AI-Powered Retail Intelligence: FarsightIQ

FarsightIQ is an AI-powered retail intelligence platform that helps retailers predict demand, optimize inventory, uncover actionable insights, and make faster, data-driven decisions. Its suite of predictive analytics, machine learning, and computer vision solutions helps improve forecast accuracy, boost sales and margins, reduce carrying costs, and enhance the customer experience.

In this episode, we speak with Scott Pearson, VP of Sales and Marketing at Jesta, about how the company is using AI to help retailers move from hindsight to foresight and then into action. The conversation focuses on FarsightIQ, Jesta’s new AI division, and how it’s designed to help retailers deal with inventory distortion, stockouts, overstocks, and the complexity of modern retail planning. Scott explains that the platform is built as a decision-intelligence layer, not just a reporting tool, so it can observe demand, anticipate outcomes, recommend actions, and still keep a human in the loop.

The episode also explores how FarsightIQ is organized into several modules, including demand forecasting, product attribution, replenishment, inventory balancing, risk detection, natural-language advising, and operational matching. Scott explains why traditional analytics often fall short, especially when retailers are dealing with fragmented data, changing channels, promotions, local events, and seasonal differences. He emphasizes that the system is designed to work with both constrained and unconstrained demand, helping retailers understand not just what sold, but what could have sold if inventory had been available.
A major theme of the conversation is that AI is most effective when it supports, rather than replaces, human judgment. Scott stresses that merchants and planners still need oversight because they understand context that AI may miss, such as local events, vendor issues, or visual changes in-store. He also explains that successful AI adoption depends on clean, focused, and well-structured data, and that retailers should start with the highest-value use cases instead of trying to solve everything at once. The episode closes with Scott’s advice to focus less on systems themselves and more on the decisions those systems are meant to improve.

Follow SourceForge:

SourceForge.net - https://sourceforge.net
SourceForge LinkedIn
SourceForge X (Twitter)
SourceForge Facebook

Interested in appearing on the SourceForge Podcast? Contact us here.

The SourceForge Podcast is the world's largest B2B software podcast.

Transcript

Copy Transcript Copied

SPEAKER_02 0:05

Hello everyone and welcome back to the SourceForge podcast. I'm your host, Bo Hamilton, and today we are getting into a problem that every retailer knows in their gut, having the wrong stuff in the wrong place at the wrong time. Empty shelves when customers are ready to buy and markdown racks full of things that nobody wanted. The industry calls it inventory distortion, and the analysts at IHL Group put the global price tag at $1.73 trillion a year. That's trillion with a T. And for decades, the answer has been forecasting tools and dashboards that tell you what happened and maybe what's coming. But uh the company we're talking to today thinks the next step is AI that doesn't just predict what's next. It actually helps you act on it. Uh Jesta has been building retail and supply chain technology out of Montreal for over 55 years. And back in December, they launched a whole new AI division called Farsight IQ. It's machine learning plus AI agents that can spot a coming stock out and help your team fix it before it happens, with of course a human still in the loop. My guest is Scott Pearson, VP of Sales and Marketing at Jesta. And what I like about Scott is he's he's not just a software guy. He spent years on the retail side before crossing over. And he's been a VP, a COO, and a CEO in retail tech over a 30-year career. So he's he's lived these problems from both sides of the counter. So, Scott, I'm really happy to have you here. Um, welcome to the Stories Forge podcast. I'm excited to chat with you.

SPEAKER_00 1:28

Thank you very much for having me. Looking forward to the conversation.

SPEAKER_02 1:32

Now, I want to start with the the basics here because this is a brand new division. Farsight IQ just launched back in December, I believe. Uh, for listeners hearing about it for the first time, can you give us the elevator pitch? What does Farsight IQ actually do?

SPEAKER_01 1:47

Absolutely. So Farsight IQ was spun off as a separate division. We'll get into kind of the why, I think, in a little bit, but it wasn't really it wasn't really started in the the core context of it. The challenge is that retailers are short on data. In fact, we recognize that retailers are often overwhelmed by data. Uh data fragmentation, to your earlier point, fragmentation, sorry, across channels is one of the most common uh issues that we have. And then teams have a tendency to kind of plan in silos. So there's different teams dedicated to different areas. There's no really kind of oversight and overview of the entire area across the entire organization. So the core premise really was simple. Retailers don't fail because they lack systems. They struggle because decisions tend to lag reality. So Farsight's really just an it's not a reporting layer. It's a decision intelligence layer that observes demand, uh, anticipates outcomes, um, recommends actions, and then allows your team to learn and to imply or apply their human judgment as well. To your earlier point about human in the loop, I think that's key. So people need to feel comfortable that the recommendations are going to make sense and drive their business appropriately. As we looked at the problem, we actually came up with six different modules. Uh, and I'll touch on them really quickly just to get into the high level. At the highest level is forecast IQ. Forecast IQ is a machine learning uh tool that is dedicated to just predicting the future, uh, demand forecasts specifically, across both short-term and long-term, preseason horizons. It's unique because it allows you to split it into seasonal goods and non-seasonal, more staple items, and it uses what we refer to as an ensemble machine learning model, which is, I guess you could think, mixture of experts. So the various algorithms come up with their own unique positioning and forecast, and then it's a blending of all of those together based on the specific seasons, products, and categories. Then we said, okay, there's other areas that I need to, in order to be effective in forecasting demand, what happens if I have a new item that I've never carried before? Well, the only way for me to really accurately forecast that demand might be to come up with similar items. So that was a challenge in and of itself. So we looked at building out style IQ. Style IQ is an AI version. You upload an image, you say, please attribute this product for me based on the attributes associated with this product category. Style IQ is very robust in that ability to extract all of this information so that I'm now looking at items that have been that are very similar to previous items, so I can start to be more accurate in my forecast demand. Replenish IQ, the third module, is around how do I determine when to best replenish an item once it's selling in the store. And rather than just looking at min-max and things like that, which are very effective, uh, when I sell out of a product in one store, I want to make sure I'm replenishing that same product. I want to start to really cater my replenishment strategy more around the likely demand so that I'm not overstocking or understocking based on the demand in that particular location. Which brings me then to optimize IQ, which is now that I have inventory in stores, I want to make sure that I'm balancing my inventory appropriately. Optimize IQ is designed to do just that, to identify are there items in the warehouse that can move to stores or items in one slow-moving store that I can move to other locations? What makes sense from a business? And we're going to recommend ways for you to balance your stock accordingly to really optimize based on, again, that forecast. Risk IQ is another kind of overlay on top of that. That's where AI and ML really come together. And AI in this context, I mean generative AI or LLMs. So, how do I take those demands and I start to really look for outliers or risks? So we have a number of dashboards and AI-built skill sets that will start to drive. Here's some recommendations. If these are overstocked, these are understocked. This is an interesting trend going on in these locations that you need to be aware of and keep your eye on. So it's really a recommendation tool around how do you avoid risk. And then ultimately, there's something else we call advisor IQ, which is more your more traditional, I guess, chat bot. If you want, you can ask it real language questions, natural language questions. It will come back with recommendations, reports, scenarios. You can explore what can I do to increase my margins in this category, or where is my highest risk and following. It's just an interpretive tool that allows you to really ask any question. And we have a couple other more operational tools, uh, match IQ being one of them, which does three-way matching. So again, we pulled all of those tools together to say what is it that retailers are really looking to try to accomplish in order to make sure they are avoiding those understocks and overstocks that we discussed in the first step of this conversation.

SPEAKER_02 6:43

Gotcha. Yeah, there's a lot of exciting uh capabilities there. Um thanks for giving us that overview. Um uh it it what kind of stands out to me is just, you know, it's the the act portion of the um the feature set. It's like it's not uh just sort of another tool that's telling you you know what already happened. It's sort of looking ahead and actually helping you do something about it through all these different um uh features um that are built into the platform. Um what's interesting to me, you know, is that Jesta isn't just some some startup kind of chasing the AI wave. You've been you've been building retail technology for over 55 years, um, much longer than a lot of these uh AI companies we we um hear about and chat with. Um so walk me through the the decision. Why spin up a dedicated AI division like right here and now?

SPEAKER_01 7:33

I think that's a great question. I think there's a few reasons. I guess I'd start with timing. You know, retailers are facing a level of volatility and complexity that traditional planning methods just weren't really designed to handle. I think everybody realizes that they need to be looking at how do I incorporate AI into my business model. Everybody else is doing it, and I can't be a laggard in this context. I need to stay ahead of the game. But where do I actually apply it? Where is it most important for my business? You know, demand channels are changing quickly. Um, channels behave differently online, handles behaves differently than stores do, marketplaces, all of these other things are going on. And the reality of it is inventory is expensive. That is your most expensive cost, is what you're buying in your inventory. So I need to really make sure that I'm highly efficient in how I manage all of that. So, and again, to my earlier point, there's probably more data than ever right now, but people still kind of struggle to turn that data into timely action. So, what can we do to help retailers make more timely decisions to make sure they're proactive rather than reactive to what's going on in their business?

SPEAKER_02 8:41

Now, um, I know, I know retail analytics isn't um particularly new, right? Forecasting tools aren't new. Retailers have been have been using dashboards and like demand planning software for for decades now. Um, from what I've read, the pitch with Farsight IQ is that it doesn't just like predicts predict what's next. It helps you act on it, right? So what's actually different here compared to some of the more traditional tools that have been used in in years past?

SPEAKER_01 9:09

Good question. Yeah, I think traditional analytics usually tells you what happened. Uh, there are forecasting and demand tools that have been around for a long time, and they give you an idea of what's likely to happen, but then you're always still laggard. You're falling behind, you're looking at what happened. Traditional forecast ties to predict what might happen. Farsight IQ is different because it's designed really to move beyond both of those things and help retailers decide what to do next. So it's not about what's happening, what's likely to happen, and what did happen, but what do I do next? So it really is helping you to do that next decision. Retail is incredibly complex. You can't treat every product, every store, every channel, you know, every demand pattern the same way. Uh basic replenishment items behave differently than fashion items, as an example. A new product, as I mentioned earlier, behaves differently than an item that you've carried for years. And stores oftentimes change also, as does e-commerce, and regions can be changing all the time. Then you need to tie in the other levels of complexity, such as promotions, various inventory constraints that you may have in a location, weather events, local factors, all of these things can distort demand. So a lot of forecasting tools try to force all that complexity through one method, and that's where things start to break down. That's why I mentioned before we're using Farsight IQ and forecast IQ specifically to really look at an ensemble to say there's a number of considerations, and we need to balance what these experts are coming up with to find the right decision in the right location at the right time.

SPEAKER_02 10:47

Yeah, I mean, there's so many, there's so many moving pieces and and um you know areas to to watch. And it's it's really uh the problem that you're uh working to address with retailers is is um you know, it's it's trillions of dollars worth of of inventory and and um the numbers are pretty staggering. I mean, uh as I alluded to and mentioned in the in the intro, IHL Group estimates that inventory distortion, that's like out of stock and overstocks um combined, costs global retail $1.73 trillion a year, um, which is pretty staggering. So that just kind of visualizes the scale for listeners. Um you have empty shelves on one side, markdown racks on the other. How does better forecasting help retailers stop sort of bleedy money in both directions at once?

SPEAKER_01 11:36

That's a great question. I do I think that overstocks and stockouts look like opposite problems, but they're usually come from the same root issue, the same root cause. The retailer doesn't have a clear enough view of the true demand. If you underforecast, you run out of inventory, you lose sales, you disappoint customers. You might also create operational pressure as teams scramble to react. If you overforecast, you create excess inventory, you tie up working capital, your increase markdown exposure, you reduce margin, all these things kind of come into play together. Better forecasting helps because it creates a stronger demand signal. And that demand signal drives almost everything downstream. All the downstream decisions, how much to buy, where to allocate, where to replenish, when to replenish. So it really helps you to do more of that just-in-time inventory. So you could be much more intelligent about don't push everything out to the stores immediately, hold some back so that you can be more responsive to the needs of the market and make sure that you're you also have that sort of feedback coming in on a regular basis. Demand is much higher than we anticipated. Make sure that you're also taking your buying cycles and the length of time it takes for you to purchase new inventory from from your vendors into account as well, and really try to strive for just in time, but but also recognize that there is volatility and you need to be on top of this on a you know daily, weekly basis.

SPEAKER_02 12:58

They they almost look like opposite problems, like uh too much stuff, not enough stuff, but they're they're sort of the same problem if you think about like you guessed, you guessed wrong about what people would buy. And if you fix the guessing portion of the issue, then both sides sort of get better. And it sounds like that's kind of what you're you're working on with far site IQ, right?

SPEAKER_01 13:17

Yeah, and then I and I mentioned earlier the the context of uh you know the real demand. And I guess I'll draw one other distinction that most planning tools are generally triggered very much toward what is considered to be constrained demand. Constrained demand is demand based on the product inventory that you had available at the time. The question is, what about unconstrained demand? If you had more inventory, what would the sales have been in the past? And therefore, how do I incorporate the opportunity, the unconstrained demand, in my forecasting tool so that I can identify if everything was always available in the right channel at the right time, what might my demand actually be? And that's where fireside IQ comes in is we have the ability to look at the unconstrained demand and then constrain it as well. So you can take a picture of that, and this is where the human in the loop comes in. Now the human can say, do I believe this unconstrained demand is accurate? And if so, let me beef up my inventory a little bit or follow what the AI model is telling me. And if I want to, you know, ratchet it back a little bit, or maybe my open to buy doesn't quite meet those numbers, I'm gonna live within my OTB and I'm gonna actually maybe constrain that a little bit more, and we're able to really provide the insights into both of those concepts.

SPEAKER_02 14:31

Now, I want to see if you're be able to give us a um a concrete sort of example to illustrate what it is you guys are are solving for um customers and clients. Um, can you give us a real example of a retail decision, maybe something a buyer or a planner deals with every week that Farsight IQ could help improve or maybe even automate?

SPEAKER_01 14:52

I think a number of things. So I guess I'll focus first on the forecast. A lot of people say forecasting is the most important. And I would argue in some cases forecasting is the most important because everything else works downstream from that. So getting your demand forecast down first and foremost is key. So most retailers I'm speaking to today are saying, let's start with the forecast. Interestingly enough, I have a slightly different point of view that maybe the best place to focus on is what I called risk IQ earlier, which is let's analyze your existing data and your existing inventory, what's available in each of the locations today, and let's start to identify where there's likely to be risk. This product is selling particularly well in these eight locations, and therefore I need to move more inventory into those eight locations, or it's really moving very slowly in these others, and which products, categories, uh, types of products are not moving, let's be more responsive to that. So I think risk IQ provides you the real-time insight, and it's a daily AI model that's running against the likely demand for each location that's changing on a daily, weekly basis, but it's also looking at the actual inventory and stock. That's where I see that stock balancing and optimization might actually provide some retailers more benefit than the forecast itself. Not to say the forecast isn't important, it clearly is. Most people want to start with the forecast. Let's look for that demand, let's look for the areas that are hitting my business the most. In many cases, it's the fashion inventory because right now I'm basing my fashion inventory on past years, past three, four year sales. And I didn't carry this product before. So how do I link this product to fast uh past products to make sure that I'm accurately forecasting that demand? And they want to focus on a very targeted area of business, which is smart, by the way. I do uh suggest that people implement the products and AI just in general. Find what makes the most sense for your business. Is it the broader forecast because everything downstream is impacted by that? Or is it targeted risks that you want to focus on? And if it's those areas, in what part of your business do you want to focus those areas and risks on? And let's really tackle it that way. So that's really where our conversations usually go is let's figure out what makes the most sense for your specific business, identify those areas that you want to move the needle, and then start to implement. You have to you have to take that first step.

SPEAKER_02 17:25

Yeah, forecasting is tough. I mean, uh the world often doesn't like to cooperate with uh with one's forecast. Um uh and retail, especially, I know, loves loves uh changing things up on you. Um like I I just think of like the TikTok trends that maybe blow up and change things. Uh yeah, there's so many shifting like pieces that in a market that can um suddenly just throw your your carefully planned forecast out of whack. Um now I know I know far side IQ pulls an outside signals like like weather and local events. So, how does AI help retailers actually respond to those curveballs instead of just sort of reading about them in the next quarter's report?

SPEAKER_01 18:04

Great question. So, in from the forecast IQ perspective, we can incorporate any third-party data that can be fed into the system, weather being a good example. Um, events, known events, are usually inside of the event calendar for each retailer, but also outside external events, let's use World Cup as an example. Clearly, that was a factor. You had a lot of people coming into various cities that maybe are tourists and don't usually come in. So you need to start to look at the population of likely demand that's coming from these external events, and feeding all that into the AI engine allows you to train against past purchases and really ascertain uh likely probabilities and then apply those into your forecast. So that in and of itself is very important. Obviously, we get into the repunishment and the forecasting. You need to be more responsive if there's a World Cup going on in your city for the following three weeks. You need to be more responsive to how you address that. So pulling all those external data sources in is I think a challenge that most people have. It's putting your finger in the air and saying, I think we just need to beep inventory up by 25%. You know, then you're gonna find yourself sitting on a lot of merchandise that maybe you know is now stale or becomes stale and has to be marked down at a later date. So you need to be more uh responsive to all of these various factors. Our tool allows you to train against those external factors and the implications of those factors uh in the past to identify what's likely to occur in the future. And that's I think a key element is really recognizing that there's many, many factors, some of which maybe default just in the training model, some of which your team recognizes needs to be incorporated in for some period of time, and that you can then trigger it into those events. So we could be doing different forecasting, for example, for the store locations in the city where the World Cup is taking place or the Olympics might be taking place, et cetera, and then trigger it back so that it's not taking that into account for the other locations. So it really is that level of flexibility that you need in the tool.

SPEAKER_02 20:05

I I want to circle back around to the the human in the loop um element here, uh, because I think that sort of maybe worries uh potential, you know, customers is is the agentic AI portion, is handing the keys over to a machine to to um act on uh a user's behalf and and start forecasting and whatnot. Um I know Farsight IQ has these these AI co-pilots that can recommend and and actually execute actions, um, of course, with a human in the loop still, but how do you help teams move faster without losing that human oversight and making sure that there's still a proper review process? Where's that like that line between the AI acting and uh a person approving the decisions?

SPEAKER_01 20:49

No, it's a great question. I think it's it's in fact was kind of a major design principle for us. You know, we don't believe that retail AI should remove people from the decision process. In fact, retail is really too nuanced for that. Merchants and planners know things that maybe the AI tool doesn't know or understand and they have an intuition around it. The local events, vendor issues, visual change going on in the store, whatever it may be. So we want to make sure that the human is always in the loop. So if I look at the forecast as an example, forecast. Each planner, buyer, whoever is looking at that tool can say, okay, I think that's pretty accurate. I'm looking at what happened last year, and I think this might be a little bit overestimated. So I want to tune that down. And they're able to go in and modify and tweak. Now we're going to track the original forecast plus the modified forecast, and we'll tell you, were you more correct or was the tool more correct? And in fact, the tool itself will then learn from those as well. So the humans always in the loop from the forecast perspective. The underlying risk IQ I mentioned and replenish IQ, optimize IQ, those are purely recommendations. Based on your current inventory positions in each location, here's what we recommend you do. And in all instances, AI isn't triggering anything. The user has to go in and say, okay, yes, I agree with what AI is doing, or I want to adjust it slightly. So in no case are we actually writing straight back to the database using AI. In all cases, we are writing back only when somebody is approving it and pushing it through. And even then, in many cases, depending on configuration, there might even be an approver above the buyer that's actually making that decision. The approver still needs to be done and approving, okay, we're going to generate new POs. That's just not going to be AI that's generating POs for you. There's going to be human in the loop, sometimes two humans in the loop in that context.

SPEAKER_02 22:44

Gotcha. Okay. So yeah, so humans, humans set sort of the the guardrails. The AI recommends and only uh automates within the the approved sort of boundaries and parameters. Um I think that's yeah, I think that's a good, you know, a good way to frame it, good approach. Um yeah, the AI isn't ultimately, it's not replacing uh the judgment call, right? It's just sort of clearing the the busywork off of one's plate.

SPEAKER_01 23:09

Yeah, I I think AI brings, I guess, I guess I'd say scale, pattern recognition, and speed, but people bring judgment, context, accountability, and experience. So it's really a combination of those. I think that is kind of key to how this becomes effective.

SPEAKER_02 23:24

Yeah, I think that's a good strategy. Um, and of course, you know, um perhaps in the future, you know, as these AI's uh models and agentic AI capabilities continue to evolve, maybe there'll be additional sort of um uh maybe like feature sets with the more judgment-based calls with some of the automations, but um until then, you know.

SPEAKER_01 23:44

Well, I think what we'll see to that, there's probably, well, not probably, what will occur and what is occurring now is building out more, I guess I'll call it skills within AI. So allowing AI to use a very finite number of skills to actually execute on your behalf. You still, as a human, need to say, I want to execute this skill, but that skill will maybe do a little more balancing or something that the buyer might not be quite as attuned to. So there are ways to leverage AI and that kind of capability as well. And you know, we're seeing more and more of this in the AI world in general, people building out various skills that would allow things to execute. But right now, we're saying most retailers are a little nervous about that. I don't want anything touching my database. I don't want to generate purchase orders that you know have a commitment, obviously financial commitment, without somebody being involved and saying, yeah, this is the way to go. And even with the skills, we there would be a lot of constraints and security tied around that.

SPEAKER_02 24:39

Yeah, let's let's uh let's zoom out to the buyers who who are thinking, like, okay, but is this is is my company even ready for this this next leap, right? Um I know every sort of AI vendor says it just like depends on the data, depends on the company. Um, but I want to get real, like how should retailers think about data readiness before they even start you know adopting AI driven forecasting? Like how clean does the the the data and the business um you know, house, metaphorical house need to be before you you move in?

SPEAKER_01 25:13

That's a great call, because I think you're right. A lot of retailers are experimenting with AI today, but not all of those pilots turn into real business impact. Uh and typically it is, as you're saying, it's it's we need to change it a little bit more from you know a project to a little bit of a science experiment. And you have to look at your data. You have to make sure that your data is ready to work through that, but not across your entire organization. Some people are going, we can't move on AI until we do a one-year project that cleans up our data and gets rid of all duplicates and does the following. I would argue that that's probably not the right approach. The best approach is probably to say, what are the, as I mentioned before, what are the lowest hanging fruits? Uh, where can I really tackle that and where's my biggest risk and the biggest benefit for me implementing AI? And let's focus on that level of data and let's you know segment it out and let's really target in on the areas that make the most sense. So your data needs to be clean, and that's part of any project as we migrate data over to where we're going to do the training. You don't want to train against bad data. So you want to make sure that if I'm going to focus on this data, it's four years worth of history. I need to make sure that all that data is clean. Totally agree. You do need to go through that cleansing process, sometimes augmentation as well. It's not, it's maybe enrichment, not just your standard data today. And that is an exercise you need to go through. But I don't see, don't think you need to boil the ocean. I think you need to really focus on, you know, where is it that you need to focus your attention on today? Let's make sure that there's data readiness in that context, and then let's apply it there only. And then really let's look at these various quality and goals signals that you're looking for and make sure that all attributes of products are complete. That's an area I do focus on with people to say, I've got your past products, your past products you didn't attribute particularly well. So let's go back and attribute them now so that any future product that has similar attributes, I'm I'm much more targeted in. So there are strategies around that that we would work with our customers to say, here's where you need to focus your attention right now. And let's identify those areas that make the most sense.

SPEAKER_02 27:19

That makes sense. Yeah. I I the data, the data um aspect is um something that I feel like could use some more um, I don't know, representation in a lot of these AI discussions because I always think like so much discussion is around AI and and what um agentic capabilities there are based on top of the data. But like not very many people are talking about the actual data itself and how how like clean and like val um uh reliable it is, right? And uh you mentioned like you know, a lot of times companies are are drowning in data. So uh I I kind of make sense that you kind of maybe gloss over that and focus in on what to do with the data. But with with AI, like you gotta, you still have to focus in on making sure your your data, the the root, the source is is valuable.

SPEAKER_01 28:04

Absolutely. I think they're drowning in data, which I said before. And I also think that the data is often siloed, which is also the other challenge. So if I have siloed data and I have a lot of it, but I there's no way for me to easily join that data together so that I have a semantic layer where I can communicate with a single layer of data, that's kind of key. So the forecast IQ product, as we pull data from various sources, we do normalize it and we put it into a structured data format that we can easily train against so that it's always consistent in training. But again, there are times when, even though I might have a ton of data, a lot of our retail customers today, I keep using attributes as an example, but I think it's the perfect example. They're carrying a particular product type and they've never bothered to identify whether it's a high top or a low top sneaker as an example. Well, then clearly your forecasts are going to be off a little bit. You're gonna be focusing on things that are a little less important, the color, the brand, that sort of thing. Not less important, but uh perhaps in the context of the broader real forecasting, I need to be looking at other attributes associated with that. Men's clothing is a good example is a picola pal, natural pal, double-breasted, single-breasted, two button, three-button, what all those types of things are, all factors that if you haven't identified that those attributes in your past products, then you need to start going back and repurposing, augmenting, and enriching your past products so that when I'm now introducing a new product, I'm aligning more closely with the actual demand based on those attributes. That's kind of a key enrichment area for sure.

SPEAKER_02 29:36

In your experience, like what separates a really useful uh AI project from something that's um just kind of experimental?

SPEAKER_01 29:45

Well, I think useful is you need to know what metric you're trying to change. You need to understand what your end result and goal is, and then really figure out all the things that you need to do to accomplish that specific goal. It's not just let's try it and see what happens. You need to say, I'm trying it to see what happens, but here's my goal, and here's the KPIs or the set of initiatives that I'm trying to improve. I'm looking to reduce stockouts by X percent, and these are the store locations that I'm having the biggest challenges with. And then we can start to isolate the store locations, the types of products, the balance and mix between seasonal versus staple goods, et cetera, et cetera. It just sort of cascades beyond that. And then based on all of that target goals that we're trying to initiate, it allows us to really dig a lot deeper into how do we make sure that we accomplish your goals. And here's the numbers we're going to be looking at. This is what we're feeding into, train the model to make sure that it knows exactly what it the end goal is. It's just not going out and saying, oh, well, you know, here's what I think, you know, it's here's what I think in order to accomplish this specific goal. That's kind of key.

SPEAKER_02 30:52

Gotcha. That makes sense. Yeah, because I and a lot of retailers I know are they're running some kind of AI pilot themselves, but um the the those pilots, a lot of them just never actually turn into making uh they never have much of a business impact, right?

SPEAKER_01 31:06

Yeah. And I also think to that point, you know, this is a little bit of a challenge. You know, every company is dealing right now with there's a bunch of tools out there today. And there's Quad Code, there's Codecs, there's this, there's that. And can I apply these tools inside of my own business? And of course, we're applying a lot of those tools in our business as well. But the the big challenge that you often have is that you're going against data that is not necessarily standardized in a way, or there's an AI front end that standardizes it in a way to really get meaningful data back in the proper way. So that's where a lot of our tools, in fact, inside of our core platform, we're actually building out that new agentic layer so that you can access data in a more effective and efficient manner, just in our core operational database. You're looking at uh you know various um MCP servers, you're looking at uh GraphQL data so that you have a more standardized language, I guess, almost to communicate with the data. So all that stuff is kind of critical as well. And just implementing tools and saying, let me point it to this data and therefore I'm gonna do better, unlikely. Not to mention the fact that these tools are changing almost daily. Um each week there's a new release of one of these models out there, and there's various harnesses that will tie into your data better. And so there really needs to be a structure. So that's in fact the advisor IQ I mentioned earlier. We built that as a harness that's looking at your data that has very targeted skills that are already predefined against very targeted data that's also predefined. So I have that language that I'm communicating with and how and through, that's all bringing back some really highly effective data. If you're just experimenting on your own, I suspect you won't be particularly happy with the results. So you need tools that are really targeted for that purpose.

SPEAKER_02 32:55

Well, speaking of results, uh for a retailer that does go all in on Farsight IQ, what does the success metrics uh, what do they actually look like? Like let's say a year in? Is it is it like fewer stockouts, is it better margins, is it just faster overall planning? What's the what are those what's the kind of the scoreboard you're looking at?

SPEAKER_01 33:16

Good question. I think kind of tied back to what I said earlier. For some retailers, success starts with you forecast accuracy. If you reprove the demand signal, like I said, downstream, everything else improves, buying allocation, replenishment, pricing, marks downs, planning, everything else. Other retailers might be looking again at specifically fewer stockouts, those types of things, but reducing excess inventory, improving sell through, protecting margin, which a lot of people sort of ignore, um, or reducing markdown exposure. So I think there's that productivity story as well. Planning and merchandising teams, spend a lot of time gathering data, reconciling spreadsheets. A lot of people are still using spreadsheets today, debating about assumptions, and then we're trying to also bring everything together so that you're looking at one common set of data, and AI can kind of surface these exceptions more easily for you when you're looking at a common set of data. So I think there's even that operational productivity set of things. So you are we seeing risks earlier? Are we acting faster? Are we connecting merchandising supply chain finance, uh, store operations around that same demand signal? And then more importantly, I guess to the earlier point we were talking about with the human in the loop, are we learning from these outcomes? We being the team, but also we being the aggregate of the AI and the team. And are we making our business more responsive? So I think that's kind of our goal. Success looks like to me is measurable operational improvements, a smarter way to run your business over in general.

SPEAKER_02 34:49

It sounds like that they're all they're all very interconnected, right? Because I feel like you know, better forecasts, you're gonna uh it's gonna lead to fewer markdowns. Fewer markdowns is gonna lead to to better margins, and then um, you know, your your planners, um, whoever's, you know, the humans in the loop, they get they get their time back on top of it, which is um always a plus. There's a chain reaction of positive outcomes, it sounds like. Um, but this is this has really been really insightful, and I appreciate your taking your time to to give us the rundown of of the Farsight IQ platform. Um, I want to circle, uh just like cap off this episode and focus the the last question on you um because you've had a long, you know, career working uh in retail, but both, like I said in in the intro, both sides of the counter. Um, you were uh COO, CEO. Um, you're on the SourceForge podcast. So, I mean, you obviously have uh been quite successful. Um, if you could go back to the start of your career, what's what's one piece of advice you'd give yourself?

SPEAKER_01 35:48

You know, sort of how my career has kind of gone anyway, that I would probably tell myself to pay as much attention to the decisions being made and not the systems. So as I transitioned from retail, the first half of my career, into technology, there's a tendency to jump into the systems and what the systems do and what they're accomplishing rather than what's the decision behind what that system is supposed to be doing. You know, so earlier in my career, especially in technology, you know, it was easier to focus on the features, modules, these screens, how it was implemented, all of those things. But over time you start to realize again, it really goes back to what we were trying to accomplish and how do we accomplish that more effectively. So I think that's probably the main thrust of what I would say if I really could go back in time. And it's kind of been my path, I suppose. But where I would really want to focus on, even when I get into technology, is are we actually delivering on the promise of what this is supposed to be doing?

SPEAKER_02 36:47

I love that answer. That's that's so true. And I feel like it's it's harder, becoming more and more difficult for uh individuals to make you know effective decisions um and improve their decision making when we when we kind of offload a lot of our our processing to these external um you know brains uh with AI and whatnot. And so it's it's um you gotta you gotta take everything to account, make the proper decision that'll set you up for success. I think that just runs into the in the DNA of um Jesta too, right? You guys have been in the business uh for 55 some years. You kind of obviously have been, I'm sure, have been watching this AI uh moment and like they're determining when to jump in and and and when to actually make uh a product and a platform that's gonna make a difference. And um, you know, you don't want to to rush it out prematurely.

SPEAKER_01 37:36

And so Yeah, I think Farsight was a a very major multi-year investment. We've been, you know, before we launched it, there was a lot of stuff going on. And of course, the market was changing very rapidly in that period of time that we were building this, and even as we were building it, we were making that iterative series of adjustments, but there's no question about it. And of course, we continue to add and build on top of it because things are still continuing to change. So we need to be uh receptive to that and and adapted to that. You know, it's it's really about helping, I think, retailers move from hindsight to foresight and then from foresight to action. So there's where I think we're layering those extra from foresight to action components as we speak today. Some of the other things I talked about skills and being able to allow the agents over time to do more work for you, but I don't think the market's quite there yet. So let's make sure your humans in a loop. They're very uh very involved in making the final decisions and using their judgment as well.

SPEAKER_02 38:34

So well, for those interested in learning more about the platform FarsightIQ as well as JESTA, where's the best place to go to learn more?

SPEAKER_01 38:43

I think the best place to go is I would suggest right now FarsightIQ.com. FarsightIQ.com also ties back into the jestis.com. Uh and they can certainly go to this contact page, and I would welcome people to schedule a meeting with us on really either of the two platforms uh or reach out to me directly. I'm happy to have people reach out to me directly as well, uh, which is just spierson at jestis.com and happy to further conversations. Looking forward to it.

SPEAKER_02 39:12

Perfect. Yeah, we'll include links to everything mentioned uh down below in the in the show notes. But Scott, thank you so much for everything you shared with us. I I really did uh I think I think this is a really valuable conversation, and I hope uh and I know listeners and viewers will get a lot out of it. So um I'm excited to hear the feedback on this episode.

SPEAKER_01 39:29

But excellent. Thank you very much for your time.

SPEAKER_02 39:32

Thank you for listening to the SourceForge podcast. I am your host, Bo Hamilton. Make sure to subscribe to stay up to date with all of our upcoming B2B software related podcasts. I will talk to you in the next one.