July 23, 2024 • 39 mins
The retail landscape is changing faster than ever. Customers demand speed, convenience, and above all, a seamless experience. This puts immense pressure on supply chains, especially with the rise of omnichannel retail. As this complex world is navigated, one thing is clear: AI is not just a buzzword; it's the key to unlocking a new era of quality and efficiency.
Captur, a company at the intersection of quality and supply chain, leverages AI to help retailers scale their operations while maintaining, and even exceeding, customer expectations. Their platform automates quality control processes by transforming internal policies into real-time, image-based verification systems. Imagine a world where every delivery is checked for accuracy, damage, and proper placement, all through AI-powered image recognition. This is the power of Captur.
Advances in mobile technology and hardware now allow for sophisticated AI processing directly on devices, which is game-changing for real-time applications, especially in environments with connectivity limitations. Imagine drones equipped with AI for optimized delivery routing, or on-site quality checks conducted instantly with a smartphone. Additionally, consider a world where merchandisers can create photorealistic 3D product visualizations simply by typing in descriptions. This is quickly becoming a reality thanks to advancements in text-to-3D generation, which has the potential to revolutionize product visualization, planning, and ultimately, the customer experience.
Traditional quality management tools, while effective, can be time-consuming and require specialized expertise. Large language models (LLMs) have the potential to automate and democratize these tools. Imagine a fishbone diagram automatically generated by an LLM, pinpointing the root cause of a supply chain error in seconds. This is just one example of how LLMs can empower teams to identify and address quality issues proactively.
However, the journey to AI adoption is not without its challenges. Companies often grapple with questions of ownership, measurement, and risk mitigation. For leaders considering AI solutions, it is advisable to embrace prototyping. Experimenting with readily available AI tools to understand their capabilities and limitations can provide valuable insights for defining specific needs and identifying potential use cases. Evaluating the strategic importance of the problem being solved is crucial. If it's a core differentiator for the business, building a custom solution might be the right approach. However, for maintaining parity with market trends, partnering with a specialized AI provider can offer a faster and more cost-effective solution. As AI becomes increasingly integrated into daily life, customer expectations will continue to evolve. Retailers need to stay ahead of the curve by anticipating how AI can enhance the shopping experience, from personalized recommendations to intuitive search functionalities.
The AI revolution is here, and it's transforming the retail value chain as it is known. By embracing this technology, unprecedented levels of quality, efficiency, and customer satisfaction can be unlocked. The future of retail is about creating seamless, intelligent, and customer-centric experiences, and AI is the key to making that vision a reality.
Mark as Played
Transcript
Episode Transcript
Available transcripts are automatically generated. Complete accuracy is not guaranteed.
Dr. Matt Waller (00:10):
Welcome to the Matt Waller podcast, where we
look at success at theintersection of technology,
logistics, supply chain, retailand CPG, also known as the
retail value chain.
I want to clarify that thispodcast is distinct from my
responsibilities as a professorin the Sam M Walton College of
Business.
Nonetheless, it aligns with myaspiration to provide practical
(00:31):
insights to professionals andbusiness by showcasing companies
and people that can enhanceyour ability to manage, lead and
strategize and marketeffectively in the retail value
chain.
Before we dive into today'sexciting episode, I'd like to
thank our sponsor, new RoadCapital Partners.
New Road invests in proventechnologies, services and
(00:53):
products that serve unmet needsin the marketplace.
They look for companies insupply chain and logistics, as
well as consumer-orientedcompanies.
For more information, go tonewroadcpcom.
I would also like to disclosethat I am a strategic advisor to
New Road.
I'd also like to recognizepodcastvideoscom for the
(01:18):
services they provide for thesepodcasts.
I'm very pleased with theirservices and now, without
further ado, let's get into theexciting episode.
I have with me today CharlotteBax, who is the founder and CEO
of Capture, which is a AI-drivenquality control solution for
(01:39):
supply chains, and I thinkyou're going to really like this
discussion If you're in supplychain management merchandising,
retailing, cpg supply chain.
If you're an investor, I thinkyou'll be interested in this.
(02:00):
This company that Charlottefounded was started in a Google
accelerator four years ago andtoday Capture already serves
Fortune 100 company supplychains across Europe, us and
APAC.
Charlotte is an expert inapplying AI to help supply
chains deliver better customerexperiences, as you will hear,
(02:23):
but unlike typical AI founders,her technical knowledge is
largely self-taught.
She started her career inproduct operations at Gap Inc.
She holds an MBA from theLondon Business School and has
been building AI products forthe past eight years.
Personally, she is passionateabout making AI accessible and
(02:48):
partnered on programs with NASAand Google to promote women in
STEM.
This particular episode today.
Again, it's looking at thequality challenge of modern
supply chains and how AI canaddress them.
But we dive into the criticaltopic of quality for modern
(03:09):
supply chains, exploringchallenges that retailers face
in maintaining service standardsacross multiple distribution
channels, from online deliveryto scheduled and in-store pickup
.
We also examine themultifaceted problem of trust
that emerges in this landscape,including customer retailer
(03:30):
trust, fraud prevention andinternal workforce trust.
Charlotte, thank you so muchfor joining me today.
Charlotte Bax (03:38):
Thanks so much for having me.
Matt, Really excited to chattoday.
Dr. Matt Waller (03:42):
Charlotte, I'm really excited about this
podcast, Really excited to chattoday.
Charlotte, I'm really excitedabout this podcast and I'd like
to ask you what are some of thebig challenges that the supply
chain faces, especially in termsof quality.
Charlotte Bax (03:59):
Yeah, I think it's such an interesting topic
and quality in the supply chainis not new, but I think it's one
that we've all probably hadexperience, acute experience
with, as customers.
Right, think about you knowwhen you order something online,
right, or you schedule adelivery, you expect it to be,
you know, fulfilled on time atyour house in the right state.
(04:22):
And if you've ever experiencedas a customer something going
wrong in that process gettingyour order and it's the wrong
thing, or it's damaged or itcomes late you know from a
customer's point of view howimportant quality is in the
supply chain, because customersare really demanding.
To say no, I might be patientonce, but I'm not going to be
(04:43):
patient twice if kind of issueskeep happening.
But for the supply chains, Ithink what's interesting is why
it's becoming harder and harderto manage.
Quality is especially in thelast five years with COVID,
we've seen lots of retailersfocusing on trying to meet the
customer where they are right,so offering new modes of
(05:08):
distribution, right.
So shop online, pick up instore, schedule delivery,
on-demand delivery, and that'screating all of these new
challenges, which I think aboutin kind of three buckets.
You know, one is theoperational challenge of just
how do you distribute to thesenew channels and meet the
customer where they are.
(05:29):
Second is experience challengeson how do you drive retention,
how do you make sure that thingsgo right every time.
And then the last one is reallythe scale challenge on how do
you do this profitably?
How do you build an automation?
Last one is really the scalechallenge on how do you do this
profitably, how do you build anautomation.
So there are those threechallenges operational,
(05:49):
experience and scale which everysupply chain face today, but
they face acutely because ofthese new types of modes.
I think of distribution.
Dr. Matt Waller (05:56):
Yeah, you know mentioning Omnichannel,
especially for brick and mortartypes of stores.
It seems like, well, one sortof tenant of quality management
is that when you introducecomplexity, quality problems
(06:19):
increase.
And so before you had aretailer that all of a sudden
you know eventually was sellingonline, maybe on their dot com.
But Omni Channel reallycomplicates that because all of
a sudden you're doing you'reselling online, but you're also
(06:39):
there's buy online, pick up atstore, buy online, deliver to
home, buy online, deliver inhome, drone delivery.
There's all these differentoptions and much of it for
traditional brick and mortarretailers is being done through
brick and mortar stores, whichis making store operations more
(07:03):
complicated as well.
Stores, which is making storeoperations more complicated as
well.
And I know, even at themerchandising level, most of
these brick-and-mortar retailersthat have gone omnichannel,
their merchants focused more on.
You know they'd have merchantsfocused on the brick-and-mortar
and then merchants focused one-commerce and over the years
(07:26):
they've started bringing themtogether.
But again, it's causing lots ofchallenges.
So you mentioned a lot of thesechallenges.
So my question is how can AIhelp to solve those?
Charlotte Bax (07:40):
Yeah, absolutely, and you're spot on.
I mean, especially when youtalk about omni-channel types of
channels.
Yeah, absolutely, and you'respot on.
I mean especially when you talkabout Omnichannel.
I think Omnichannel that hasreally been the focus for the
past 10 years, and even in mytime at Gap Inc, we were going
through that transformation oftaking the store merchandise
team and the online team andbringing them together and
thinking about unifiedassortment, unified distribution
.
I should just have one pool ofinventory, right.
(08:02):
But now everything's changingagain and changing on these
retailers and don't just havethe store and online.
We've got all these micro hubsand micro sites.
How do we adopt a hub and spokemodel to serve that?
I think AI and the other hand,too, you've got also rising
(08:25):
theft and kind of fraud in themix, right, and so where AI can
really help, I think, if wethink about operational
experience and scale,operations-wise, you're seeing
pretty early adoption of a tonof different technologies, like
how do we distribute to remoteplaces with drones, right.
(08:48):
So AI can really help in kindof distribution.
You're seeing people use AI forautomated picking and sorting.
You're seeing robotics at theactual distribution centers
themselves, right.
So I think there are physicaloperations challenges that are
kind of earlier on the AIadoption curve, where you are
(09:10):
seeing retailers and differentcompanies test and actually do
successful drone deliveriesright at the moment.
And then the scaling challenges.
So where can you use AI todrive automation and better
decision making across the org?
Where can you use AI to driveautomation and better
decision-making across the org?
There's a very interestingapplications there of things
like um route optimization right, can you try and make a
(09:32):
decision on where to routesomething the best um using AI
so that you can do that at scaleand very rapidly and kind of
replicate what it might.
Take a team you know, uh, 10minutes to do it in.
Take a team you know 10 minutesto do in seconds.
(09:55):
And there's also someinteresting kind of workforce
safety elements to think about.
When I'm scaling, my riskincreases of safety right and my
costs increase.
And I've seen companiesimplement successfully computer
vision to do kind of monitoringright of the work environment.
And as we start scaling andit's harder for us to have
oversight, are things safe, arethings running well?
(10:16):
But then the last piece of it,the experience piece, is one
that I think is earliest on inthe adoption curve, which is
kind of the most nascent.
When we talk about applying AIto improve customer experience,
most retailers today will befamiliar with things like
chatbot automation, right, and akind of experience test point
(10:37):
with the customer how do I meetthe customer at every sort of
communication channel, how do Ifulfill distribution.
But that piece of how do weimprove the experience is an
area where I'm really interestedin.
How can we use kind ofreal-time verification,
real-time data, to know that ouroperations are happening as we
(10:58):
expect?
Right, whether that be driving,better data visibility, data
capture, better oversight intoprocess.
So AI, the bottom line, isalready being adopted in all
these places, right, and itgenerally just enables the
company to scale efficiently andyou no longer.
(11:20):
The potential is that you nolonger have that kind of
accepted belief of as operationsgrow, complexity grows, issues
grow.
It's kind of what if you couldscale and actually reduce
defects and missing orders inreal time.
Dr. Matt Waller (11:40):
So what are some of the challenges of
adopting AI in supply chainwithin large enterprises?
Charlotte Bax (11:49):
Yeah, ba, I'd be curious to hear your thoughts
too on where you think we are inspeaking to retail execs in
terms of AI adoption.
A few of the things that wehear as blockers because people
are buying AI.
It's a new type of technology.
They're kind of unsure abouthow to assess it.
The three things that I hearconsistently are kind of who
(12:11):
owns it right?
This sits at the intersectionof maybe the business team is
driving the decision making, butthey require heavy
collaboration with the technicalteams and the product teams.
So one is just a challenge ofownership and cross-team
collaboration.
The second is how do we knowit's working?
So we implement AI right.
(12:33):
How do we know when it's livein production, it's working and
it's not hallucinating, right,it's not making the wrong
decision?
So what are the checks in place?
And then the third thing isquite an interesting it's a
growing trend I'm seeing of whatare the legal implications,
what are the data implications?
Ai requires the capturing of,you know, a lot of data to work
(12:59):
and I see business teams unsureabout data security, data
privacy, ramifications.
So it's a classic.
Like you know who owns it, howdo we measure it and what's the
risk?
Dr. Matt Waller (13:13):
Yeah, companies are concerned especially with
employees using publiclyavailable AI because of its
potential of leaking IP,essentially, and trade secrets
(13:38):
and things like that.
So a lot of these companies, Ithink, are starting to build
their own systems.
They may be using somethingthat's already out there an open
source tool, but they'rebuilding their own systems and
then optimizing on their data aswell, Because a lot of these
(14:01):
companies depends, but a lot ofcompanies in supply chain retail
, cpg, et cetera they've got alot of data.
They've actually got a hugeamount of information to train
large language models on, butfrom what I've seen, everyone is
(14:25):
still very much in a.
You know, we're on the verybeginning of the curve on how to
learn this.
The adoption, the initialadoption, has been incredibly
high, but the application from apractical perspective has been
(14:45):
a little slow, from what I cantell.
Now.
Some companies are faster thanothers, for sure, and I think
many companies have been usingcertain types of AI, like neural
networks and things like that,for a long time.
But this is different, and itreally is different in so many
ways, and so I think there'sstill a lot of experimentation
(15:12):
going on.
But I'd like to know what areyou personally excited about in
terms of AI?
Charlotte Bax (15:18):
Well, a couple trends.
I mean, you and I were chattingabout personally how we're
using Gen AI in our day-to-daywork, our research, right, and
so I think first OpenAI and nowa plethora of other tools from
Microsoft, google, have justmade AI so accessible, and we've
(15:41):
seen it.
You know, after ChatGPT, youhad this big wave and suddenly
you know, 2024, 2025,implementing an AI strategy is
now top of the board deck formost companies, and I think
that's so exciting becausepeople obviously see the
potential and even if they'reusing it in their own personal
day-to-day work or projects,it's just really sped up the
(16:06):
awareness of AI.
But, like you said, it's not AIhas been around for 20 years in
applications, but it's justthis type of accessible AI,
general AI.
So I'm excited about mindsetsopening and shifting and people
being open.
In terms of the actualtechnologies, I'm really excited
about two things.
(16:27):
One, our phones are so good nowright, and mobile devices are
so good.
Hardware from people likeNVIDIA and Qualcomm has really
sped up, and so now you can do alot of AI on what they call
edge, so actually doing thecompute on the device that
enables you to tackle challenges, like we see at Capture, where
(16:51):
the AI needs to happen in realtime in the moment and you might
have constraints like offlinelow signal.
So the ability to use these newkind of advances in mobile and
do edge compute in very realapplications is very exciting.
And then also when you talkabout companies adopting AI,
(17:17):
especially if they've got theirown data set to play with or to
start with, there are somereally interesting realms in
synthetic data which allow andI'm sure people personally might
have played around with kind oftext-to-image generators.
Dolly right, you type in I wanta picture of a cat with a
birthday hat on, and they'repretty bad, and they're bad in
(17:40):
funny ways now, but you can seehow they get better and better.
But I've seen some recentadvancements from Meta which are
quite astonishing, where youcan type in text and get a 3D
output.
So if you think about processesin retail, like products online
, giving people a 3D experienceof what a product looks like,
maybe you can start to generatethat and create that.
(18:00):
Merchandiser could do that.
They don't need to be aspecialist graphic designer.
Things don't need to takemonths.
So I'm excited about those waysthat people are thinking about
AI and how we can rapidlyprototype.
Dr. Matt Waller (18:13):
You know, I use an AI tool called PoE.
It gives me access to hundredsof applications and I love it.
I have it on my phone, I haveit on my computer.
I'm currently building a houseand I started thinking about and
you'll see how this applies tomerchandising in a house.
(18:35):
And I started thinking aboutand you'll see how this applies
to merchandising in a moment.
But I started thinking aboutour outdoor cooking area, patio,
and I know I'm going to have agas grill.
I also know I'm going to have abig green egg Love it.
So if I'm going to have a biggreen egg, then I'm going to
need lump charcoal.
(18:56):
I also want a trash can outthere.
There's a few things I want.
I also want a staging area formy firewood for my house Not
permanent, but staging area andall in this brick enclosure.
So I entered into Claude Sonnet3.5, the dimensions of
(19:23):
everything I'm going to buy, howmuch space I have for this
brick structure, et cetera, etcetera.
Right, yeah.
I said, given all of this, comeout, come up with a description
of how I could lay all this out.
Charlotte Bax (19:41):
Brilliant.
Dr. Matt Waller (19:42):
And it came up with one and there was something
I didn't like about it.
I said, well, I don't like this, and it changed it.
And again, this was not visual,it was just a description, but
detailed.
It took into account thedimensions of everything.
And then I said, okay, nowrewrite this so that I can take
the text and put it into animage generator to generate an
(20:07):
image of what this would looklike.
So it did.
I took that text, I put it inan image generator.
One nice thing about Poe is, ifyou do that, it'll generate an
image for you know, using thetool you pick.
But then at the bottom it'llsay would you like to try one of
these other image generators?
(20:28):
So I kept collecting them untilI found the one I liked.
I probably went through 15.
And then I did some.
The final version looked like aphotograph of exactly what we
want.
So I was able to print that outand give it to my builder and
(20:54):
say you know, here's what I'mlooking for.
And my point of this is that ingoing through that process I
realized that a retailer, ane-tailer, an e-commerce company
could easily provide this kindof text in advance or even use,
you know an image generator.
In other words, you click theitems you want, you say here's
(21:16):
how much space I have, orwhatever, and it could generate
options for you.
That's for the consumer, butthen for the merchandising.
Right now, planet grams arestill very archaic from the one
(21:37):
part seed.
It seems like there's a greatapplication here using that same
kind of logic.
Right when you put thedimensions in, you say the case
pack has this much, we've got tohave at least this many face,
this much capacity given thesize of the case.
Back, on and on and on.
You would think that you couldcome up with an AI tool to
(22:01):
optimize that.
Charlotte Bax (22:02):
Absolutely, Absolutely To try and do.
I mean, that's what AI is greatat.
It's great at taking multipletypes of input, whether numbers,
text, images, combining ittogether and giving you a
starting place.
And yes, you're going to havean expert come in and shift
things.
But I mean how empowering foryou to say, actually, I don't
(22:24):
need to start with hiring areally expensive architect or an
interior designer.
I mean, I can at least get offthe starting blocks myself,
designer.
I can at least get off thestarting blocks myself, and that
makes me more informed consumerand that makes me more likely
to purchase.
Probably because you don't feellike I don't know where to
start.
I've got this kind of image inmy head.
(22:45):
I don't know if it's going towork.
Dr. Matt Waller (22:48):
That's right.
Charlotte Bax (22:49):
Right.
So it probably drives higherconversion if you can empower
your customer to have moreknowledge?
Dr. Matt Waller (22:57):
I would think so.
Now back up the supply chain,or even before that stage,
charlotte?
Well, you're the founder andCEO of Capture, and tell us a
little bit about what Capturedoes.
Charlotte Bax (23:15):
Yeah, so Capture?
We're really at thisintersection of quality and
supply chain.
The origin of the company isreally a combination of my
background in retail with GapInc, my experience on some of
these supply chain challengesand also experience in AI and
(23:37):
how you can apply it.
What we're trying to help ourcustomers and retailers solve is
how can they scale operationswhile maintaining quality and
how can they actually makequality their USB and get rid of
this accepted truth inoperations of as you scale and
add complexity, experience,customer experience suffers,
(24:02):
right.
You always plan to see likekind of an increase in support
tickets, right With growth inthe business and that doesn't
have to happen.
And so how we actually do thatand what capture is?
It's a quality control platform,automation platform, and we
take all these processes thatretailers do today.
Would that be like education ofthe frontline workforce,
(24:25):
building policies.
I got you know what's myinternal policy for returns,
what's my internal policy fordeliveries, and those are off,
those are written in pdf docsand they're pretty stagnant.
So we take those policies, weingest them, we understand them,
we turn them into machinelearning models and we apply the
(24:45):
policy in real time throughphotos that are taken on the
spot by customers or workforce.
So, if you take a photo of adelivery in front of a door,
what capture is going to bedoing is automatically checking,
well, is it delivered to theright place?
You know, hitting theguidelines from what the.
(25:05):
You know the company specifiedwhere the customer wants it, is
it the right address?
And.
But it's doing all that in realtime.
Then, as the output, you know,the company can be really
confident that they're going toscale while managing quality.
We've got a full audit historyof what's going on, um, but at
the end of the day, it's allabout helping companies drive
(25:26):
growth, drive retention, becausethey're improving quality.
The supply chain, um and thisis part of why I'm really
excited about these mobile andwhat we can do with edge compute
, because real-time verification, real-time AI, simply just
wasn't possible beforehand,right, and that's why retailers
(25:48):
have been trying to solve thisin various different ways
forever.
I mean, what was yourexperience?
You know, how would you thinkquality and this kind of quality
oversight has changed in thepast 20 years?
Has it changed at all?
Really?
Are we still doing things thesame way?
Dr. Matt Waller (26:07):
Well, it's interesting, when I first got
involved in quality it was backin the 80s, and you know it was
back in the 80s and you know itwas a hot topic.
(26:27):
Because back then quality, youknow, everyone started realizing
Because by 1980, I'd say 1980,most people thought of Japanese
products, like cars, as beinglower quality.
By 1985, everyone realized, oh,we were wrong about that.
They may have been low onfeatures and premium, but in
(26:51):
terms of reliability, durability, et cetera, et cetera, they
were way ahead of everybody.
Charlotte Bax (26:57):
Yeah.
Dr. Matt Waller (26:59):
And so there was a big push because everyone
thought Japan was going to takeover the world and become the
world's manufacturer, and soeveryone, especially in the US,
there was this huge push toimplement things like
statistical quality control,quality management.
(27:20):
Later, kaizen Lean was muchlater in Six Sigma, but they're
all very similar in many ways.
I think that retailers startedimplementing it a lot in the
early 90s.
(27:40):
Certain retailers, I thinkcompanies, have taken their eye
off of it in some ways They'vebeen a little more focused on
things like product management,which is very important.
It is, and I actually think AIcan help with product management
in a lot of ways.
(28:00):
But I can see AI helping inespecially large language models
in areas not necessarilystatistical process control I
think there's easy ways to dealwith that but in terms of
because a big part of qualitymanagement has to do with
(28:23):
generating, for example.
For example I'll give you onetool just as an example.
I think you look at like afishbone diagram, right when you
you start with a problem andthen you try to say why did that
occur?
And then why did that occur?
And you go back at least fivewhys you know.
(28:45):
Well, large language modelsshould be able to generate that
automatically.
Charlotte Bax (28:53):
Yeah, right.
Dr. Matt Waller (28:53):
So that automatically?
Yeah, right, yeah, so that'sthe thing.
And so if you're looking atsome error and the supply chain,
it doesn't really matter whereit is, and you apply a tool like
a fishbone diagram and you goback five levels, well, large
language models should will dothat very easily.
(29:16):
So I'm thinking that at thispoint there's lots of
opportunities in that way.
But also to your point earlier,I think marrying computer
vision with AI is a hugeopportunity in the supply chain,
all areas in the supply chain.
Charlotte Bax (29:40):
Absolutely, it's probably true.
You said, too, on equality.
Maybe hasn't been the emphasisin the past 10 years because
actually people were just tryingto get.
They were just really focusedon getting to parity with speed
of delivery, the push for thesame day and speed is one very
(30:02):
important element of the overallexperience.
How convenient is this purchase?
But with that traditional kindof even Japanese manufacturing
systems, and in manufacturingyou've got the concept of
quality management systems thathave existed for a long time and
(30:25):
they're very important tenantsof especially highly regulated
industries or manufacturingindustries, right.
So why don't retailers and whydon't new on-demand platforms
have quality management systems,right?
What is the new system?
And it's going to be acombination of things.
(30:45):
Like you said, there'ssomething you can do with LLMs
helping to prompt people, youknow, help people work through
decision making, the fishbonediagram and then computer vision
can be used as this tool, whichreally gives you eyes on the
ground, right, and virtually canyou know what's going on with
(31:07):
every delivery and can youvalidate the right things
happening.
So I think we'll definitely seea shift.
And this is how retailers wantto differentiate, right.
They want to differentiate byexperience.
If everyone is almost on paritywith speed.
How are you going to driveretention Really?
(31:27):
It's going to be aboutdelivering the customer a
different experience or aconsistent experience.
Dr. Matt Waller (31:32):
So, charlotte, what advice do you have to
companies that are consideringthe decision of build versus buy
in-house when it comes to AIsolutions?
Charlotte Bax (31:46):
Yeah, it's such a hot topic.
Everyone's really trying todefine this internally.
So, on one hand, you've, um,the kind of existing product
tenant which is you buildinternally for for strategy,
kind of is a strategic or or youbuy for parity, right.
(32:09):
So are you trying to kind ofmatch the market in shifts in
demand and maintain featureparity, or are you trying to
strategically, internally, makea big bet about your product.
So that's, I think, more fromthe product and technical teams,
what an existing tenant thatthey would look at.
(32:30):
That they would look at.
So, for example, a companymight choose to work with a kind
of customer support AI botbecause they know they need to
provide a customer support tool,a CRM, but it may not be the
(32:51):
thing For a retailer, it may notbe the thing that they want to
strategically differentiate with, right, they need to provide a
level of service, but they'renot a customer support platform
internally, right.
So that's one way to thinkabout it.
I really encourage and I'mseeing more and more of this,
(33:12):
which is great I encouragecompanies to prototype and play
around with what you can do withlarge vision models, large
language models, but there'ssuch a large gap between
productizing AI and reallyhaving it in production, because
it involves all of this extrabits around.
(33:34):
How are you going to monitor,maintain it continually, update
models, how are you going tobuild machine learning teams and
infrastructure around this?
And we've even had companiescome back to us, right, who have
tried to build their own modelsand just said it's way too hard
from a process and managingperspective.
Like prototype, so that you cankind of understand what you
(33:57):
want, just like you did withyour furniture right in your
backyard.
You know prototype.
But then if you want to trulysee the benefit of AI, launch it
in production and have a reallyscalable AI system, that's
where partnering can be yourbest bet, because, really, teams
(34:20):
if they underestimate the kindof the resource that goes into
and building all the scaffoldingthat you need for not just
building but continuallymonitoring and maintaining AI
systems, that can be a rudeawakening right Six months, 12
months down the line ininvestment.
So partners are the best there.
(34:40):
We're not just buying off theshelf.
Right, you have an ability toco-collaborate.
You have the ability to know,you know co-collaborate on
roadmap and that sort of thing.
But those are the type ofthings that I'd suggest leaders
think about.
But I guess the last elementwith AI is to think about is how
to define your problem.
(35:01):
So AI problems can be really.
They can demand really reallyhigh accuracy.
So it's kind of what's the riskif this goes wrong?
High accuracy, so it's kind ofwhat's the risk if this goes
wrong?
So if a chatbot gives the wrongarticle recommendation to a
customer, it's not that big adeal.
(35:21):
It's not ideal, but it's notthat big a deal versus
applications of AI where you'reneeding to reroute large
shipments or you're doingmedical diagnosis right.
There you think about theaccuracy has to be 99.9%, right,
Because the cost of failure isso high.
(35:42):
And so that's another criticalthing for business leaders to
think about.
Like, where is our problem onthe kind of risk of failure
scale?
Because the higher the risk,the more control you're going to
want to have over thingsinternally, right?
Um, and then you know what typeof problem is this?
Is this a image problem?
(36:03):
Is this a language problem?
Is this a combination problem?
So, hopefully that, yeah, thatstarts to frame it.
But it's a different type ofpurchase from things we've seen
in the past, for sure what aresome of your thoughts around how
this is changing consumerbehavior?
(36:25):
so I'll be really interested tosee how this is changing how the
next generation shops.
You think about high schoolstudents today.
They've already adopted ebt tohelp with homework prompts,
right, I bet they're researchingusing llms instead of using
(36:46):
google, so they're alreadylearning these behaviors, just
like the generation before Ilearned how to use iphone.
Right, they're using, they'relearning this behavior of prompt
writing, whether they know thatconsciously or not.
And so there are going to be somany interesting things, even
how consumers are going to wantto navigate retail websites.
(37:07):
Right, retail websites arebuilt on a very hierarchical,
structured approach.
Right, I go to the category, Iselect the thing, whereas the
next generation are used tofreeform text.
You know, I'm looking for a redshirt that's great for summer
(37:28):
and can be worn with jeans,right?
So we haven't even started tosee really the full impact of
that consumer shift.
But I think as the currentgeneration, the kind of
teenagers now, as they gain morepurchasing power, right, you'll
start to see a bigger demand onretail to actually change the
(37:51):
shopping experience.
On retail, to actually changethe shopping experience.
The past five, 10 years haveall been about as we talked
about, consumer demands forspeed, convenience.
So probably the next wave isgoing to be all about ease of
(38:13):
experience.
So who knows what websites willlook like in 10 years' time?
But I bet they'll look quitedifferent.
Dr. Matt Waller (38:19):
Well, Charlotte , thank you so much for taking
time to join us today.
I really appreciate it.
Charlotte Bax (38:24):
Thanks so much, Matt.
Yeah, I really enjoyed theconversation.
Dr. Matt Waller (38:27):
If you're finding value in this podcast.
We greatly appreciate yoursupport by subscribing to our
YouTube channel podcast.
We'd greatly appreciate yoursupport by subscribing to our
YouTube channel.
Additionally, following us onApple and Spotify and leaving up
to a five-star review would beimmensely helpful.
We welcome any feedback orquestions related to the podcast
(38:48):
, as well as suggestions forfurther topics and guests.
You can leave your comments onour YouTube channel and rest
assured that I will read eachand every one of them.
Please also take a moment tocheck out our podcast sponsors,
as they play a critical role inkeeping this podcast running.
For more information onspecific topics, timestamps or
links to articles mentionedduring the podcast, head over to
(39:09):
mattwallerpodcastcom.
Welcome to the Matt Waller podcast, where we
look at success at theintersection of technology,
logistics, supply chain, retailand CPG, also known as the
retail value chain.
I want to clarify that thispodcast is distinct from my
responsibilities as a professorin the Sam M Walton College of
Business.
Nonetheless, it aligns with myaspiration to provide practical
(00:31):
insights to professionals andbusiness by showcasing companies
and people that can enhanceyour ability to manage, lead and
strategize and marketeffectively in the retail value
chain.
Before we dive into today'sexciting episode, I'd like to
thank our sponsor, new RoadCapital Partners.
New Road invests in proventechnologies, services and
(00:53):
products that serve unmet needsin the marketplace.
They look for companies insupply chain and logistics, as
well as consumer-orientedcompanies.
For more information, go tonewroadcpcom.
I would also like to disclosethat I am a strategic advisor to
New Road.
I'd also like to recognizepodcastvideoscom for the
(01:18):
services they provide for thesepodcasts.
I'm very pleased with theirservices and now, without
further ado, let's get into theexciting episode.
I have with me today CharlotteBax, who is the founder and CEO
of Capture, which is a AI-drivenquality control solution for
(01:39):
supply chains, and I thinkyou're going to really like this
discussion If you're in supplychain management merchandising,
retailing, cpg supply chain.
If you're an investor, I thinkyou'll be interested in this.
(02:00):
This company that Charlottefounded was started in a Google
accelerator four years ago andtoday Capture already serves
Fortune 100 company supplychains across Europe, us and
APAC.
Charlotte is an expert inapplying AI to help supply
chains deliver better customerexperiences, as you will hear,
(02:23):
but unlike typical AI founders,her technical knowledge is
largely self-taught.
She started her career inproduct operations at Gap Inc.
She holds an MBA from theLondon Business School and has
been building AI products forthe past eight years.
Personally, she is passionateabout making AI accessible and
(02:48):
partnered on programs with NASAand Google to promote women in
STEM.
This particular episode today.
Again, it's looking at thequality challenge of modern
supply chains and how AI canaddress them.
But we dive into the criticaltopic of quality for modern
(03:09):
supply chains, exploringchallenges that retailers face
in maintaining service standardsacross multiple distribution
channels, from online deliveryto scheduled and in-store pickup
.
We also examine themultifaceted problem of trust
that emerges in this landscape,including customer retailer
(03:30):
trust, fraud prevention andinternal workforce trust.
Charlotte, thank you so muchfor joining me today.
Charlotte Bax (03:38):
Thanks so much for having me.
Matt, Really excited to chattoday.
Dr. Matt Waller (03:42):
Charlotte, I'm really excited about this
podcast, Really excited to chattoday.
Charlotte, I'm really excitedabout this podcast and I'd like
to ask you what are some of thebig challenges that the supply
chain faces, especially in termsof quality.
Charlotte Bax (03:59):
Yeah, I think it's such an interesting topic
and quality in the supply chainis not new, but I think it's one
that we've all probably hadexperience, acute experience
with, as customers.
Right, think about you knowwhen you order something online,
right, or you schedule adelivery, you expect it to be,
you know, fulfilled on time atyour house in the right state.
(04:22):
And if you've ever experiencedas a customer something going
wrong in that process gettingyour order and it's the wrong
thing, or it's damaged or itcomes late you know from a
customer's point of view howimportant quality is in the
supply chain, because customersare really demanding.
To say no, I might be patientonce, but I'm not going to be
(04:43):
patient twice if kind of issueskeep happening.
But for the supply chains, Ithink what's interesting is why
it's becoming harder and harderto manage.
Quality is especially in thelast five years with COVID,
we've seen lots of retailersfocusing on trying to meet the
customer where they are right,so offering new modes of
(05:08):
distribution, right.
So shop online, pick up instore, schedule delivery,
on-demand delivery, and that'screating all of these new
challenges, which I think aboutin kind of three buckets.
You know, one is theoperational challenge of just
how do you distribute to thesenew channels and meet the
customer where they are.
(05:29):
Second is experience challengeson how do you drive retention,
how do you make sure that thingsgo right every time.
And then the last one is reallythe scale challenge on how do
you do this profitably?
How do you build an automation?
Last one is really the scalechallenge on how do you do this
profitably, how do you build anautomation.
So there are those threechallenges operational,
(05:49):
experience and scale which everysupply chain face today, but
they face acutely because ofthese new types of modes.
I think of distribution.
Dr. Matt Waller (05:56):
Yeah, you know mentioning Omnichannel,
especially for brick and mortartypes of stores.
It seems like, well, one sortof tenant of quality management
is that when you introducecomplexity, quality problems
(06:19):
increase.
And so before you had aretailer that all of a sudden
you know eventually was sellingonline, maybe on their dot com.
But Omni Channel reallycomplicates that because all of
a sudden you're doing you'reselling online, but you're also
(06:39):
there's buy online, pick up atstore, buy online, deliver to
home, buy online, deliver inhome, drone delivery.
There's all these differentoptions and much of it for
traditional brick and mortarretailers is being done through
brick and mortar stores, whichis making store operations more
(07:03):
complicated as well.
Stores, which is making storeoperations more complicated as
well.
And I know, even at themerchandising level, most of
these brick-and-mortar retailersthat have gone omnichannel,
their merchants focused more on.
You know they'd have merchantsfocused on the brick-and-mortar
and then merchants focused one-commerce and over the years
(07:26):
they've started bringing themtogether.
But again, it's causing lots ofchallenges.
So you mentioned a lot of thesechallenges.
So my question is how can AIhelp to solve those?
Charlotte Bax (07:40):
Yeah, absolutely, and you're spot on.
I mean, especially when youtalk about omni-channel types of
channels.
Yeah, absolutely, and you'respot on.
I mean especially when you talkabout Omnichannel.
I think Omnichannel that hasreally been the focus for the
past 10 years, and even in mytime at Gap Inc, we were going
through that transformation oftaking the store merchandise
team and the online team andbringing them together and
thinking about unifiedassortment, unified distribution
.
I should just have one pool ofinventory, right.
(08:02):
But now everything's changingagain and changing on these
retailers and don't just havethe store and online.
We've got all these micro hubsand micro sites.
How do we adopt a hub and spokemodel to serve that?
I think AI and the other hand,too, you've got also rising
(08:25):
theft and kind of fraud in themix, right, and so where AI can
really help, I think, if wethink about operational
experience and scale,operations-wise, you're seeing
pretty early adoption of a tonof different technologies, like
how do we distribute to remoteplaces with drones, right.
(08:48):
So AI can really help in kindof distribution.
You're seeing people use AI forautomated picking and sorting.
You're seeing robotics at theactual distribution centers
themselves, right.
So I think there are physicaloperations challenges that are
kind of earlier on the AIadoption curve, where you are
(09:10):
seeing retailers and differentcompanies test and actually do
successful drone deliveriesright at the moment.
And then the scaling challenges.
So where can you use AI todrive automation and better
decision making across the org?
Where can you use AI to driveautomation and better
decision-making across the org?
There's a very interestingapplications there of things
like um route optimization right, can you try and make a
(09:32):
decision on where to routesomething the best um using AI
so that you can do that at scaleand very rapidly and kind of
replicate what it might.
Take a team you know, uh, 10minutes to do it in.
Take a team you know 10 minutesto do in seconds.
(09:55):
And there's also someinteresting kind of workforce
safety elements to think about.
When I'm scaling, my riskincreases of safety right and my
costs increase.
And I've seen companiesimplement successfully computer
vision to do kind of monitoringright of the work environment.
And as we start scaling andit's harder for us to have
oversight, are things safe, arethings running well?
(10:16):
But then the last piece of it,the experience piece, is one
that I think is earliest on inthe adoption curve, which is
kind of the most nascent.
When we talk about applying AIto improve customer experience,
most retailers today will befamiliar with things like
chatbot automation, right, and akind of experience test point
(10:37):
with the customer how do I meetthe customer at every sort of
communication channel, how do Ifulfill distribution.
But that piece of how do weimprove the experience is an
area where I'm really interestedin.
How can we use kind ofreal-time verification,
real-time data, to know that ouroperations are happening as we
(10:58):
expect?
Right, whether that be driving,better data visibility, data
capture, better oversight intoprocess.
So AI, the bottom line, isalready being adopted in all
these places, right, and itgenerally just enables the
company to scale efficiently andyou no longer.
(11:20):
The potential is that you nolonger have that kind of
accepted belief of as operationsgrow, complexity grows, issues
grow.
It's kind of what if you couldscale and actually reduce
defects and missing orders inreal time.
Dr. Matt Waller (11:40):
So what are some of the challenges of
adopting AI in supply chainwithin large enterprises?
Charlotte Bax (11:49):
Yeah, ba, I'd be curious to hear your thoughts
too on where you think we are inspeaking to retail execs in
terms of AI adoption.
A few of the things that wehear as blockers because people
are buying AI.
It's a new type of technology.
They're kind of unsure abouthow to assess it.
The three things that I hearconsistently are kind of who
(12:11):
owns it right?
This sits at the intersectionof maybe the business team is
driving the decision making, butthey require heavy
collaboration with the technicalteams and the product teams.
So one is just a challenge ofownership and cross-team
collaboration.
The second is how do we knowit's working?
So we implement AI right.
(12:33):
How do we know when it's livein production, it's working and
it's not hallucinating, right,it's not making the wrong
decision?
So what are the checks in place?
And then the third thing isquite an interesting it's a
growing trend I'm seeing of whatare the legal implications,
what are the data implications?
Ai requires the capturing of,you know, a lot of data to work
(12:59):
and I see business teams unsureabout data security, data
privacy, ramifications.
So it's a classic.
Like you know who owns it, howdo we measure it and what's the
risk?
Dr. Matt Waller (13:13):
Yeah, companies are concerned especially with
employees using publiclyavailable AI because of its
potential of leaking IP,essentially, and trade secrets
(13:38):
and things like that.
So a lot of these companies, Ithink, are starting to build
their own systems.
They may be using somethingthat's already out there an open
source tool, but they'rebuilding their own systems and
then optimizing on their data aswell, Because a lot of these
(14:01):
companies depends, but a lot ofcompanies in supply chain retail
, cpg, et cetera they've got alot of data.
They've actually got a hugeamount of information to train
large language models on, butfrom what I've seen, everyone is
(14:25):
still very much in a.
You know, we're on the verybeginning of the curve on how to
learn this.
The adoption, the initialadoption, has been incredibly
high, but the application from apractical perspective has been
(14:45):
a little slow, from what I cantell.
Now.
Some companies are faster thanothers, for sure, and I think
many companies have been usingcertain types of AI, like neural
networks and things like that,for a long time.
But this is different, and itreally is different in so many
ways, and so I think there'sstill a lot of experimentation
(15:12):
going on.
But I'd like to know what areyou personally excited about in
terms of AI?
Charlotte Bax (15:18):
Well, a couple trends.
I mean, you and I were chattingabout personally how we're
using Gen AI in our day-to-daywork, our research, right, and
so I think first OpenAI and nowa plethora of other tools from
Microsoft, google, have justmade AI so accessible, and we've
(15:41):
seen it.
You know, after ChatGPT, youhad this big wave and suddenly
you know, 2024, 2025,implementing an AI strategy is
now top of the board deck formost companies, and I think
that's so exciting becausepeople obviously see the
potential and even if they'reusing it in their own personal
day-to-day work or projects,it's just really sped up the
(16:06):
awareness of AI.
But, like you said, it's not AIhas been around for 20 years in
applications, but it's justthis type of accessible AI,
general AI.
So I'm excited about mindsetsopening and shifting and people
being open.
In terms of the actualtechnologies, I'm really excited
about two things.
(16:27):
One, our phones are so good nowright, and mobile devices are
so good.
Hardware from people likeNVIDIA and Qualcomm has really
sped up, and so now you can do alot of AI on what they call
edge, so actually doing thecompute on the device that
enables you to tackle challenges, like we see at Capture, where
(16:51):
the AI needs to happen in realtime in the moment and you might
have constraints like offlinelow signal.
So the ability to use these newkind of advances in mobile and
do edge compute in very realapplications is very exciting.
And then also when you talkabout companies adopting AI,
(17:17):
especially if they've got theirown data set to play with or to
start with, there are somereally interesting realms in
synthetic data which allow andI'm sure people personally might
have played around with kind oftext-to-image generators.
Dolly right, you type in I wanta picture of a cat with a
birthday hat on, and they'repretty bad, and they're bad in
(17:40):
funny ways now, but you can seehow they get better and better.
But I've seen some recentadvancements from Meta which are
quite astonishing, where youcan type in text and get a 3D
output.
So if you think about processesin retail, like products online
, giving people a 3D experienceof what a product looks like,
maybe you can start to generatethat and create that.
(18:00):
Merchandiser could do that.
They don't need to be aspecialist graphic designer.
Things don't need to takemonths.
So I'm excited about those waysthat people are thinking about
AI and how we can rapidlyprototype.
Dr. Matt Waller (18:13):
You know, I use an AI tool called PoE.
It gives me access to hundredsof applications and I love it.
I have it on my phone, I haveit on my computer.
I'm currently building a houseand I started thinking about and
you'll see how this applies tomerchandising in a house.
(18:35):
And I started thinking aboutand you'll see how this applies
to merchandising in a moment.
But I started thinking aboutour outdoor cooking area, patio,
and I know I'm going to have agas grill.
I also know I'm going to have abig green egg Love it.
So if I'm going to have a biggreen egg, then I'm going to
need lump charcoal.
(18:56):
I also want a trash can outthere.
There's a few things I want.
I also want a staging area formy firewood for my house Not
permanent, but staging area andall in this brick enclosure.
So I entered into Claude Sonnet3.5, the dimensions of
(19:23):
everything I'm going to buy, howmuch space I have for this
brick structure, et cetera, etcetera.
Right, yeah.
I said, given all of this, comeout, come up with a description
of how I could lay all this out.
Charlotte Bax (19:41):
Brilliant.
Dr. Matt Waller (19:42):
And it came up with one and there was something
I didn't like about it.
I said, well, I don't like this, and it changed it.
And again, this was not visual,it was just a description, but
detailed.
It took into account thedimensions of everything.
And then I said, okay, nowrewrite this so that I can take
the text and put it into animage generator to generate an
(20:07):
image of what this would looklike.
So it did.
I took that text, I put it inan image generator.
One nice thing about Poe is, ifyou do that, it'll generate an
image for you know, using thetool you pick.
But then at the bottom it'llsay would you like to try one of
these other image generators?
(20:28):
So I kept collecting them untilI found the one I liked.
I probably went through 15.
And then I did some.
The final version looked like aphotograph of exactly what we
want.
So I was able to print that outand give it to my builder and
(20:54):
say you know, here's what I'mlooking for.
And my point of this is that ingoing through that process I
realized that a retailer, ane-tailer, an e-commerce company
could easily provide this kindof text in advance or even use,
you know an image generator.
In other words, you click theitems you want, you say here's
(21:16):
how much space I have, orwhatever, and it could generate
options for you.
That's for the consumer, butthen for the merchandising.
Right now, planet grams arestill very archaic from the one
(21:37):
part seed.
It seems like there's a greatapplication here using that same
kind of logic.
Right when you put thedimensions in, you say the case
pack has this much, we've got tohave at least this many face,
this much capacity given thesize of the case.
Back, on and on and on.
You would think that you couldcome up with an AI tool to
(22:01):
optimize that.
Charlotte Bax (22:02):
Absolutely, Absolutely To try and do.
I mean, that's what AI is greatat.
It's great at taking multipletypes of input, whether numbers,
text, images, combining ittogether and giving you a
starting place.
And yes, you're going to havean expert come in and shift
things.
But I mean how empowering foryou to say, actually, I don't
(22:24):
need to start with hiring areally expensive architect or an
interior designer.
I mean, I can at least get offthe starting blocks myself,
designer.
I can at least get off thestarting blocks myself, and that
makes me more informed consumerand that makes me more likely
to purchase.
Probably because you don't feellike I don't know where to
start.
I've got this kind of image inmy head.
(22:45):
I don't know if it's going towork.
Dr. Matt Waller (22:48):
That's right.
Charlotte Bax (22:49):
Right.
So it probably drives higherconversion if you can empower
your customer to have moreknowledge?
Dr. Matt Waller (22:57):
I would think so.
Now back up the supply chain,or even before that stage,
charlotte?
Well, you're the founder andCEO of Capture, and tell us a
little bit about what Capturedoes.
Charlotte Bax (23:15):
Yeah, so Capture?
We're really at thisintersection of quality and
supply chain.
The origin of the company isreally a combination of my
background in retail with GapInc, my experience on some of
these supply chain challengesand also experience in AI and
(23:37):
how you can apply it.
What we're trying to help ourcustomers and retailers solve is
how can they scale operationswhile maintaining quality and
how can they actually makequality their USB and get rid of
this accepted truth inoperations of as you scale and
add complexity, experience,customer experience suffers,
(24:02):
right.
You always plan to see likekind of an increase in support
tickets, right With growth inthe business and that doesn't
have to happen.
And so how we actually do thatand what capture is?
It's a quality control platform,automation platform, and we
take all these processes thatretailers do today.
Would that be like education ofthe frontline workforce,
(24:25):
building policies.
I got you know what's myinternal policy for returns,
what's my internal policy fordeliveries, and those are off,
those are written in pdf docsand they're pretty stagnant.
So we take those policies, weingest them, we understand them,
we turn them into machinelearning models and we apply the
(24:45):
policy in real time throughphotos that are taken on the
spot by customers or workforce.
So, if you take a photo of adelivery in front of a door,
what capture is going to bedoing is automatically checking,
well, is it delivered to theright place?
You know, hitting theguidelines from what the.
(25:05):
You know the company specifiedwhere the customer wants it, is
it the right address?
And.
But it's doing all that in realtime.
Then, as the output, you know,the company can be really
confident that they're going toscale while managing quality.
We've got a full audit historyof what's going on, um, but at
the end of the day, it's allabout helping companies drive
(25:26):
growth, drive retention, becausethey're improving quality.
The supply chain, um and thisis part of why I'm really
excited about these mobile andwhat we can do with edge compute
, because real-time verification, real-time AI, simply just
wasn't possible beforehand,right, and that's why retailers
(25:48):
have been trying to solve thisin various different ways
forever.
I mean, what was yourexperience?
You know, how would you thinkquality and this kind of quality
oversight has changed in thepast 20 years?
Has it changed at all?
Really?
Are we still doing things thesame way?
Dr. Matt Waller (26:07):
Well, it's interesting, when I first got
involved in quality it was backin the 80s, and you know it was
back in the 80s and you know itwas a hot topic.
(26:27):
Because back then quality, youknow, everyone started realizing
Because by 1980, I'd say 1980,most people thought of Japanese
products, like cars, as beinglower quality.
By 1985, everyone realized, oh,we were wrong about that.
They may have been low onfeatures and premium, but in
(26:51):
terms of reliability, durability, et cetera, et cetera, they
were way ahead of everybody.
Charlotte Bax (26:57):
Yeah.
Dr. Matt Waller (26:59):
And so there was a big push because everyone
thought Japan was going to takeover the world and become the
world's manufacturer, and soeveryone, especially in the US,
there was this huge push toimplement things like
statistical quality control,quality management.
(27:20):
Later, kaizen Lean was muchlater in Six Sigma, but they're
all very similar in many ways.
I think that retailers startedimplementing it a lot in the
early 90s.
(27:40):
Certain retailers, I thinkcompanies, have taken their eye
off of it in some ways They'vebeen a little more focused on
things like product management,which is very important.
It is, and I actually think AIcan help with product management
in a lot of ways.
(28:00):
But I can see AI helping inespecially large language models
in areas not necessarilystatistical process control I
think there's easy ways to dealwith that but in terms of
because a big part of qualitymanagement has to do with
(28:23):
generating, for example.
For example I'll give you onetool just as an example.
I think you look at like afishbone diagram, right when you
you start with a problem andthen you try to say why did that
occur?
And then why did that occur?
And you go back at least fivewhys you know.
(28:45):
Well, large language modelsshould be able to generate that
automatically.
Charlotte Bax (28:53):
Yeah, right.
Dr. Matt Waller (28:53):
So that automatically?
Yeah, right, yeah, so that'sthe thing.
And so if you're looking atsome error and the supply chain,
it doesn't really matter whereit is, and you apply a tool like
a fishbone diagram and you goback five levels, well, large
language models should will dothat very easily.
(29:16):
So I'm thinking that at thispoint there's lots of
opportunities in that way.
But also to your point earlier,I think marrying computer
vision with AI is a hugeopportunity in the supply chain,
all areas in the supply chain.
Charlotte Bax (29:40):
Absolutely, it's probably true.
You said, too, on equality.
Maybe hasn't been the emphasisin the past 10 years because
actually people were just tryingto get.
They were just really focusedon getting to parity with speed
of delivery, the push for thesame day and speed is one very
(30:02):
important element of the overallexperience.
How convenient is this purchase?
But with that traditional kindof even Japanese manufacturing
systems, and in manufacturingyou've got the concept of
quality management systems thathave existed for a long time and
(30:25):
they're very important tenantsof especially highly regulated
industries or manufacturingindustries, right.
So why don't retailers and whydon't new on-demand platforms
have quality management systems,right?
What is the new system?
And it's going to be acombination of things.
(30:45):
Like you said, there'ssomething you can do with LLMs
helping to prompt people, youknow, help people work through
decision making, the fishbonediagram and then computer vision
can be used as this tool, whichreally gives you eyes on the
ground, right, and virtually canyou know what's going on with
(31:07):
every delivery and can youvalidate the right things
happening.
So I think we'll definitely seea shift.
And this is how retailers wantto differentiate, right.
They want to differentiate byexperience.
If everyone is almost on paritywith speed.
How are you going to driveretention Really?
(31:27):
It's going to be aboutdelivering the customer a
different experience or aconsistent experience.
Dr. Matt Waller (31:32):
So, charlotte, what advice do you have to
companies that are consideringthe decision of build versus buy
in-house when it comes to AIsolutions?
Charlotte Bax (31:46):
Yeah, it's such a hot topic.
Everyone's really trying todefine this internally.
So, on one hand, you've, um,the kind of existing product
tenant which is you buildinternally for for strategy,
kind of is a strategic or or youbuy for parity, right.
(32:09):
So are you trying to kind ofmatch the market in shifts in
demand and maintain featureparity, or are you trying to
strategically, internally, makea big bet about your product.
So that's, I think, more fromthe product and technical teams,
what an existing tenant thatthey would look at.
(32:30):
That they would look at.
So, for example, a companymight choose to work with a kind
of customer support AI botbecause they know they need to
provide a customer support tool,a CRM, but it may not be the
(32:51):
thing For a retailer, it may notbe the thing that they want to
strategically differentiate with, right, they need to provide a
level of service, but they'renot a customer support platform
internally, right.
So that's one way to thinkabout it.
I really encourage and I'mseeing more and more of this,
(33:12):
which is great I encouragecompanies to prototype and play
around with what you can do withlarge vision models, large
language models, but there'ssuch a large gap between
productizing AI and reallyhaving it in production, because
it involves all of this extrabits around.
(33:34):
How are you going to monitor,maintain it continually, update
models, how are you going tobuild machine learning teams and
infrastructure around this?
And we've even had companiescome back to us, right, who have
tried to build their own modelsand just said it's way too hard
from a process and managingperspective.
Like prototype, so that you cankind of understand what you
(33:57):
want, just like you did withyour furniture right in your
backyard.
You know prototype.
But then if you want to trulysee the benefit of AI, launch it
in production and have a reallyscalable AI system, that's
where partnering can be yourbest bet, because, really, teams
(34:20):
if they underestimate the kindof the resource that goes into
and building all the scaffoldingthat you need for not just
building but continuallymonitoring and maintaining AI
systems, that can be a rudeawakening right Six months, 12
months down the line ininvestment.
So partners are the best there.
(34:40):
We're not just buying off theshelf.
Right, you have an ability toco-collaborate.
You have the ability to know,you know co-collaborate on
roadmap and that sort of thing.
But those are the type ofthings that I'd suggest leaders
think about.
But I guess the last elementwith AI is to think about is how
to define your problem.
(35:01):
So AI problems can be really.
They can demand really reallyhigh accuracy.
So it's kind of what's the riskif this goes wrong?
High accuracy, so it's kind ofwhat's the risk if this goes
wrong?
So if a chatbot gives the wrongarticle recommendation to a
customer, it's not that big adeal.
(35:21):
It's not ideal, but it's notthat big a deal versus
applications of AI where you'reneeding to reroute large
shipments or you're doingmedical diagnosis right.
There you think about theaccuracy has to be 99.9%, right,
Because the cost of failure isso high.
(35:42):
And so that's another criticalthing for business leaders to
think about.
Like, where is our problem onthe kind of risk of failure
scale?
Because the higher the risk,the more control you're going to
want to have over thingsinternally, right?
Um, and then you know what typeof problem is this?
Is this a image problem?
(36:03):
Is this a language problem?
Is this a combination problem?
So, hopefully that, yeah, thatstarts to frame it.
But it's a different type ofpurchase from things we've seen
in the past, for sure what aresome of your thoughts around how
this is changing consumerbehavior?
(36:25):
so I'll be really interested tosee how this is changing how the
next generation shops.
You think about high schoolstudents today.
They've already adopted ebt tohelp with homework prompts,
right, I bet they're researchingusing llms instead of using
(36:46):
google, so they're alreadylearning these behaviors, just
like the generation before Ilearned how to use iphone.
Right, they're using, they'relearning this behavior of prompt
writing, whether they know thatconsciously or not.
And so there are going to be somany interesting things, even
how consumers are going to wantto navigate retail websites.
(37:07):
Right, retail websites arebuilt on a very hierarchical,
structured approach.
Right, I go to the category, Iselect the thing, whereas the
next generation are used tofreeform text.
You know, I'm looking for a redshirt that's great for summer
(37:28):
and can be worn with jeans,right?
So we haven't even started tosee really the full impact of
that consumer shift.
But I think as the currentgeneration, the kind of
teenagers now, as they gain morepurchasing power, right, you'll
start to see a bigger demand onretail to actually change the
(37:51):
shopping experience.
On retail, to actually changethe shopping experience.
The past five, 10 years haveall been about as we talked
about, consumer demands forspeed, convenience.
So probably the next wave isgoing to be all about ease of
(38:13):
experience.
So who knows what websites willlook like in 10 years' time?
But I bet they'll look quitedifferent.
Dr. Matt Waller (38:19):
Well, Charlotte , thank you so much for taking
time to join us today.
I really appreciate it.
Charlotte Bax (38:24):
Thanks so much, Matt.
Yeah, I really enjoyed theconversation.
Dr. Matt Waller (38:27):
If you're finding value in this podcast.
We greatly appreciate yoursupport by subscribing to our
YouTube channel podcast.
We'd greatly appreciate yoursupport by subscribing to our
YouTube channel.
Additionally, following us onApple and Spotify and leaving up
to a five-star review would beimmensely helpful.
We welcome any feedback orquestions related to the podcast
(38:48):
, as well as suggestions forfurther topics and guests.
You can leave your comments onour YouTube channel and rest
assured that I will read eachand every one of them.
Please also take a moment tocheck out our podcast sponsors,
as they play a critical role inkeeping this podcast running.
For more information onspecific topics, timestamps or
links to articles mentionedduring the podcast, head over to
(39:09):
mattwallerpodcastcom.
Popular Podcasts
Fudd Around And Find Out
UConn basketball star Azzi Fudd brings her championship swag to iHeart Women’s Sports with Fudd Around and Find Out, a weekly podcast that takes fans along for the ride as Azzi spends her final year of college trying to reclaim the National Championship and prepare to be a first round WNBA draft pick. Ever wonder what it’s like to be a world-class athlete in the public spotlight while still managing schoolwork, friendships and family time? It’s time to Fudd Around and Find Out!
Crime Junkie
Does hearing about a true crime case always leave you scouring the internet for the truth behind the story? Dive into your next mystery with Crime Junkie. Every Monday, join your host Ashley Flowers as she unravels all the details of infamous and underreported true crime cases with her best friend Brit Prawat. From cold cases to missing persons and heroes in our community who seek justice, Crime Junkie is your destination for theories and stories you won’t hear anywhere else. Whether you're a seasoned true crime enthusiast or new to the genre, you'll find yourself on the edge of your seat awaiting a new episode every Monday. If you can never get enough true crime... Congratulations, you’ve found your people. Follow to join a community of Crime Junkies! Crime Junkie is presented by audiochuck Media Company.
24/7 News: The Latest
The latest news in 4 minutes updated every hour, every day.