Kinaxis Inc. (KXS) Earnings Call Transcript

All right. Hello, everyone, and welcome to our webinar today. Our topic today is the next phase of supply chain AI from promise to accountable outcomes. So I'm excited about the topic today. My name is Justin King, I'm Field CTO with Kinaxis. And we have the privilege of having some interesting support. A research director from IDC, Eric is joining us with some really interesting research that he's done. And this isn't across a couple of dozen of respondents. This is thousands of companies or thousands of responses across companies around the world, different industries, different geographies. So really excited about the perspective that we're going to get today.

So as we kick off, Eric, why don't you introduce yourself to everyone? And maybe just give us a little bit of perspective of what you guys do generically, and then we can dive into this particular research project that you are working on.

Yes, of course, excited to be here, excited to talk about, gosh, what's really kind of a front-of-mind concept for all of us. So for those that don't know me, I actually kind of come more from the practitioner side of supply chain. I worked at Nike for 21 years, just kind of navigated all parts of the company, different geographies and functions, and as you can imagine, you might do in that amount of time. I have a pretty deep background in supply chain planning, in particular, and have kind of brought that over to IDC, where we -- as you mentioned, we research across just about every company in the world that has a supply chain. And we kind of dive deep on data and information and just bring that sort of practitioner side that I bring, but also IDC brings some really heavy data focus throughout supply chain and gives us an opportunity to kind of spot trends, have some insights. We work with a lot of people who kind of look like customers of people like yours. So just nice to be here, and excited to chat.

Yes, absolutely. So this is not going to be a sales pitch, right? This is really a research-led discussion. I'm going to give some perspective on -- from what our customers have been talking about and asking. And of course, opinions, there's no shortage of opinions in AI and what it means and what's real and what's not, what's hyperbole and how organizations can start to be successful with AI. And I think that's really where some of your research starts to uncover where supply chain leaders are, where their heads are at. And we're going to talk today about what comes after initial adoption of AI. And it's the stuff that's not as exciting and sparkly and shiny, right? It's the things like trust and governance and data readiness and guardrails and human oversight and maybe most importantly, how do we get to real measurable outcomes, right, that are aligning to our business objectives and our KPIs. And I think there's a lot of nuggets of important information that we can extract from the research that you guys are doing. So let's jump right into that. Eric, why don't you kind of set the stage in terms of who was this -- what was your out -- who are you reaching out to? Who's responding to this survey? And I think what you found is that in terms of AI readiness and adoption, most everyone was saying they had started something, but at the same time, no one was really saying that they were a leader, right? Everyone was kind of clustered together somewhere in the middle. So I'm curious on what's your perspective in -- what does that tell us, right? Where are companies on their journey? Do you think we're overestimating, underestimating where you're at? Do you think that's fair in terms of where the market is today with AI? So why don't you start with painting the picture of this research project and who your respondents were and some of the basic learnings in terms of the pool of responses?

Yes, yes. Great point. And actually, before I even answer that, maybe I'll even go back because you just triggered kind of a thought to one of the many roles I held at Nike. I was a Strategy Director for a little while. And I like the way you just worded something, which was sometimes maybe we're not talking about what sounds like the glamorous and exciting stuff. And that was the way I viewed that. There were times where people would come to me and say, "Oh, you're a Strategy Director, that's really cool. How can I get that kind of job?" And I would often tell people actually a lot of it is the discipline. A lot of it is building a strong foundation. And when you do that, then you have something you can really build a larger structure on top of a strong foundation. So sorry, you just kind of reminded me of that thought. I thought that was worth underlining because I think it's going to show up through our conversation today, too. I mean, you just mentioned adoption. And I mean, probably most of our listeners today, I mean, are like us. We've all been dabbling with whether you're using ChatGPT or you're using Claude or whatever other tools you're using. And then, of course, supply chain vendors are now embedding these tools within our supply chain functions. We've all kind of played with it. So to your point, we were asking, I think it was over 2,000 respondents around the globe, kind of all geographies, all industries, different levels of companies, executives, directors, different functions like logistics and planning. And we had almost 98% of people say on some level, they're touching AI or AI is touching their supply chain, let's say. And so the question isn't really adoption, per se. It's just kind of how much, how far. And then in particular, are we realizing value from it. And I think you maybe said too, so maybe forgive me if I'm repeating, but we saw through this data that only like 12% of the respondents, almost barely more than 10% kind of say, hey, we're kind of in the lead here and would -- I think back to my Nike days, we were very competitive. We wanted to be at the leading spot of a lot of things we did. And so there's only like 10% of companies saying that they view themselves as a leader in the space. So it's clearly something emerging, and people are still figuring it out. And then something else we saw -- and forgive me for maybe throwing around a few numbers, I'll flash up some slides here in a minute to try to land home on those. But like 50-plus percent, more than half of companies are saying, what's actually slowing us down from going farther and really becoming a leader is just a trust issue with our data, with our systems. Do we even think as we play around with, say, agentic AI, is it really going to do what we want it to do? Is it going to give us the outcomes we want? So kind of seeing all of that.

So okay. So you hit on a really interesting topic, right, and that's trust. And what we're seeing, and even as a software company, what AI has enabled is it's so much easier for us now and for our customers themselves to do a pilot, right, to do a POC and to kind of do some vide coding and say, look, this is the realm of what's possible. Which is amazing because it's democratized access to our data, it's democratized access to mathematical solves and even like machine learning routines and such. So we can quickly and easily see what's in the realm of possible. But of course, that's being built with no security in place, no governance in place, no logging, no -- none of that structural stuff that I was talking about that's going to enable that trust, right? So as you look at your response, I think you said about half of them are saying, "Hey, trust is a real barrier." What do you think companies are most concerned about, right? Do you think it's like a technology issue? Or is it more of a process, governance, operating model, that sort of thing?

Yes. And I think back to those days when I was maybe a Strategy Director, one of the issues with really realizing value from your systems is, is your organization even ready? Do you have the skills? Do you have the ways of working even across functions? You mentioned democratizing, I mean, as agentic AI, as large language models, different things like that, as they help us to cross functions, if I'm an executive at a company, I'm asking, well, okay, but if I unleash all of this, are my teams even going to work together? Are they going to get the right outcomes? What's happening to roles? How do we shift roles? Do I have the skills to leverage all of this? So I think that's just part of the conversation. And if I may, maybe I will share something. Give me one moment, apologies. Hopefully, this is showing now. Can you see that on your screen?

I do, yes.

Okay. Well, so I think one of the things -- I'm going to come back and answer your question, too. But -- so we see -- we talked about everybody's piloting. And I like that you said proof of concept, whether you want to call it walk before you run or whatever it is. But companies are saying, well, let's at least start to see what we can do in that sort of art of the possible. But what's interesting is today, if we ask these 2,000-plus respondents in companies all around the globe, like I said, every geography, every industry, every company size even we looked across, we get consistent answers. You might think maybe some more mature companies might answer differently than others. But we actually got pretty consistent data where we see that companies are kind of saying, today, in terms of automating things, having autonomous supply chains, we have these conversations about always-on supply chains and lights off warehouses and things like that. But the actual autonomy right now, companies are saying it's only like maybe 6% have any form of sort of, I'll call it, real autonomy. Meanwhile, 40% say in the next couple of years, they aim to get there. So that's a heck of a jump. And so the question then, if they're telling us out of one side of their mouth, they're having problems trusting the data, trusting the tools, maybe even trusting themselves to get the outcomes they want. But then they're also saying, yes, in the next couple of years, we plan to just kind of 7x the amount of autonomy we have. So one of the things I think -- and here's where I'll kind of come back to your question is if we look at it through the lens of governance, I'm asking even from my old roles, people I know, I'm sort of challenging even -- it's not just the technical side that needs to solve this. I would suggest even there's sort of three, I don't know, legs to a stool or something like that, that the capabilities are coming. The tech is advancing perhaps faster than we've even sort of prepared ourselves to use it. Now that -- there's an optimistic side to that because that then means, wow, there's a lot coming that I can really improve my supply chain and improve the outcomes. And so I think what needs to happen is there's sort of a business strategy. Where do we think we want to even use these tools, where do they best apply? And by the way, I know this slide says agentic, but we're talking AI overall even. I would even almost expand out and talk about a complete AI strategy. Where do we believe from a business win, and there's even force ranking from time to time. Are we going after logistics? Are we going after, I don't know, warehousing? Are we going after supply chain planning? Where do we think we really want to sort of place our bets here in terms of the business side? And how are we working together? That matters. But then also, what I'm suggesting is there's two other pieces. One is the technical side of things, where even what are the guardrails for these tools? What are they allowed to do? When is human in the loop? We talk about that. How are we going to allow autonomy versus where do we actually want to engage, like I said, with sort of human insights. I think it will be something we may talk about today, too, where are we bringing in domain expertise from the humans and things like that. And then I think an under sort of, I don't know, an undertracked side of all this, we all know that there's some kind of economics emerging here in terms of the use of compute power, whether it's tokens, whether it's value-based pricing, all these things that we're starting to see emerge. So what are the guardrails even around -- I myself, the first month, my company gave us the sort of, let's call it, Claude tools. I tapped out my tokens within like the first few days because I was just trying to do all kinds of things with it. So how do we put guardrails around even where is the spend and how much are we spending? So I think forgive me for rambling a bit there, and maybe I'll even stop sharing here for a minute. But I think just companies maybe contemplating their governance around these issues and how do we even -- what's our own internal strategy. Because then that informs even what are the partners we work with and how do we work with them and all of that.

Yes. I think to -- from what we're hearing from our customers is the magnitude of the impact of the decision can change even inside of supply chain, right? And so you could have thousands of decisions made maybe inside of a warehouse that are much lower risk. And you can look at that in an aggregate and say, okay, is efficiency increasing, is accuracy increasing. But then you can pivot to the polar opposite side and say, well, in planning, if one decision means I'm cutting a $1 million purchase order, well, the risk and the trust behind that decision has to be -- are both much higher, right, because there's real constraints in play. If they were not considered, then I cannot produce the finished goods. There's customer commitments. Like if I miss this customer shipment, there's financial trade-offs. There's operating all these different things that are behind one decision versus just being able to roll it up and look at it in a consolidated number. So planning is a really interesting one, I think, because of the magnitude of information that's behind every decision, the context that's behind every decision and in turn, the risk of making the wrong decision. So I agree with you, I think trust is a big part of it. And I think that then leads a little bit into the data side of it, right? So does your agentic system have the right data to make a decision that you can trust. And I think, if I'm not mistaken, your data had some pretty strong opinions from your responses on how important -- well, not only how important data readiness is, but maybe it being a bit of a barrier, right, and kind of holding companies back a bit. So what can you tell us from what your respondents are saying about maybe their readiness and how they saw the importance of their supply chain data in order to move into more agentic decision-making?

Yes. I think that question definitely came out pretty strongly in the data, Justin. It's where we look at -- and I would even go back. Data quality has just kind of -- it's an evergreen field. It's something we always need to focus on in supply chain. We had data management teams in my time at Nike. I'm sure it's still there. It's something we need to keep focusing on. So I almost look at it as a coin with two sides because on the one hand, this survey work that we did in the field would say 62% of people, almost 2/3 of our over 2,000 respondents are saying that better data quality, better integration would actually be an accelerator for AI. So there's certainly a focus we need to have on data. And I would even say, again, from that sort of business practitioner lens, we need to find the right partners, too, that are helping us with that and have contemplated these issues and thought that through. But on the flip side, I would also say -- and now just personal opinion, this is not necessarily the data. I wouldn't say that we should use that as an excuse to not move forward either. Because sometimes at IDC, I think we find -- some of us have a personal point of view that even the act of moving forward with projects in this space sometimes itself creates improvements in our data quality. And so it's kind of like it's a chicken or the egg kind of discussion. Do I have to clean up my data to use AI? The answer is yes. But does the act of moving towards these solutions also kind of yield sort of data improvements? And the answer is also yes. So I guess it's something that absolutely the sort of survey findings have sort of fleshed out in terms of AI readiness. It's kind of a what's my data quality question. But I also think it's not this like go/no-go checkpoint. It's just something we need to really solve and continue to move forward.

Yes, that's a good point, Eric, because AI can even help us with our data quality issues, [ can it not ], right? So you've heard concepts like the self-killing supply chain where maybe you're working with a certain lead time. And you can look historically and figure out, am I building my supply chain plan based on accurate information. AI can help us project it, detect seasonality, all these different things. So I couldn't agree more, and it's -- I think there's two ways to look at it. One is more traditional data quality, like is the data there? Is it proper in the field within some sort of bounds? And then it is the context correct, right? I mean, are we defining our constraints correctly, our dependencies correctly, our demand signals correctly? And then you got the unstructured data side of it, where there's just having unstructured risk signals available, right, which can give us directional guidance to picking alternate suppliers in those sorts of cases. So there's a lot of use cases where AI can use not clean, structured data to help guide decisions in the same way that more structured data and operational context is important as well. So as we -- so we kind of covered the data side of it, right? And now we're looking at the foundations that are in place, and now we're looking for decisions to come of that. But then we need to balance that with measurable outcomes, right? It's not just better decisions, but it's about the right KPIs, the right business objectives and making sure we're just not using technology for the sake of technology, but we're actually achieving our business objectives. So in a way, AI needs to be accountable to the business, not just accountable to technology that checked the box, we successfully deployed it, right? So I think some of your findings were -- we were kind of highlighting the clear ROI, one, is important; and two, it would be a driver to accelerate that investment, right? As we look at it less about decision-making and more about achieving business objectives. Did you glean that from the data as well? What are you hearing from -- in terms of the [indiscernible]?

Yes. I think both from a data standpoint, yes. And then even anecdotally for myself, it's -- as executive teams or, let's say, execution teams are clear on your -- whether it's your metrics, your KPIs, your strategic imperatives, whatever it is at your company, when we clearly defined the goal, we had a lot better chance of achieving it. And I think we've even seen that -- I'll just even again say for myself in my own piloting with AI. When I don't get the outcome I thought I was going to get, I often find, oh, okay, I didn't make it clear, what I even really wanted from you. I'm sort of using quotes as the AI as a person. But -- and so I've learned to better articulate what the thing is that I want out the other end. And so now that's back to that governance and that strategic role of are we aiming at days in supply chain? Are we aiming at speed? Are we aiming at cost? What is it exactly we're solving for? And of course, we've got tools now that are starting to even maybe solve for more than one variable at a time, but there's still -- what's my strategic priorities? What are the real outcomes I want? And so you're right. I mean -- and I'm just kind of looking down at my notes, we had -- more than half were kind of saying, "Hey, if I'm going to move forward with these tools, I want to have what's the clear return I'm getting." But also, there's a lot of folks who are saying, "I'm sort of not sure what my business case is just yet, and so that's causing me to maybe slow down a little." And so there, again, I think whether it's through the lens of business governance and knowing clearly what my strategic priorities are so that I can even inform the AI itself to try to solve for those. And that's how it becomes accountable. I have to even know what my goals are to hold it accountable to my goals. And then besides that, even having a clear business case, why am I implementing these tools? Am I trying to have better execution across functions? Am I trying to reduce my days in inventory? It's just important to know what we're even trying to solve for.

Yes, that makes a lot of sense. So would you say that have we just not figured out how to measure the impact of AI the right way yet? Is that what it boils down to?

That's hard. I mean -- and that -- let me be fair almost to AI. I used to manage strategic projects, and measuring the real impact of them has always been a big issue. Like can we really define ROI of a specific project? And there's ways we all got around doing that, and there's best practice, of course. But yes, I think it's a good question you asked, like how do we even rightly assess what the AI did and what the outcome of it was? And I just think that continues to be a part, again, if I could sort of belabor the point of governance, like, okay, how am I going to measure this? How do I look at it? How do I determine if it did what I wanted it to? These are just kind of the things I need to be thinking about.

Yes. No, that's really good. Yes. I think -- so if I were to summarize, right, I think it's better insights, faster recommendations, increasing that speed does not necessarily correlate to getting better results, right? And so there's an opportunity here for organizations to, I guess, close that gap between just making a decision and genuinely achieving those business outcomes and objectives, right, that move the needle.

Well, it's interesting. And here, I'll be fair to people in your role, for example. We all ask on the business side, I want to do it faster, I want to do it easier. I want to do more. And so those are specific outcomes. You just talked about, can I get to a decision quickly? I don't remember your exact words, so I don't want to put words in your mouth. But I might measure from, "Oh, wow, it used to take me a week to come out of an S&OP cycle and answer an executive's question." I've literally been in those meetings. And you go back and you have meetings to follow up from the meeting and you have slide presentations you're building and questioning, and now the data is stale a week later. So it's certainly a value from some of these tools to get to information quickly, get to decisions quickly. But I think there's something underneath what you just said. And again, I don't want to put words in your mouth. But just because I get to a decision quickly doesn't always mean it with quality. And so there's both sides of this. And I can imagine from your side of the world, we're not so consistent on the business side, let's say, of demanding tools that give me automation, give me speed, help maximize my resources. But then on the other hand, am I really actually testing for, oh, was this the decision I wanted to make? Maybe it would be worth waiting a week. If I took 2 days of inventory out of my supply chain, there's a real value to that. So of course, we're all going to say we want both hands. So give me tools that are fast and accurate.

Yes. So we're ambitious, right? I think that's what you're saying. And you talked a little bit ago about, I guess, the ambition a bit outpacing readiness and seeing that in some of the data. So as an organization, like how do I decide when AI should come in and help me -- just recommend and give me guidance, when AI should help me act, right, and go ahead and do things autonomously? And maybe somewhere in the middle, like when humans should take the recommendation and take some guidance, but always be approving each step, right? That's kind of the governance side of it and the gaps there. So what's your view? And maybe, did the data tell you some things about kind of ambition versus readiness? And how do we land on deciding where to inject AI in this process?

Yes. Part of it is that governance, for sure. It's -- I'll go back. If you had -- maybe there's some monetary guardrail, for example. Like if I'm making a decision to purchase $1 million of product versus I'm making a decision to start a $1 billion project or something if I was in project manufacturing. Well, that's very different. And so there's some sort of scaling of decisions. And deciding for each company and each industry, that's going to be different. So I can't necessarily be overprescriptive about that. For your company or your industry, that -- it needs to be contemplated. Where are we okay allowing some automation, where do we feel like we need to be involved in the decision. And that's actually just kind of a roles and responsibility discussion. So if we kind of look at it that way, these are conversations we've had for decades in business. What are the roles and responsibilities? When does something need to escalate up the chain? When can you make a decision at your desk, if you're a junior person, a middle executive, a senior executive? Now we're just making that decision with agents in a way having the role. And so there's sort of roles and responsibilities and, I don't know, size of decisions governance piece. And then I will just kind of circle back on the data quality part again that -- I think there's a testing or there's some kind of internal effort and an effort with partners, too, to say, am I getting the outcomes I expected? If I do -- if 9 times out of 10, I've proposed something like a small PO change, let's say, okay, you're going to change the mode to airfreight on something, something like that, or reduce this purchase order by x units. If there starts to be a certain sort of statistical acceptance rate on some of this stuff, then if I was back in my roles leading teams that were planning around the globe, I'd say, hey, if we -- 90 times out of 100, we've accepted the recommendation, maybe we need to start thinking about automating that, too. So there's a lot of elements of roles and responsibilities and data quality that are just going to keep being in every part of this conversation, I think.

Yes, absolutely. Yes. We don't want to slow down because the opportunity is certainly there. But in all of our development and rollout, we have to make it safe. We have to make it explainable. We have to make it useful. We have to make it tied to a real operational environment. And someone said autonomy without governance creates risk. But governed autonomy creates confidence, right? And I think that's what we're looking for, for our team.

And by the way -- and forgive me for interrupting, too, but you've just reminded me something that I think is really important for anyone listening to us today. And I'll sort of like come from the place of having had roles like this. It's become my perspective, it's an evolving perspective that we give humans grace and we give machines a lot of judgment. And the reason I bring that up, you just mentioned explainability, and that is very important. We need to understand why it was a decision. But I think we're starting to kind of embrace the notion of acceptability, too. And let me say what I mean by that. Like take a Tesla automated car. I literally have two teenagers in my house. And so we've moved to the place of their driving. And it's like, okay, as a dad, you wonder, are they safe and have I trained them well and all of this. And if you look at perhaps automated driving, when we notice it making a mistake, it's headline news. On the other hand, if we looked at it statistically, it might actually be safer than my teenage drivers. So what -- where is the level where there's some kind of acceptance, let's say, back to these PO changes? If it turns out that it can do it more accurately, then maybe I'm freed up to do higher-level thinking and -- rather than sort of use fear-based thinking on, well, it got this one wrong out of 1,000. I might look at that and say, actually, that's still 99.9% accuracy. Maybe I can accept that. So just sorry for kind of tangenting on that. But I think on all of this, as part of the governance, there's even a contemplation of what's our rate of acceptability or something, what's our accuracy? And if it's within these guardrails of accuracy, I'm good.

Yes. No, that makes sense. Yes. So let's dive into that kind of the human side of it, right? As we're evaluating and as we're -- and maybe even evaluating our changing roles, right, as this technology comes in. I think some of your survey questions leaned heavily into like how it's going to change individual roles and such. Did any of that surprise you? I think I saw all the way as much as 80% seeing it more opportunity than threat, right, which I think that's a good thing. Did anything surprise you about the data that you saw coming back?