Confessions Of Supply Chain Executives | The End Of Inventory Guesswork

Physical AI.

Physical AI.

It's the new topic du jour.

It sounds cool.

Take two things that don't normally go together, artificial intelligence and the physical world, mash them up and suddenly you have a great marketing hook.

But what if, Just what if that hook is actually as good as advertised?

Welcome to Confessions of Supply Chain Executives, the podcast where we get brutally honest about the challenges, failures and celebrate the victories in retail supply chains.

I am your host, Chris Walton.

Most retailers won't admit it, but few, if any, know where their inventory actually is at all times.

I don't mean in a philosophical sense either.

I mean in a literal operational sense.

They scan inventory at the docks.

Some, not all, might scan it as it leaves the back room and gets placed on shelves.

And finally they scan it at the register.

And everything in between.

Well, quite honestly, it's anyone's guess.

I was first introduced to the concept of physical AI by Wiliot's Amir Kojniati on this very podcast series back in February.

So I've asked him back.

But this time we've also invited K. Irwin, the national practice lead for Connected Assets and smart Supply chain practice at AT&T to the discussion because the two companies just formed a unique relationship to help retailers make the concept of physical AI work AT scale across retail.

Amir and K, welcome to Confessions of a Supply Chain Executive.

Amir, how are you doing today?

Doing well.

Thanks again for having us.

Really appreciate it.

Yeah, it's great to have you back.

We had such a great discussion.

I'm really excited to get into this concept with you even more since you kind of teased it for us the first time.

But Kay, how are you doing as well?

I'm doing great, thank you.

And I appreciate being invited to the party.

Yeah.

So, K, our audience knows Amir a little bit.

Tell us about yourself.

What all does your role entail at at&t?

So I'm in our at&t connected Solutions group and I head up a team that handles anything location related.

So we help our customers track products, assets, shipments, providing location and sensor data on any mobile things they need to find and keep track of.

Who knew, who knew that AT&T extended into such areas.

That's great.

All right, well, thanks for joining us.

I'm excited to have this power duo with us today.

All right, so let's start at the top because I want to make sure every listener that listens to our podcast on a regular basis, from a first year analyst all the way up to a chief supply chain officer, I want them all to be on the same Page by what we mean when we say physical AI.

So physical AI, Amir, what does Wiliot and you particularly mean by that term?

And how is it different from the IoT and RFID conversations retailers have been having for the past few decades?

Yeah, great, great question to kick things off here.

So physical AI by definition is our ability to take the physical world, turn it digital, and then be able to make sense of it using AI and agentic type algorithms.

For us, historically, we've seen barcodes RFID tags in the market.

We've seen certain use cases where you know exactly where they are at a point in time, mainly what we call snapshots, and then the items go dark.

There's a lot of brownouts through the supply chain.

With physical AI, you're building on that concept of digitization, but you're able to trace and track the journey of every asset with no brownouts and then at the same time be able to get full visibility on the sensing and the state of that item.

So whether it's an item that's a dry item or if it's an item that's a perishable, you, you can still have the same use case behind it, but get the insights around the sensing as well.

So, Amir, I want to push you a little bit on that because, you know, why, why, why describe that as physical AI?

Why isn't it just sensing or IoT as we traditionally talked about it, prior to the advent of all the discussions around AI?

What is the AI part actually doing in this, this framework, so to speak?

Well, it's getting better and better by the number of data points that are ingested in.

Historically, data points live all around us.

You have data points from barcodes, you have data points from RFID tags.

The AI component of it mixed in with the sensing capability allows you to be predictive, not reactive.

Historically, being able to take snapshots is a reactive state.

If you have somebody in a warehouse walking around with a RFID reader as a wand and they're manually scanning the left side and then the right side of the warehouse, there's manual intervention, there's.

It's error prone, they may get distracted and miss something.

That data doesn't come in.

With physical AI, specifically, drilling down on the AI component of it, you're able to get full visibility with the Williot stack today.

But also the AI portion with the data coming in starts to become predictive and start to look into trends.

So if there are certain levels of restocking stack stock to light when you look at it, or if there's certain levels of volatility in a transport and that same methodology is happening time after time.

You can predict it ahead of time and then save any kind of spoilage, any kind of issue, or from a supply chain standpoint.

And procurement, you can get ahead of any kind of stock out based on the trends in the data coming in.

So the AI component is that insight and the visibility to stay predictive versus the historic reactive system state.

Got it.

Amir, so if I were to summarize what you said, it's really about generating insights and possibly even actions based on the data that's being understood or acquired by whatever system you're deploying.

Absolutely.

Yep.

Okay.

Got it.

Okay, Got it.

Thank you for that.

All right, K, so I want to talk about this relationship that you.

That your two organizations, AT&T and WilIOT, have formed from AT&T's perspective.

What was it about Wiliot that made the relationship so compelling?

And why now?

And why physical AI specifically?

Why did you two enter into this?

Well, AT&T and Williot are working together to help businesses gain deeper operational visibility through physical AI.

And Williot brings this sensing technology that is doing wonders for many of our customers.

And AT&T helps deploy, scale it with connectivity, services and support.

So the combination is really working well for our joint customers.

And you know, this item level intelligence really provides businesses a clearer view of where their products are throughout the supply chain.

So we really value this relationship and it's working out really great.

Got it.

Kay, I'm curious too, because you said you're your job, role or your remit really is focused on location and understanding everything that happens and the data you can collect from around it.

When you look at where physical AI fits into AT&T's strategic portfolio, how do you describe, describe physical AI internally to get what you need and get the resources committed to you in the ways you need them to be?

We have physical AI helps connect what's happening in the real world to what companies can actually see and act on.

So.

So AT&T provides that visibility, that connectivity to make those decisions faster.

So it's very important to AT and T and to our customers to have that improved awareness in the fields.

Catch issues sooner, respond faster, and connectivity really allows that to happen, especially on these mobile assets that our customers are looking for and tracking.

It includes not just the location data, I guess, but also sensor type data, you know, other data that goes along with that physical AI layer.

Right.

And then when.

And then the powers when it gets combined with that location data.

Too right.

The sensing data, that's where the one plus one equals three.

I'm curious too, Is, is this relation, I mean, this is a retail podcast.

It's a retail focused podcast.

But is the relationship solely retail or is there a broader industrial play between the two of your companies as well?

Kay.

Well, retail is a natural fit, but I've seen other customers.

It's really any business that makes moves, stores, sells a lot of products, or manages even other assets and equipment in the field, they can all benefit from physical AI.

So retail is probably number one.

But we're seeing manufacturing, obviously logistics.

Some of those types of companies have a real interest and a need and an roi.

So that's great.

Okay, so, so, all right, so I imagine there's a lot of lot of executives listening to this podcast right now who some of them are probably a little skeptical.

And by that I mean there's been a lot of people that have heard or seen a lot of relationships announced in the media that, you know, never delivered.

And so I want to ask you about Mirror.

Let's go back to you.

What, what makes, what makes this relationship different from the relationships they've heard about in the past?

Well, when you look at the way that physical AI is getting traction in the market and specifically Will yet is stepping into really lead that initiative.

We want to really collaborate with the right types of organizations.

And AT&T and K's group are absolutely one of those.

So we take our cutting edge technology and as we take it into market, we have the tech, we need the expertise and definitely somebody that has the track record to take that tech and then amplify it into large organizations that are putting trust behind this technology.

I've went through historically at the start of the podcast, really around the technology ladder.

So we went through what barcodes meant, what RFID meant.

We really feel we're on the next frontier with Williot.

And there's a lot of enterprise level organizations that have started to take this from a deployment standpoint forward.

So taking that from a POC or pilot to a deployment means a lot of trust in technology.

And you need a reliable partner to collaborate with to actually take that to market in a consistent way that it's not running into any challenges, especially at the scale that we're moving at.

And that's where this collaboration is so important and it's working out great.

Kay, what has to be true, say in three years time for this relationship to be considered a success in your mind, in the mind of A, AT.

And T, I would Say that in, in three years success means our customers are getting real operational value from physical AI and at scale.

As Amir said, you know, if we can, if we're moving from pilot phase and testing phase to that broader adoption and, and our customers are finding measurable business impact, that's.

I think we're right on track there.

Got it.

Okay, so that's really important.

That's why I started off the questions I did, everyone who's listening, because I want to understand the relationship within the context of what we're talking about, the timeframes with which we're talking about them, which, Kay, if I heard you right, you just said we want to see actual physical AI proving itself in the market within a three year time frame.

Is that right?

That is definitely true, yes.

Okay, so then my next question for you then is this supply chain visibility problem that I talked about at the outset, it's not new.

You know, retailers have been trying to solve it, I mean, as long as I've been in retail, I mean, I've been in retail for 30 years.

And most retailers can tell you where their inventory is in the system, but not where it actually is.

So let's get into, let's get into the nitty gritty now and why that gap exists and what it ultimately costs.

So Amir, why do supply chains still operate with limited real time visibility despite massive investments in technology over these past few decades?

Well, I think the attempt and the ambition is still there from what was set out to be done in the early stages.

The difference is where does the previous or the legacy technology stack have limitations?

And then really how does that match up to the expectations?

If you look at legacy technologies around vision or you look at rfid, they are very good at pinpointing those scans in real time.

But if the ambition is to have full traceability through the supply chain, we know that there's limitations behind it.

There's infrastructure limitations when it comes to the CAPEX required and the hardware required to have full visibility at all times.

Outfitting a trailer that moves from a manufacturing site to a DC or from a DC to a, to a back of store, if there's no outfitting or no infrastructure to drive, track and trace through that process, you have a brownout spot.

So automatically you're at a disadvantage to, to leading up to that objective.

If you're looking at some level of sensing, understanding temperature, volatility of a perishable through the supply chain, you're also at a disadvantage because you need to calibrate readers differently.

You have to change the tag composition from the IC makeup.

So you have all these variables that have to go in perfectly to achieve that objective.

Now, as we've set forth on this journey, we've understood some of those limitations and our hope was as we start to become a puzzle piece in this equation, we're addressing the challenges and shortcomings from the legacy tech so that we can not only build on the foundation there, but also we can really focus in on the areas that we see the true problems.

So that handshake from a source to a distributor, that that's a handshake that should be automated and there should be visibility through that process because it's so crucial.

Equally, when it goes from a DC to a store, that handshake needs to be streamlined and we need visibility around that.

If any one of these nodes is broken down, you have a discrepancy in the supply chain and then you have a, a lot of questions and finger pointing.

So it sounds to me like the main gaps or the main problems that still exist in trying to answer that question, I said is it really comes down to the cost and the actual infrastructure being able to be implemented in a way that is ROI efficient, for lack of a better way to put it.

So what is the breakthrough that then comes, it enables you to then take this to that next level.

Is it that the increase in technology and continuous sensing that you've been talking about, what is it specifically?

So we have to start on actually the value that we're selling through and gear it kind of backwards to the problem and the cost and the dollars that are going into to the problem.

We match that up with the value of the tech and then we start to build the ROI storyline behind that.

What we've seen is we've broken it down basically into five categories and they go hand in hand.

Two of them are basically hugging.

I wouldn't even say they're.

It's a handshake.

It's.

It's the ability to understand when a shipment is leaving and a shipment is received.

So those are hand in hand.

It's your ability to automate and count within a zone or a set location what we call automated cycle counts, inventory counts.

It's your ability to have visibility on any kind of reusable asset that's in your facility.

So think about plastic crates, roll cages that transport different cases, cartons, packages within a facility or between facilities, tracking the assets specifically that have additional smaller assets within them.

And then the last component is that the condition monitoring freshness topic, which is really around temperature monitoring today.

And of course we can do humidity, get exposure to light, all that.

But temperature is the core foundational point.

So from these five solution sets we are starting to see recurring problem patterns that have costs associated to them.

And then from there we can back into one of the solution sets and then start to drive an ROI behind the investment that the organizations are making and then the time that they're going to realize the ROI behind that investment.

Got it.

So it's about having the framework and something you've got to fight the framework of where the operational issues are and where they can best be attacked.

So, so Kay then, from a connectivity standpoint, why is physical world sensing at scale so hard and are there infrastructure requirements that people are, that people have traditionally underestimated, that there's now a new punchline to the joke, so to speak?

Well, I, from experience I know that retailers IT teams do not typically have excess resources.

I came from retail and it was, yeah, so it's, if it's something that you're, the company is relying on for your business operations and ultimately for sales, knowing where your products are, it just, it takes more than just connecting devices, although that's very important.

The reliable coverage, the secure data transportation, the tools to manage those devices, networks and support are very important and they really can facilitate the deployment at scale.

If you don't have those tools and support available, it makes it very difficult to work consistently across real world environments in retail.

Yeah, that's a great point.

I mean people, I think people that haven't been in retail underestimate the complex complexity that's involved and just how many different touch points there are between so many different constituencies in the process, let alone the customers that are involved in the process too.

So Amir, you mentioned the five, you mentioned the five buckets.

Then can, are we at a point where we could put a dollar figure on the problem?

Like maybe even, you know, look at some of those individually.

Like what does, let's start at a high level first.

What does poor supply chain visibility actually cost a mid to large scale retailer annually?

Well, I think you have, you have some variability here because you look at the types of items that each retailer might have.

Even if you break it down to like a grocer that's working with handful of different sources of how they're getting their product, vegetables or have a different type of loss on a truck than a, than steaks do, for example.

We've, we've looked at, at high level averages and something in the sorts of mid 200 millions is a reality in loss.

When, when you look at items that potentially are misrouted, a byproduct of that misrouting is items that then go through volatility with temperature and spoil.

The other factor that we've seen through this equation is dwell time.

So let's take that same example.

Whatever that item might be, and it ends up at a dock door and it doesn't go to refrigeration or a freezer in a point in time.

We're averaging numbers in the mid 30 minute ranges.

You are now exposed to that dwell time issue.

And then ultimately it's spoilage.

So it's a pretty significant number when you just average it out.

And these are not just specifically on the higher end of products.

If you take it to the higher value end, with let's say the stakes of the world focused on those, you will start to see even higher numbers.

So for us to start with that type of nine figure number and then work backwards with a solution that has minimal infrastructure costs, you start to see the ROI much quicker than something that you're getting into that requires heavy levels of infrastructure investment.

And then maybe even a shortcoming on the technology stack.

Let's say you can't do the sensing or you're dependent on some level of cameras to triangulate where items are within a facility.

You don't get the same outcome with the full technology stack that we offer.

Right.

Is the bulk of it then waste and spoilage, is that where most of the opportunity comes from?

In the examples you just shared, when.

It comes to freshness, we see that as one of the primary drivers when it comes to any kind of dry item, which also then translates to freshness.

It's a lot around misrouting.

So you might have an item that comes from one facility, you still need the assurance that it's getting loaded on the right truck the right time.

And if it isn't, you're efficiently taking it off and putting it through the right dock door.

Many of these facilities, they have anywhere in the ballpark of 50 to 100 dock doors on each side of the building.

And when you look at the type of volumes that they're pushing out of these facilities, you want to get it right.

And if somehow it's not right, you want to catch it right away, pull it off, and then move it to the right location that those are the main drivers.

I would say from a, from a cost perspective.

Got it.

Okay, so then, so let's talk about the before and after then like so let's say, let's say I'm bought into this and I, you know, I've got continuous physical AI visibility into my supply chain to avoid those types of issues that you just described.

How does my day to day operation as a retailer overseeing my inventory actually change?

Well, it becomes more efficient and then back on the theme that we started with, it becomes predictable and repeatable in a consistent way.

We have a customer right now that we're working very close with with the collaboration with AT&T.

They, they do a lot of asset tracking on their roll cages.

So one of, one of the use cases that I went through in that example, you want to know where the assets are at all times, but also you want to be predictive internally to know when those roll cages are coming back to the facility so that you're efficient in your operation.

If those roll cages are lost in the field, if they don't make it back on the truck and return back to the manufacturing site, you're going to have a disruption in your day to day operations within the facility.

And what physical AI is creating now on the reusable side is the ability to then predict and say you have 200 roll cages out in the market.

Their planned to come back to the facility within three working days.

If they're not back at the facility, we're gonna have this level of downtime because we have to push around this many products into this many trucks and then repeat the process.

So that level of visibility, it not only helps with the supply chain overall, but it helps with day to day business operation.

And any operator within the facility is gonna be dependent on that information so that in an accurate way they can plan their day.

And how the facility operates, it also.

Changes just how you're managing the business too.

Right.

It takes you from a react because what you're describing is 100% a reactive state most of the time.

Whereas now you can actually become proactive to identify the problems in advance.

Absolutely.

And that's where the value of the data ingestion comes in, coupled with the AI component.

So you can take on more data and then you can start to predict these things before they become problematic.

Okay, so that, that makes sense to me.

But you know, but by the same token, I've heard the phrase continuous visibility, you know, for years.

Like I've heard it, it's, I mean at least I've probably been talking about it for 10 years and it's always great in a PowerPoint deck.

But the proof is in the pudding.

When you start getting down to specific use cases as we've started to get into.

So I want to go.

I want to keep going deeper on this, Amir, if you'll.

If you'll indulge me.

So let's start with inventory accuracy.

Let's get really deep into inventory accuracy.

How does physical AI change what a retailer knows about their inventory and what they can do with that information that they couldn't do you before?

So there's.

There's a couple aspects to this.

So when you look at inventory counts and the visibility behind it, one within.

Let's start with the example in the facility.

Then we can move to the example of something in transit, because both of them are equally as important.

And then one, one of them, the latter didn't even exist.

When you look at what it takes to scale up and support that from a inventory perspective within a facility, with the read ranges that we're getting right now, which are about three times what you get with standardized rfid and with barcodes, it's much less because you have to have the actual line of sight to read the barcode.

You're getting longer read ranges to outfit a zone you're looking at anywhere in a ballpark.

You can go from a very simple device into the mid-50s, maybe less, into a more complex device that has all the network components and everything built in, maybe for a few hundred dollars.

So you couple all that in and compare it to where it is, almost 10 to 15 times the cost.

With RFID, you can outfit these devices all over a facility.

You're getting three times the read range, and you're getting full visibility within the facility when an item moves from the left side to the right side with no manual intervention.

So the visibility within the facility is something that is realistic, it's tangible, and it's completely reduced from error because there's no manual intervention behind it.

Now you take that same inventory that now you can pinpoint within anywhere in the facility.

When it starts to move from the back of the facility to a staging site to a doctor, you have visibility on where that item is staged when it enters a truck and leaves that dock door.

And then being able to flag it in real time if it goes to the wrong truck and then move it back and then put it in the right truck.

So you're not losing any inventory within your facility.

And you're also catching things predictively before they go in transit.

And then if you take that same example with the very limited lift on the infrastructure and you place it on a truck, now you're able While an item is in transit to be able to get visibility on where the item is at all times.

You can couple that with the telematics of many of the organizations that are out there.

And then with a reliable network, you can get all the information, real time, of where the assets are and what state they're in.

So it's more than just this asset is on this truck.

It's how are all of these assets doing on the truck and how is the condition of all of them through this process, and then equally when they're received in an automated format, you know when it's taken off the truck and you know when it's actually within the facility.

So it is building on a foundation of historic inventory visibility.

It went through this door.

It went through this gate, and I know it went through there.

And now you're getting full visibility throughout the process of not just one point in time where it entered or left a door.

Wow, that was really well said and really well articulated.

It actually makes complete sense to me.

And it's kind of like one of those, like, yeah, why wouldn't you do this?

Why wouldn't you want to try to get that type of visibility that you just described?

All right, so you've mentioned it a couple times, and I want to click into this one, too.

Freshness and food safety, you know, we talked about already, like waste spoilage, huge opportunity.

Walk me through what, what the technology the latest technologies are doing to help improve temperature monitoring and freshness tracking.

So, so a lot of this is built.

These capabilities, I want to preface it, they're built around the standards and the compliance points that are already out in the market.

So when you look at right now what the FDA is putting out with Fisma, the direction it's going over the next 12 to 18 months, there's stringent requirements.

Not only that they're putting in, but all of the food processors, the grocers, the QSRs, are really putting as a baseline foundation of everything that needs to be put in place.

So as we've looked at the technology stack and you look at the capabilities around temperature, around humidity, we, we want to make sure we're abiding by that direction, so we're following the standards.

And at the same time, when you're able to get visibility and information on the asset and you go into the platform, we're hitting all the check marks that, that you would be looking at from a compliance standpoint on anything that's a perishable.

So tying it now back to the capabilities, everything that we went through on the inventory side, you can absolutely do with the sensing capability as well.

The question is, what are the items and the requirements behind them?

If it's a dry item, if it's a package of, let's say, any kind of retail item, shoes, garments, different things, you're not going to need the condition they're in unless something drastic is happening and there's liquid involved and humidity and then there could be mold.

So maybe you want some metrics of the temperature and the state of that, that truck and the position it's been angled in.

But if it's not in that state and it's just a dry item, the inventory use case and the visibility through the process is enough.

If you get into any kind of perishable, you want to couple the inventory use case with the temperature and humidity, the condition that it's in.

And that's what we're really basing the, the freshness topic around.

Right.

And the other point, I gotta think the other important part of this too, Amir, is like you actually have a realistic record of what's happening right now because you know, and you can trace it back too, which is important with all the regulations coming.

Like right now, you just kind retailers and grocers only know what they know in a lot of ways.

And that's not always that much information when it comes to understanding how an item has been kept in temp throughout the entire supply chain process.

Yeah, we want to be a source of truth in everything that we do.

So that's the value that and the value of this collaboration is that we have a reliable partnership.

So when we look at how data is transmitted, there's no downtime.

We're working with a very credible partner in this ecosystem.

But then on the other side of it is that when you do get that data in a reliable format, that ledger with all of the different records, product provenance, where it came from, all the different nodes as it went through the process, and then being able to just triangulate that to a single source, that gives you all the reporting that that's so valuable.

And when you take it even one step further, there's some other mandates right now that are in play over the next about five years called digital product Passport out of the eu.

You have to be able to trace an item all the way back to the source.

You can't just stop, you know, from the last location that it's been in.

So whether it is a perishable or it's just our example of building on where the foundation of just standard inventory is that doesn't require any kind of freshness tracking.

You still need the product provenance and you gotta prove that it's going through the process.

So we're, we're proud to say that being able to get that level of visibility, you could be the, the using that as your source of truth around the compliance side.

Yeah, that's a great, that's a great point.

There's a whole lot of ancillary benefits here to this approach.

And K Kay's been waiting patiently.

I want to get her back in here.

But, but Amir, I do want to ask you one more question before we go to K, because I think this is important too.

And it's a, it's a, it's a topic that you actually haven't brought up yet really overtly and that is shrink.

Either shrink from loss or shrink from theft.

Like does physical AI, as you've been describing it, do anything to help tackle those issues?

And if so, how so?

Absolutely.

So, so I would start on the topic before we go to the back of store and, and those, those use cases that, that are maybe a little bit more controlled.

The things that are really out of control.

Many times is when you're going through the process and you're working with a 3 PL or you're switching too many hands with an item, leaving a source and it's, it's on its way in its journey over days, weeks, whatever it might be to, to an end destination.

There's a lot of gray market diversion that, that happens through these processes.

So I would say the first use case really around shrink or any kind of theft is really having the visibility that every time there's, what I'm calling these handshakes from one node to another is is it going to the source it's supposed to be going to?

Can we be the stamp of approval showing that that handoff was made in an automated format and then when it leaves that node and goes to the next one, does it follow that sequential process?

And if anything breaks down, being able to flag it, send the event trigger and say that something went wrong here.

The number of days it sat idle, the number of times it moved around, they don't match up to what the expectation was.

One, one step further is really our light detection capability.

So we could put these tags in boxes, in packages, if those are ever opened through the product's life cycle, through its journey, you could flag that and say there's been tampering, something is wrong there through the process.

So that's kind of step one and then step two to address the shrink and the theft topic is maybe it went through its journey, but something did not go right at the back of store level.

And if something was, for example, on a shelf and it's a carton and it's planned to be sold and somehow it's opened before it's actually sold and taken out, that event trigger showing that that package maybe was opened in the facility is another trigger that something has gone wrong through the process.

Because in theory, that event should never have happened unless a customer took it out of the store and you're recognizing it leaving the store.

So there's a couple use cases there, but I would say there's more weight right now on the supply chain aspect of it being able to track the item through the process and make sure that nothing goes wrong through its journey than when it actually gets to the back of store.

Because that's the last node.

And in many times, if the store operation has processes in place, all the hard part is done.

That's just basically a point of sale discussion to just position the product and push it out.

Yeah, yeah, right.

I'm thinking of all those horror stories I hear about too, of entire trucks going missing, you know, as well, because that's a, that's a big opportunity and a big loss for the retailers too.

All right, kay, so from AT&T side, the other question that's pop into my head is what role does the network, quote unquote, play in enabling all the various use cases that Amir just described?

Because to me, a sensor that's, that's, that's not connected to anything doesn't help anyone.

But am I oversimplifying that?

Well, no, I, I agree with you.

Because back in the day, we waited months to get reports and information and try to make observations well after the fact.

Right.

And then try to figure out what had happened, how we handled that.

Now, the network sends the data faster than ever from wherever, anywhere.

And that data becomes much more useful and timely for retailers to be able to evaluate and make decisions quickly.

So connecting those sensors and pixels across locations so businesses can see what is happening throughout their operations, they're able to act on it faster with that information right in front of them and resolve most of those issues that Amir was just going through.

Being able to know right away and act on it quickly.

Yeah, and that's the Venn diagram here that we're talking about, or the intersection that makes the unique relationship between the two of you so compelling.

Because I think Oftentimes, most of us in the pundit space take for granted the actual infrastructure that needs to exist in and around these technologies, to deploy them, to connect them, to monitor them, to maintain the sensors at an enterprise scale level.

That gets really, really complicated.

So, Kay, I want you to break that all down for me.

Like what is AT and T's specific contribution to making all of this work operational?

Well, we help make it work in the real world with the connectivity and the deployment for that ongoing support for customers at scale.

The tools that are required to monitor the data, the devices remotely and manage them at scale.

We also have that as a managed service.

So, you know, we take that, that support and that monitoring off of the customers and be able to proactively say there may be an issue right here.

So we're really helping turn this promising technology into something a business can actually run day to day.

So Kate, what does that actually look like?

So in practice, like if, say I want to say I'm buying into what Amir described and I want to deploy physical AI infrastructure in my retail environment, what does that look like?

How long does it actually take?

What can go wrong?

Who's doing the work ultimately?

Well, AT&T does provide really the end to end services for deploying the solution for customers.

So we do have field service personnel nationwide and we have a number of international partners as well.

We're installing some internationally for.

So that piece of it, you know, we have, we test it, we deploy it, we validate it and know that it's installed and working properly.

So the duration of that really depends on the design.

It depends on the size of facilities, the quantity of facilities, the area covered.

Are you just looking at that, you know, receiving and shipping accuracy?

Do you want inventory everywhere in your store or your warehouse?

So it depends on the use case as well.

But I think one of the really important things is that validation and fine tuning to ensure that, I mean, that's where the pitfalls can happen if we can have that service upfront to know we are meeting the customer's APIs, we are achieving their business goals and their KPIs from the actual installation of the infrastructure and the entire solution.

So, okay, that kind of takes me to my next question.

So it sounds like AT&T is responsible for the actual overseeing the entire deployment of the situations like this or technology concepts like this.

What happens post deployment though too?

Is AT&T still involved in that?

Are you still involved in the monitoring and any maintenance work that needs to come into play as you continue to improve this and hone it and make sure it continues to work at scale.

Yes, we are actually.

We are.

After deployment, we're keeping things, you know, monitored, running smoothly.

We're helping customers provide reports on that, monitoring performance.

We do dispatch, you know, trouble tickets if there's ever any hardware or, you know, cabling or installation failure or network.

So we are able to stay ahead of that as our customers, you know, grow and scale.

Got it.

Got it.

Well, that gives me a lot of confidence then, with the two of you guys working hand in hand to make this, make this work.

So, Amir, I'm curious too.

From Juliet's perspective, What does having AT&T in the ecosystem unlock that you all just couldn't do on your own?

Well, there's a handful of things, so I would say I would generalize a couple different categories.

One is that their ability and their reliability on the data side is so, so important, critical.

Especially when you look at all things we talk around, around physical AI.

The foundation is being able to take the data from that physical product and put it in a platform and be able to then run the agentic algorithms behind it.

We can't even get the foundational piece of that going without reliability on the data side.

The second piece that I would generalize that's equally as important is that you can have great tech.

That tech could work.

If you don't have a go to market to actually get into the field and deploy it at scale in a consistent way without any breakdowns, you don't have success.

So those are foundationally the two pieces that I would say this collaboration is so instrumental for the success and the work that we're doing right now.

Yeah, this is really great stuff because this is where the rubber meets the road.

When you start talking about technology that's fun to talk about versus technology that's scalable and deployable.

All right, so let's talk about the commercial model then.

So, you know, we've got people bought in.

We've talked about the use cases, we've talked about the value each partner is bringing to the table.

If I'm a retailer, how do I actually buy this?

You know, is it hardware plus connectivity plus software?

Like walk me through, like how this all works.

So it does have all of those components and we can actually at and often rolls that all into a managed solution for our customers.

So depending on if they want that capex model or an OPEX model, we can do that for the entire solution.

Or we can sell hardware upfront.

You know, network is obviously opex Typically services and support are, but we do have an arm within AT&T called Capital Leasing.

So really anything can be rolled into that operational model, which is, is nice for a lot of retailers.

So the other, the other angle you got to think about is the integr.

There's also a lot of integration points when you start talking about a technology like this.

I mean I could think about, you know, the, the acronym SOUP comes to mind for me.

Like the wms, the oms, the erp, K. What is, what is the, what is the integration side of that looks like and how does that all get handled?

Well, definitely with all of this automation, it's not going to work unless we have that integration to the customer system.

So the goal is to fit into their systems that they use already, not make them start over, but really works best when the data flows from the platform.

They're already running inventory, fulfillment warehouse operations into the Wiliot platform, into the assignment of pixels per product.

All of that automation is key and that means integration into their systems.

So we perform that service for many of our customers and it's really critical.

Otherwise you're adding time, saving time at the same time as adding time.

But integration obviously allows that automation and free flow of data back and forth.

Ideally you're not ripping out and replacing anything.

You're just working alongside or in connection with, in concert with what already exists.

Ideally that's correct.

We are not ripping and replacing, we are just adding visibility and adding that integration to make it all seamless for the customer.

Got it.

Amir, who owns this on the retailer side from your experience, like you know, is it, it the supply chain, some combination of both.

But like where does the buck actually stop inside the retail organization where you've seen this deployed successfully?

It's a great question.

So it's, it's a combination of all.

So I think it's an all hands effort and that's what is, is valuable.

But also it, it slows down some of the sales cycles when we look at the different stakeholders that have to be in play.

When you look at discussion points, anything around supply chain.

So you have all of the warehouse managers, you have the, the folks, the executives on the supply chain side that are, throughout the process they're identifying the dollars out and they're trying to figure out a tool to get, get the right ROI behind their problems.

However, anytime you have any kind of platform integrated with their systems, you have it involved.

So we'll, we'll start usually either at the CIO VP of IT level and then it Actually then goes down to the tactical levels within their teams.

What we see is that typically the origination of any of these projects, it does fit in somewhere between supply chain merchandising, the folks that really realize the day to day pains.

And then very quickly within our first few meetings, you have someone from the IT side already shouldered into the discussion so that they're talking around the compliance and what that all means.

One, with the legacy investments they've already made because you have to draw lines and integrate within those.

And then two, what this means for future roadmap and how these systems are going to play together in a way that doesn't break down investments they've already made.

This has been a really, really interesting discussion.

We've gotten into a lot of nitty gritty on the factors at play here.

So I want to, I want to take the last time we have with you, say the last five, six minutes of this podcast and I want to get tactical for our audience.

I want to give them things that they can take away and actually go back into their organizations and potentially do so.

Amir, I want to start with you.

I want you to give us the 30 day action plan.

If I'm a chief supply chain off, let's say I'm a chief supply chain officer, I'm listening to this episode.

What are the specific action steps I should take within the next 30 days if I'm buying into what I've been hearing?

Well, there a couple categories.

One is if you've already made an investment in legacy technologies, but you have any of the shortcomings that we've went through during this session.

So summarizing it here around visibility of an asset leaving being received in an automated format, visibility or assets within your four walls, any kind of misuse, misunderstanding, loss of visibility of your own assets that sit within so RPCs, rolling cages, or if you have requirements around temperature or any kind of freshness, visibility on humidity, you want exposure of items that maybe are getting open prematurely, they need light detection capabilities.

If you have any investment in technologies that you've made historically and it hits any one of those and there's shortcomings, try to understand what are those shortcomings.

And if you can actually quantify it without our help, that's a very good first step to get the discussion going.

Then the second part of that is if you can't quantify it, that's something we can get involved.

We can evaluate the workflows and then start to build a model that makes sense and a charter to take K made a very good point that a lot of the technology, the hardware out there doesn't have to be net new.

It could be things that we integrate with.

So it could be devices that are already in place, like an RFID reader that can energize our tag.

And then we would look with our collaboration into a low cost bridge that could receive it via Bluetooth, low energy, and then transmit the data.

So we don't always have to start at ground level.

There might be investments and workflows that we can already build on based on limitations of legacy technology.

The other workflow really around here is that maybe there has been no investments on technology and traceability solutions.

Maybe there's barcodes, maybe not maybe, maybe there's nothing.

There's no barcodes, no rfid.

And you're starting net new.

That blueprint to still evaluate any one of those five solutions sets and starting to go through a methodology to quantify also is a good starting point for us to jump from.

So it really doesn't matter if you're on a journey of an investment you've made or an investment you haven't made.

Our problems are common through the supply chain.

So it's just about identifying it down to those, those, those simple five and then we can work it from there.

Yeah, that's a great point too.

Like there's no sense recreating the wheel.

Like if you, if you've got a problem, chances are somebody else has it.

And you all have probably helped a lot of people solve it at this point too.

So there's a lot to leverage in that regard.

All right, K, so keep me in this.

Let's keep in the same vein, like say I'm the chief supply chain officer.

What's the first conversation I have?

So Amir kind of talked about, like, how do I, how do I think about this strategically and what do I want to do?

But what's the first conversation I have?

Do I start with you?

Do I start with Williot?

Do I call my own internal IT team?

Like map the journey out for us?

Well, I think customers need to start with what is the business problem?

What is the problem I'm having?

We help customers do that quite a bit.

In fact, we provide workshops around, you know, ROI and prioritizing technology deployments, you know, from a higher level, strategic level, but then also drilling into what problem are they facing.

The logistics folks, the store folks, the, you know, executives.

What is it, you know, what is your goal for shrinkage?

You know, do you have a compliance issue?

You know, do you have to have that traceability and store that data and feed it to a government agency or your partners.

So what are those goals and those KPIs and bring us in early.

You know, we, as Amir mentioned and you did, the problems are common across retailers and any supply chain organization.

So we're happy to help with that.

But you know, bring us in to design and pilot and scale ultimately the solution for you.

So it sounds like they should call you first, Kay, I'm going to ask at the end, like who should people get in touch with?

But it sounds like it sounds like they should call you first, am I right?

It's a lot of my job honestly is talking with the customers about their problem and digging into, you know, what is the goal and how do we solve it.

And here are some options and here are pros and cons and that's great.

We love to be brought in early.

And I just want to add to Kay's point there, that's very important to note is because they're on the forefront of really taking it to scale.

So a common issue is too much work being done on the pre sales without a vision of what direction you're going at some scale.

So that that blueprint that they walk through the hand inand trusted advisor methodology is the right one because they're looking at it big picture.

And then for us to be brought in to then solve the challenge with the technology, I like to say is the easier part because we know it works.

We just need to then have the viability and the trust that it can be taken to scale.

And that's why that exercise is so important.

Yeah, you two are dropping nuggets here.

I always love when the nuggets get dropped in like minute 53 of the podcast to YouTube because.

But you're right, I mean like as an former executive, like having to phone a friend there that understands, you know, everything that's going on at a large scale across the industry, where you can leverage the capabilities like Kay, like you just described is hugely valuable.

And there, there just aren't that many friends you can phone in that regard in for a lot of these executives.

So it's great to hear.

All right, Amir, we don't call this podcast confessions of a supply of supply chain executives for nothing.

So my last question for you is what is the bigger hurdle to physical AI adoption?

Is it the state of technology or the organizational change that's required to rally around it?

And how long do you think it will be until we see everything that we've discussed here today as commonplace throughout the industry.

Just being very direct, I think it's the organizational side because technology is going to evolve.

We're going to continue to have iterations.

You know, we're going to have Gen4 coming out in a, you know, a couple years.

And so when you look at that, then it's going to have a successor based on that.

For us, it's not a technology limitation.

We know it works.

We know it works successfully.

The value is there.

It's more around working through organizations and the hurdles they have.

You hit a very good point earlier.

What groups are we talking to?

It's all of them.

So if it originates in IT and supply chain, we still have to get them to have their collaboration and their nodding heads to go forward with the project.

And that works across all industries.

It's agnostic.

Where I think it's going over the next foreseeable future.

I would say we're already there.

I wouldn't say right now that we're limited at all.

You've seen some big announcements with some major retailers not too long ago, late last year.

We have some really transformational projects that we're working on with this collaboration.

So it's not a matter of a when.

It's just more.

How do you dial up that volume?

So there's more implementations and more scale, not only with these clients that we have, but also with net new ones.

But I do see a transformation over the next 12 to 18 months that we're going to be hitting the retail, grocery, quick service, restaurants, the logistics space and the automotive space with this technology because those are the ones that are struggling with legacy technologies, with many of the shortcomings that we've outlined here.

Right, right.

That's a great point too.

And with, with, with more implementations comes more scale.

All right, so K, we kind of kind of joked about it before.

Thank you for that.

Amir.

We kind of joked about it before.

If people want to get in touch with you, what's the best way for them to do that?

Kayl, go to you first.

Well, AT&T has a big website.

You can search on that for any kind of solution we offer.

But please contact me on LinkedIn.

I'm happy to talk through use cases and try to find the biggest impact for this technology for your company.

Yeah, yeah, for sure.

And Amir, you're definitely on my speed dial for this topic.

So what's the best way for people to get in touch with you?

Same as K. I think we have a lot of assets, really good information on the Will website.

Our collaboration is also there so you can work between the organizations to get started.

But personally, LinkedIn also, I'm pretty active there, so send the message and we'll be happy to follow up.

Wonderful, wonderful.

Well, thank you both.

Thank you so much.

Amir Kojnati of Wiliot and kay Irwin of AT&T.

This has been an incredibly valuable discussion.

We got to levels deeper here than I think I ever have on any podcast.

Thank you for educating our audience on what it takes to succeed with physical AI in commerce.

And of course, today's podcast has been produced with the help of Ella Sirjord.

I am Chris Walton.

This has been Confessions of a Supply Chain Executive.

Never forget Omnitalk fans.

Confessions are almost always good for the soul.

Be careful out there.