:55
three, five years? Obviously, it's a changing
16:57
market. It's crazy. The last couple of years
16:59
have been crazy. Yeah. If you were to hypothesize
17:03
that, where do you think it's going? Well, AI
17:05
influencers is already a thing, even though people
17:08
don't fully believe it yet. I think that software
17:15
agencies are going to, a lot of them are going
17:18
to go out of business. because business owners
17:20
are going to be able to make software themselves
17:22
so you're going to see a big flood of apps and
17:25
new stuff coming out because there's going to
17:27
be a bunch of normal people launching their own
17:29
apps and so maybe the new game is more sass related
17:33
and less programming related i still think there
17:35
will be programmers but i feel like it'll only
17:37
be like very advanced niche stuff or old stuff
17:40
like like like cobalt right for banks or something
17:44
but so yeah i think At least my industry is going
17:48
to, like, we need to pivot hard because, like,
17:51
engineers now are not, I mean, they know code
17:54
now, but maybe in five years they'll still know
17:56
code. But in 10 years, they might not know how
17:58
to program, you know, and that's kind of weird.
18:01
Outside of that, man, I think it's very good
18:03
and very bad in a lot of ways. And it's really
18:05
hard to say. Yeah, that's fair. Yeah, just like
18:08
the car did, the horses that will probably open
18:10
up new jobs and, you know, maybe business use
18:14
cases that. we're not even sure of yet you know
18:17
but yeah it's definitely interesting is there
18:20
any hype takes about ai that that you disagree
18:24
with or you want to kill i don't think ai is
18:27
quite to the point yet where you can fully automate
18:29
social media i mean you can for like those info
18:32
talking head videos not talking head but like
18:35
the ones where it's like they say a story like
18:39
you know oh this person went missing in 1994
18:41
because whatever like yeah those can be automated
18:44
but i mean like truly valuable content i don't
18:47
think can be automated to a high enough quality
18:51
extent yet right i think that's coming i think
18:54
it's close but i still think that there needs
18:56
to be that human element the other thing is is
18:59
because we're posting so much ai content now
19:01
the ai is just rewriting ai content yeah so it's
19:04
just learning from itself at this point So as
19:08
time goes on, I believe the information gets
19:10
more and more diluted. And that's a problem in
19:14
and of itself. I think that kind of covers it,
19:17
at least my view. I know just being on LinkedIn
19:20
even, but more so Instagram, Facebook, TikTok,
19:24
it's all AI -generated Sora content now. A lot
19:28
of it, yeah. It's hard. It's hard to just have
19:32
real, genuine content that's not AI -derivative.
19:36
Yeah. To be real. It cuts through the noise pretty
19:39
well. So we talked about this a little bit earlier.
19:42
How do you think AI is going to influence development
19:44
in the future? I mean, now we're using Cursor,
19:47
right? We're using MCP servers to get information
19:51
on Kubernetes clusters or whatever. But we don't
19:55
have new junior engineers that are getting hired
19:58
at the same rate they were just a couple of years
19:59
ago. What happens to engineers and developers,
20:03
you know, in the next five, 10 years? They become
20:05
prompt warriors or they switch industries, right?
20:09
I mean, even our guys, like I handed them Cursor
20:12
like a couple months ago. And before that, it
20:15
was like Microsoft Copilot, right? Like two years
20:17
ago, something like that. And even Microsoft
20:21
Copilot. Like, dude, the difference in productivity
20:23
between my guys was insane when I gave them Copilot.
20:27
It was like, it's not like they don't know how
20:29
to write code. It's just that they can write
20:31
it so much faster. They can be like. I want this,
20:34
this, and that. And there's a huge difference.
20:37
It's like I see a lot of business owners who
20:39
don't know programming, who try to program with
20:42
Cursor, and then they jump on Cursor, and it
20:44
doesn't work. Cursor sucks, blah, blah, blah,
20:46
and it's horrible. And I'm like, you don't know
20:48
how to talk to it. It's really what it is. I
20:51
mean, there are some things like, yeah, it probably
20:53
isn't the best app, but understanding how to
20:56
architect. how to describe how it works with
21:02
data, what order to build things in. That's huge,
21:06
right? Because, you know, you get a certain database
21:10
model set up, certain APIs set up, and all of
21:12
a sudden you add something crazy in the mix and
21:16
you have to redo everything, right? And so it's
21:18
like a lot of people just jumping in like they
21:22
don't know that. So to answer your question,
21:24
man, I mean, like I said, we're implementing
21:26
AI more and more. All of our dev workflows, we're
21:30
implementing it in code review and debugging
21:32
and our CICV type of stuff to make sure that
21:36
nothing broken gets pushed. We're implementing
21:38
it and helping them even. on issues you know
21:42
they get an issue they talk to cursor cursor
21:45
spits out some code they review some of the code
21:47
they write some code themselves they tell it
21:50
to write a certain function a certain way they
21:52
sip their coffee they wait they test it you know
21:55
and for testing too it's huge for like unit testing
21:58
and integration testing like you write out a
22:01
whole api in django and then you're like hey
22:03
unit test this you know integration test this
22:06
like that's huge the speed of which it can write
22:11
out some of that code that, you know, before
22:13
it would take, you know, weeks is insane. Yeah,
22:16
I agree. I know for me, being able to vet the
22:19
output that Cursor gives you and not just trusting
22:21
it, like you mentioned business leaders that
22:24
may go use Cursor for the first time, but they
22:25
don't really understand what they're, even understanding
22:28
what to ask, but then understanding what the
22:30
output is that they're given. Like you need to
22:32
have some awareness around that, at least currently,
22:34
maybe in the future you won't to the same degree,
22:36
but right now you, you know. it still can hallucinate.
22:39
It could still mess up. So you need to know how
22:41
to frame the question in the right way so it
22:43
gives the output that you want, but then also
22:45
vet the output that it's giving you. And if you
22:47
don't know anything about the subject that you're
22:48
asking it to help with, you don't know if it's
22:51
a good output or not necessarily. Right, exactly.
22:54
Any closing thoughts? Not really. I think the
22:58
world's going to change a lot really quickly.
23:00
And we're along for the ride. Everybody needs
23:03
to pivot. Every business, every industry needs
23:05
to pivot, not just ours. Right. So the faster
23:09
you can kind of get on board and start testing
23:11
new stuff, I think the better off you're probably
23:14
going to be. Yeah. Be ready for it because it's
23:16
coming regardless. Like you'd even mentioned,
23:18
you know, the father and the son, the father
23:21
may not want AI. He probably is talking to an
23:25
AI chatbot on a website every day. He's probably
23:28
getting phone calls from AI chatbots. He just
23:30
doesn't know it. He's already interacting. He's
23:32
already in that world. He just doesn't know.
23:33
Yeah. Yeah. All right. Well, appreciate it. Thank
23:36
you, Austin, for coming on. Really appreciate
23:37
it. I appreciate it, Brian. Definitely. All right.
23:46
That's the conversation with Austin Reed from
23:48
Horizon .dev. My biggest takeaway is his framing
23:52
that automation isn't magic. The win is picking
23:55
a real pain point that already exists and getting
23:59
a quick win live and then iterating. Also, the
24:02
AI is cheap now mindset is a trap. If your business
24:06
is doing real volume, you need systems that can
24:09
actually hold up with guardrails, approvals where
24:13
it matters, and someone who can sanity check
24:16
the output. If you want to connect with Austin
24:18
or check out what they're building, I've got
24:21
links in the show notes to his LinkedIn and to
24:23
horizon .dev. If this episode was useful, share
24:27
it with a DevOps, ops, or business friend who's
24:29
trying to automate workflows without creating
24:32
a bigger mess. Hit follow wherever you listen
24:35
so you don't miss the weekly news recaps plus
24:37
these guest interviews. We'll be back with a
24:40
regular Ship It Weekly News episode later this
24:43
week. See you then. Thanks.
