Latest / The Retail Podcast / Why Retailers Are Already Falling Behind on AI

Transcript

Why Retailers Are Already Falling Behind on AI · The Retail Podcast

0:00
That the status quo is no longer acceptable.
0:02
AI is. This is the big one.
0:05
I want folks to really understand that this is a
0:07
categorically new way of work and folks need to start working
0:12
on it right now. Hello and welcome back to the
0:25
Retail podcast. Now this conversation is due to
0:29
a conversation with Jason, who's the current president of the Mac
0:32
Alliance about influential people in this world of AI.
0:37
Now Kelly, who you're going to meet in one second, do you can
0:40
see is in my doing my research. He's probably one of the most
0:43
influential voices in modern commerce architecture and across
0:47
his roles at Oracle, ATG, Commerce Tools, and now the next
0:52
next generation of AI driven innovation, he's probably shaped
0:57
how global AI leaders think about this new world.
1:01
Kelly, have I missed anything? I mean, when you lay it out like
1:05
that, I mean, what are you supposed to say?
1:08
Yeah. Well, listen, thank you so much
1:10
for coming onto the podcast. I really do think it's a
1:13
pleasure and a delight to take a deep dive.
1:15
Before we get started, what do you, when you do these types of
1:20
podcasts and only you're going to say, well, what does your
1:22
community audience want? Well, what do you traditionally
1:25
want to get across? What would you like the listener
1:27
to leave this show understanding.
1:29
That the status quo is no longer acceptable.
1:33
I think so many folks think that they can just put a little
1:35
Clippy type assistant on their website and call it a day.
1:39
And AI is this is the big one. This is the big one.
1:42
Now it's going to take years and years, but there's already a
1:45
very stark divide in the labour industry between people who are
1:49
AI native and can use AI to its fullest extent versus those who
1:53
don't. And I want folks to really
1:55
understand that this is not just a smarter Google.
1:57
This is not just a way to auto complete some code.
2:00
This is a categorically new way of work and folks need to start
2:05
working on it right now. What do you think, because
2:07
obviously through through your extensive experience, you've
2:11
seen the monolith era, you've seen the headless era, and now
2:16
it's like the composable Mac era.
2:18
Are these a natural progression from each one?
2:21
So if I have been going up, I'm a retailer and I've gone through
2:24
this process, am I, am I going to have to reinvent my stuff?
2:28
Am I going to have to like start from scratch?
2:30
How much of A connection is but is there between these
2:33
transitions and where the world is going now from a technology
2:36
perspective? I mean, you have to go back and
2:38
and look at computer science as as a new industry, as a new
2:42
field of study. It's relatively brand new.
2:45
I mean, it's really only been since the 60s maybe and really
2:48
started to take off in the 90s and 2000s.
2:51
If you look at computer science versus any other established
2:54
field, whether it's archaeology or medicine or law or take your
2:59
pick, they've had decades and decades and decades to come up
3:03
with best practises, right? If, if, if an engineer needs to
3:06
go build a bridge, they know how to do that.
3:08
That's a well established task at this point.
3:11
And there's not a lot of innovation around the edge.
3:13
And software is unique in that the entire foundation is all
3:16
bits and bytes. You can RIP it out and rebuild
3:19
it. And over time, the technology
3:21
has improved dramatically, right?
3:22
Do you remember Bill Gates famously saying 64 kilobytes
3:26
ought to be enough for anyone? And now in your phone, right, We
3:29
practically have a TB of, you know, of storage.
3:33
So it's matured. And historically, back in the
3:36
90s, it made a lot of sense to have monolithic applications
3:40
because the complexity was on the inside and networking was
3:44
really slow and really painful. So there is no means of like
3:48
exposing little bits of functionality up to the world
3:51
because you just couldn't call a function remotely.
3:54
It didn't work. So that incentivized everyone to
3:57
build big applications that did a little bit of everything.
4:00
And then we saw the rise of headless.
4:02
My boss at Commerce Tools, founder, CEO Derek Horig,
4:06
invented the headless and the mid 2000s changed the industry.
4:10
And headless was pretty revolutionary in that the web
4:14
head was decoupled from the back end.
4:16
And when I first started building in a large scale
4:20
e-commerce websites, it was pretty crazy because you
4:22
literally embed business logic right in the UI, right?
4:25
Right. In your JSP file you would have
4:27
business logic. And now the thought of that, it
4:30
just, you know, makes your skin crawl, right?
4:32
Because you would just never do that anymore.
4:34
So headless was a big innovation.
4:36
And then on top of that, microservices and microservices
4:40
is another iteration of headless.
4:42
It's further iteration. With headless, you can have a
4:44
monolithic back end or microservices back end.
4:46
Microservices are the better way to do headless because now you
4:49
have little applications that each own their own domain, their
4:52
own data, their own everything. So you can just update your
4:55
shopping cart API with its own application or just your
4:58
promotions API or your pricing. You can go in and make granular
5:02
changes to business logic and functionality without
5:05
redeploying the whole application.
5:06
And now with AI, it's even better.
5:09
It's, it's almost like like the headless we always wanted
5:13
because now you just expose up your data and functionality via
5:16
model context protocol, which then calls the APIs.
5:20
And now AI is the giant head. It's the orchestrator, it's the
5:23
everything. And it just magically knows how
5:25
to call that API's. So feels like we're, we're
5:29
evolving towards something a little bit more stable.
5:32
Although who knows, maybe 20 years from now the kids will be
5:34
talking about how antiquated Mach was back in the day.
5:37
But still, this feels pretty cutting edge.
5:39
We've got the tech to support it.
5:40
You know, you can make remote calls on a heartbeat.
5:42
Everyone's exposing MCP now, or everybody should be.
5:46
And with AI, you know, MCP pairs really, really nicely with it.
5:50
What I'm hoping this this isn't being too topical, but I'd love
5:54
your view, just your personal view on MCP, sorry.
5:58
In terms of UC, the universal commerce protocol that Google
6:01
has released, I'm just curious what your thoughts are on that.
6:05
Do you think it's something because like Amazon was out, but
6:07
now it's in and a few of the major platforms are going down
6:11
that route? I'm just curious on your
6:13
thoughts on it. Well, if we want to go back to
6:16
what is it, November of last year, Stripe teamed up with Open
6:21
AI and came up with a Gentech commerce protocol and they did
6:25
so with, as far as I can tell no participation from industry
6:28
whatsoever. Their attitude is like, hey,
6:31
we're open AI, we're just going to do it.
6:33
And clearly their lack of, of inexperience in the industry
6:38
caused ACP to be half baked and not worked terribly well.
6:43
And just their attitude towards industry has been not very nice.
6:46
They have not been very great to deal with as as a new
6:49
monopolist. So, you know, given that vacuum,
6:52
general dissatisfaction with ACP, Google then came in and at
6:56
NRF this year in January released UCP.
7:00
And just as an example of the difference between these two
7:03
organisations, it was rumoured, I didn't even see it, but it was
7:06
rumoured that Open AI had two attendees at NRF, which is the
7:10
world's largest retail conference, 10's and 10s of
7:13
thousands of people, right? Everyone there, everyone there,
7:16
and they send two random people from Open AI apparently.
7:20
And then you compare that to Google and you've got Sundar,
7:22
the CEO on stage at doing a massive keynote to kick off the
7:27
launch of UCP. And there must have been 200
7:30
people from Google there. That's the difference between
7:33
these two organisations. Google cares.
7:35
Open AI clearly doesn't care and you can talk to people.
7:38
Google like what a novel concept that you can actually find time
7:41
and talk to people and talk about the protocol.
7:44
Google's been trying commerce many different ways over the
7:46
decades, some successful, some not, but they've been learning
7:50
and they have real, credible, serious people there who
7:52
understand commerce. Now there's still some gaps for
7:55
sure, but UCP is a much, much, much better protocol than ACP
8:00
and it shows that they consulted with industry and it shows that
8:03
they at least halfway know what they're doing, which is great to
8:05
see. But UCP also is much more widely
8:10
supported now. They've been much more open.
8:12
They donated it to a foundation which has been great ties in
8:16
with the merchant centre. So if you're already a Google
8:18
customer and you upload your product catalogue for
8:21
advertising purposes, you can now auto sell.
8:25
It's also an extendable protocol, so Talon One, which
8:28
was a promotions vendor, just recently sold to Adyen.
8:32
They built a loyalty extension to UCP and I would expect to see
8:36
many more extensions over time, which is great.
8:39
So I give credit to Google here for really knocking it out of
8:43
the park with UCP and I'm cheering them on to win this
8:47
battle. Now one of the areas that I know
8:49
we've obviously spoken before was around all the management
8:52
and how you're potentially looking at that whole back end
8:57
process system technologies to create a unified way.
9:02
Can you just talk me through what, what, what are you
9:04
actually practically looking at and, and what are you putting
9:07
together or what has been put together?
9:09
Well, ACP and UCP are selling channel standards only.
9:12
So they are just order capture and they sit between the
9:17
commerce platform, whatever that happens to be.
9:19
You know, often something like Shopify or commerce tools, and
9:22
then it sits between the payments vendor, you know, so
9:25
think Stripe or Adian or somebody like that.
9:27
And then finally, it's it's with the AI head.
9:30
So think of your ChatGPT or your Google Gemini, something like
9:34
that. And there's kind of a triangle
9:36
there between those three. And the ending goal of that of
9:39
those two protocols is to get you to a thank you for your
9:42
order page, right? That the goal of those is now
9:46
there's a lot more work, obviously behind the scenes.
9:48
And this is something I fault both for not really taking
9:51
seriously is you have to surface up inventory data and shipping
9:57
data natively and quickly so that you are able to guarantee
10:01
that the item is there sitting in a warehouse for fulfilment.
10:04
And Alex, if you buy something, it says it's going to show up in
10:08
your doorstep on Wednesday, right?
10:10
Great. That's what you need to sell
10:11
something. Otherwise you just have a
10:12
product catalogue, which doesn't really help anyone.
10:15
So that's one thing. Those are two things.
10:17
The other is once you capture an order, you have to do something
10:20
with the order, right? The order has to go somewhere
10:23
and that can go right back down to fulfilment via an OMS. It can
10:28
go through the commerce platform.
10:29
So there's definitely a lot of areas of of improvement there.
10:33
We have Pipe 17 where I'm President.
10:35
We surface up inventory in many cases directly from the Three
10:38
PL. We do so authoritatively.
10:40
We do the same from shipping. And then we built a network so
10:43
that you can pass orders to this network directly from ACP or UCP
10:47
or via the commerce platform and then go all the way back to the
10:51
Three PL or WMS for fulfilment. So we get it.
10:54
We get the order from the thank you for your order page all the
10:57
way through somebody's doorstep. Basically it's the back half of
11:00
ACP and UCP. And where do you think so
11:02
obviously, these are early days and retailers are sort of
11:05
grappling around with what to do.
11:07
What do I invest in now that will be fruitful in five years
11:12
from now? What's your view like?
11:13
What's the direction of travel that you're seeing that the
11:17
industry is going when it comes to to to this new world?
11:20
I'm seeing a lot of inertia, and the nurses do.
11:23
Nothing. Wait to see who's going to be
11:26
the winner or leader. Which is just waiting and
11:30
waiting and kicking the can down the road and hoping that the
11:32
next person in your seat has to deal with it and not you.
11:35
Because look, you, you have to make some pretty substantial
11:38
changes here, right? Like a lot of folks have legacy
11:41
OMS going back 30 years. That doesn't work, clearly
11:44
doesn't work very well. They've got home grown commerce
11:47
platforms or some other mess, you know, that doesn't work very
11:50
well. So if you're a VPAV commerce at
11:53
a brand, what your goal is, is to do as much work around the
11:56
edges as possible without having to make any difficult changes.
12:00
So that's what leads to this Clippy phenomenon of, oh, we've
12:03
got an assistant now and then. You know what's even worse is
12:07
people do AB test visits with and without the assistant and
12:12
they find, of course, that the assistant dramatically over
12:14
converts relative to your normal control group.
12:17
And that's just because of selection bias.
12:19
Because if somebody actually goes to your.com and is
12:21
incentive, it's it's is interested enough in buying
12:25
something from you that they then go to your assistant and
12:28
engage with an assistant. That's a much higher conversion
12:30
intent just off the bat. So then that gets reported back
12:33
to the board as look, our AI work is actually paying off.
12:35
Look at this, we got so much higher conversion rate.
12:37
Isn't that awesome? And they're not.
12:39
They're just not doing anything. So brands need to be doing a
12:42
couple of things right now. They need to be enriching and
12:46
cleaning up their product catalogue data.
12:48
That's a really, really big one. Product catalogue data is a
12:51
mess. Always has been, always will be.
12:53
I sit on the board of Bluestone PIM and that's their entire
12:57
pitch. Is your product catalogue data
12:59
is a mess. Let us help you fix it.
13:00
And I see it all every single day of my life.
13:03
It's an absolute mess. So first people need to actually
13:06
get it together and with AI now it's actually pretty easy to do.
13:09
All of these Pims have pretty advanced AI capabilities where
13:12
you can just have it go through and clean up things.
13:14
And you also need, you know, and and like you need to add yes, no
13:20
flags that AI can understand your clothing and items better.
13:25
So your shirt, for example, right, You know how many
13:28
buttons, right, that should be a distinct field, right?
13:31
And then button type, you know, is, is bone equals true, is bone
13:35
equals false, You're right. Is plastic true, right colour
13:40
stitching, you know, thread colour.
13:42
I mean, there are 1000 attributes on a shirt and AI
13:46
thrives on that because people want very specific items, right?
13:49
Maybe they'll upload a photo of an existing shirt that they
13:52
like. And your AI assistant will then
13:54
distil that down to its component parts and look for,
13:58
you know, 50 different search criteria when it's searching the
14:01
wide world of shirts and it come back with a shirt exactly like
14:04
yours, but a different and brand, right?
14:06
Like that's what people want these days.
14:07
They don't just want another button up shirt with five
14:10
attributes doesn't work very well.
14:12
So that's one thing. Second thing is they have to
14:14
publish their data or syndicate their data out to the AI heads.
14:17
And you can go directly to ACP or UCP.
14:21
And that's a big important step. You also have to focus on AEO or
14:26
Geo. So making sure that you are
14:28
publishing a lot of original content out there on your own
14:32
domains, but also on third parties.
14:34
It's ratings and reviews, it's Reddit, it's X, it's all over
14:38
the place so that the LLMS can suck all that data in and return
14:43
better products when necessary. You also have to invest in up to
14:47
date order management, right? So selfishly you, you need
14:50
something much more modern to deal with AI because is I think
14:54
we're going to see not just two heads with open AI and Gemini
14:58
and maybe Claude. I think AI is going to give rise
15:01
to 1000 different heads where you're going to have, you know,
15:04
video game developers and you're going to have some 16 year old
15:07
kid who builds a hit video game and he's going to try to
15:10
monetize that game through in app commerce.
15:12
And you're going to see that rise just like we saw Minecraft.
15:15
It's already, yeah, I mean, for me in Roblox, for any parent,
15:19
they know that Roblox is the conduit into so much, such a
15:23
large ecosystem that is trying to transact, that is trying to
15:27
bring you on a consumer journey rather than a video game
15:31
journey. And it's in.
15:32
It's in. So look, there's a lot that has
15:35
to happen between now and then, but right now, at a very
15:38
minimum, cleaning data and publishing it up, those are the
15:41
top 2 just to get ready. The other thing is it's, it's a
15:45
it's a people issue and a lot of folks out there are just not
15:49
using AI in their day-to-day life.
15:50
Maybe they're using it as a better Google Assistant.
15:53
But there are a lot of ways, and I write about them on LinkedIn
15:56
all the time that you can do categorically different, better
16:00
work, right? It's an entirely different work
16:02
style with AI versus not. And so many of these retailers
16:06
are, are just doing, you know, like a lot of retailers even
16:10
block access to ChatGPT and Claude, like they just block
16:14
really. Like, are you supposed to run a
16:15
business and block AI these days?
16:17
But they do. It's because they're overzealous
16:19
lawyers. And what about the the, you
16:22
know, security and and, and and right.
16:24
You know, and, and I think that's just so out of date now.
16:28
And I think the AI native companies are gonna run circles
16:31
around those who are not. Yeah.
16:33
I've got, I've got 2 questions and I'll let you decide in which
16:35
order you answer them because one of them sort of relates back
16:39
to a couple of things you were saying down in a couple of
16:42
minutes ago. And the other is just sort of
16:44
thinking about this in the context of the business.
16:47
So you spend your time podcasting on LinkedIn, you
16:52
communicate a lot to the industry and you have your own
16:55
communities, communities that you are actively participating
17:00
in. What's the one thing that you
17:01
find people mostly disagree with you if they do in that?
17:05
Perspective, it's always the timing of of these big shifts in
17:10
the market. So I've been on.
17:12
That for me, you think it's going to happen sooner or later
17:15
or what's the what's the thinking there?
17:17
Social media rewards certainty. I do not have certainty.
17:21
Nobody does. Anybody claiming to have
17:23
certainty is lying. Yeah, there are some people who
17:26
say you have to do this right this very moment.
17:29
You know, There are others who say this is just a fad.
17:32
It's going to pass, you know, give it 10 years before we start
17:34
to see real improvement and it's still not going to fundamentally
17:37
change the market. And I could argue this entirely
17:40
both different ways, right? And I do like, I argue my, you
17:44
know, like I'll give you an example.
17:45
Last week I had to make some 1800 calls for insurance for my
17:49
cellphone provider airline. And I go through all the
17:53
authentication process with all three of these providers.
17:56
So you have to enter in your ID number and date of birth and PIN
18:00
codes and everything else. And then in all three cases, you
18:04
get to a human agent and they ask you what your name is.
18:09
Security purposes, you said, I've said that three times, but.
18:12
Like what are we doing here guys?
18:14
Right. And these are big household name
18:15
companies. So, you know, I think that and
18:18
this technology to be able to hand off an authenticated user
18:21
has existed for decades and I see big companies completely
18:25
unable to actually implement any of this stuff whatsoever.
18:28
On the other side, you know, I look at my day job and I look
18:32
around startup world where I live and we're all in this very
18:35
deep. You know, friend of mine just
18:37
raised 25,000,000 for his, his company Series A.
18:41
And he said they're at six people and they just, they raise
18:45
the money. And I said, well, you know, what
18:46
are you going to do hiring wise? And he said we might add a 7th
18:49
or 8th person, right? But like, that's an example of
18:52
where, you know, startup world is very different.
18:55
And in this world, we are killing it with AI and it's
18:59
dramatically improved our efficiency in our lives.
19:01
But then it's going to take longer.
19:03
And it depends on where you are in the organisation and you
19:07
know, what size of company you're at.
19:09
Because if you are working at 100,000 person, you know,
19:12
Fortune 10 company, you probably, you know, don't really
19:15
know what's happening with AI and you're probably going to
19:17
disagree with a lot of my posting because you think I'm
19:19
nuts. If you're in the startup worlds,
19:21
you're saying, well, yeah, of course, we've been doing this
19:23
now for two months. Where have you been?
19:26
Right. So it's it's a matter of where
19:29
you are. And I get disagreements on both
19:31
sides of that of you're late or you're way too early.
19:33
And I do get that a lot. Interesting.
19:36
I mean, put personal, personal and today I'm X Verizon, X
19:40
Microsoft and I can see in both of those roles that I've had in
19:43
those companies over the last 1315 years, given the technology
19:47
just today, without going forward, I would probably have 5
19:50
or 6, six agents working with me in my role to do what I'm doing
19:55
right in my, my mental model for this is if even if you're an
19:59
enterprise now, you would benefit from 5 or 6 agents
20:02
helping you. I've got about 10 doing this
20:05
little business or agency and that's helping me do all the
20:09
bits. But when I think about my
20:10
corporate career, I definitely think that, you know, the world
20:13
has moved on. One thing that you mentioned,
20:16
the second question that I want to ask is risk mitigation.
20:19
I'm, you know, C level C. EIO, whatever CXO, I have to
20:23
protect myself, right? I'm petrified of getting fired
20:26
and I have to mitigate risk. And then everything you said
20:29
about UCP about, you know, people just sort of not doing
20:33
anything and waiting for the next person to do so it's
20:36
probably because they don't want to take the risk.
20:39
What are some of the ways I can mitigate risk from your
20:41
perspective in going into this new new world?
20:45
The whole the whole framing is wrong that you just gave which
20:49
is sorry, I didn't mean to no. No, no, I love it.
20:54
I love it. And historically, you would have
20:56
had to have spent a lot of money to go all in on a new trend.
21:00
Historically you would have had to and you'd go hire Accenture
21:03
or Deloitte or, you know, pick your pick your big GSI and it
21:06
would cost millions of dollars and it would be a six month
21:10
project, 12 month project, you know, big presentation, you
21:14
know, perfectly pressed pants, the whole thing, right?
21:16
That's historically how it was done.
21:18
And in that framing, it is risky if you go forward with something
21:22
and it doesn't workout because you've invested a lot and then
21:26
you lose a lot of money. Now you can get a random
21:28
developer sitting in Dayton, OH. Sorry, I don't mean to pick on
21:32
Ohio, but you know, take your random corporate developer and
21:35
say go build me a UCP extension. Go do it and you can go pull up
21:40
plot code for 20 bucks a month and build it.
21:43
And your investment of time is basically 0.
21:46
But people don't think that way. People think in terms of big
21:49
projects that have to be managed and funded.
21:52
And you need McKinsey and to tell you that it was a good idea
21:54
and, and, and, and what we do in startup world.
21:57
And the way to do it natively with AI is before you even have
22:01
a discussion, just shut up and go build it.
22:03
Just build it. Because again, like I'm building
22:05
like I built myself a, an Onyx to UCP bridge on my own and it
22:11
works. I shouldn't be doing that.
22:12
But it was so easy to do. And AI makes this all so and
22:16
simple. Yes, there are complications for
22:19
sure, but you know, your investment is an order of
22:22
magnitude lower. It's one person's time for an
22:26
afternoon to go prototype and build something.
22:28
That's what you need to be doing on a regular basis.
22:31
And people just don't think that way.
22:32
It's an old corporate mindset of we got to do this the right way.
22:36
Big projects, big teams, the right way to do it is let's
22:39
spend an afternoon on it, prototype it, let's see it.
22:42
And then if that works, like, yeah, maybe then we spend a
22:44
little bit more time productizing it.
22:46
But until then, it doesn't really make a whole lot of sense
22:48
to do anything. Yeah, I got you.
22:50
Agile plus work in small pods go fast.
22:53
It's a right. It's the it's the hyper speed of
22:55
the old way of iterate a technology innovation and how
22:59
you sort of iterate your way. OK, makes sense to me.
23:01
What's the one question I should have asked you that I didn't?
23:04
That's a really good. Question, I guess what people
23:08
individually, Yeah, I guess what individuals can do now to
23:12
upskill themselves with AI. OK, level.
23:14
Level how? Do you do it?
23:16
What I've been doing is I've got a cheap Laserjet printer and I
23:21
print off documentation for Claude and for others, and I
23:25
will just take that and sit by the lake and read.
23:28
That's what you have to do. And then I'll go play.
23:30
And I spend a lot of Saturdays just learning and doing.
23:34
But you have to do it right. That's the key is you have to
23:37
actually do it. And what's great about AI is if
23:39
you get stuck, you ask your Claude, hey, what am I doing
23:42
wrong here? It'll tell you, right.
23:45
I mean, how cool is that? But people just don't aren't
23:48
like really opposed to learning new things, especially people
23:52
who are further along in their careers and who are used to
23:54
doing things a very specific way get very stuck.
23:57
And they just are very dismissive of this type of, of
24:01
iteration and learning. But you just, you got to put in
24:03
the time and work and do it and play with it.
24:06
Nobody's going to teach you how to do it.
24:08
Yeah, I. It's interesting.
24:10
Because a lifetime and. A lot of what you've said
24:12
resonates to me, but because I've been out of corporate for
24:15
such a long time, I've sold. I may be buying into what CEO S
24:19
or execs say on conference stages when they all understand
24:23
sort of like the Amazon plague book of failed forward
24:26
experiment fast. But listening to you, it feels
24:29
like those days that I've got I've then are still, you know,
24:33
are still living. They haven't really disappeared
24:35
and there are a lot of companies who are still scared of change
24:38
or taking a risk in this world, which which is which is the
24:42
shame. But again, maybe that's what
24:44
separates the people Who Will Win and the people who will lose
24:47
in the AI generation. OK, listen, Kelly, believe it or
24:51
not, we've always been talking for 30 minutes, which is double
24:54
my normal show length. So I don't know what that's
24:57
going to. Do to the algorithm, but it was
24:58
a. Conversation that I love and I
25:00
can continue having. So thank you so much and maybe
25:03
I'll ask you now. You'll be open to coming back in
25:05
three to six months and we can have this conversation again,
25:09
seeing where we are. I absolutely love it.
25:11
Sounds great. Thanks for having me.
25:13
Thanks, Kelly.