7. Part 2 - Before you buy Pricing AI with Suzanne Valentine

Suzanne Valentine: Foreign.

Host: Hi everyone, and welcome back to another

Narrator: episode of the Sweet Spot by Pricing Works, where we aim to inspire you to achieve pricing excellence through conversations with fascinating guests.

Host: Today we will focus on the second

Narrator: part of our interview with Suzanne Valentine, the senior director of pricing AI at, uh, PriceFX. If you haven't checked out part one yet, please do that first. If you've already listened to part one and are keen to hear more, let's dig in.

Host: First step is to make sure that we can get a coherent list of, like, the principles and the methodology for setting and managing pricing today that we can all agree on. And that reflects how we do business. Right.

Suzanne Valentine: As a company and the objectives. I think, you know, part of it is like, what are we trying to achieve with pricing? And that really varies depending on the size of the business, the type of the business.

Host: Right, exactly. And it could be that we also have objectives that we're not able to achieve today yet, or we're not able to monitor how well we're doing on different levels. Right. So objectives is a very nice then bridge to optimization, because optimizing pricing, like, what is the proof of that, that we've optimized our pricing is that we've achieved the objectives. Right. The aspirations that we have around our business objectives.

Suzanne Valentine: Right, yeah. And so even if you're not starting with optimization, defining the objectives and you know, articulating the business rules which are obviously foundational for the optimization, you know, that alone is an important starting point.

Host: Right, right. And, uh, so we are at the optimization stage. I know you've mentioned earlier what the AI can do for us, but maybe just to dig into that a little bit more, what is the value of having AI do the optimization for us and what can it come up with that we humans cannot? So what is kind of the value of going that extra trust type? Yep.

Suzanne Valentine: Yeah. This is such a great question. So first, the reality is that not every pricing problem needs complex machine learning. Sometimes, as we've been discussing, a good set of rules or a simple statistical model is perfectly sufficient and frankly easier to explain for some organizations. But that said, uh, there are some situations where leveraging very sophisticated science is appropriate and warranted. So for example, our question core optimization approach leverages a technique called multi agent AI optimization. And it's one that not everyone's familiar with, but it's a very modern and scalable approach that involves breaking a problem into different objectives and constraints that each have an agent essentially representing them. And the agents are, as we run the optimization, the agents are Essentially collaborating to find solutions that maximize objectives such as profit and revenue, while upholding as many of the business guardrails as possible. So I think there's a nice, you can kind of picture this in your head. You've got, you know, these, these agents that are responsible for achieving some sort of revenue goal and maybe there's also some profit goals as well. And instead of having to choose one or the other, you can have a blend within this optimization model and then you have all the little agents representing various business constraints and maybe you've prioritized the constraints to have some sort of hierarchy and so that all of these agents together are working to find the best solution to this problem that they can. But this obviously requires a fairly sophisticated engine. When we talk about what algorithms or AI will bring the most value to a business, we focus on three main factors. The first is complexity. So if you're dealing with thousands of SKUs, multiple customer segments, regional variation, dynamic market conditions, that's where sophisticated optimization really shines. But if you got 50 products with straightforward pricing logic, you may not need optimization or you may not need optimization or even AI to get started there. The second main consideration is really data availability. So machine learning models are hungry for data. If you don't have rich transaction history, customer behavior data and market context, you're not going to get the best results from advanced optimization. It's probably overkill. Or you might get some results that just don't make sense. It's probably better in a case like that to start with something simpler. And then the third consideration is really the business impact. Where can optimization move the needle the most? So you should focus your science investments on high margin products, high volume segments or areas where you're leaving money on the table versus, you know, trying to optimize down to every last product in the, in the long tail.

Host: Right, right. And that makes a lot of sense. And you know, these little helpers, these agents, they sound amazing. I wish I could delegate all of my time. I believe soon I will be able to do. I think that's, you know, what we're heading towards. But maybe for those of us who are not that used to having this kind of technology, help us out, act as assistants to us. And obviously we're trusting them with important work.

Suzanne Valentine: Right.

Host: With important tasks, objectives, etc. Some work we want them to do for us. I know that agentic technology, I mean we're talking about it now a lot, but it's been around for a long time. It's come up in the context of pricing perhaps a little bit more recently. But can we just talk a little bit about what agents actually are in the sense of where else have we seen agentic technology and how can we transfer that to pricing so that we know what, what these little helpers are and whether we can trust them and why? Because it actually has been around and we know what we're doing. It's not super, super new.

Suzanne Valentine: Yeah, you know, everybody has a slightly different mental model of what, you know, what's an agent, what's not an agent. I heard someone say recently that agents have agency. And I think that's a nice way to think about it. So it's, it's typically more than just, you know, automation of, uh, some sort of algorithm. There's some sense of, you know, you've given this agent or, you know, the algorithm a broad set of context and they are, uh, synthesizing that context in order to, with agency, determine what the next step should be. And the tricky thing here, I think is really like how, how much should you trust the, your, you know, your, your brilliant coworker? Um, so, you know, agents can work with. I think there's a whole conversation to be had on how agents can work with humans by having, you know, figuring out where the human should be in the loop. So, you know, the most successful AI implementations are really leveraging agents to amplify human judgment by handling the routine cognitive load and then freeing the human professionals for the strategic decisions. For example, rather than reviewing every single decision, you can design systems where agents are handling fairly routine operations while the humans are monitoring performance dashboards. Pricing managers should be spending their time on analyzing margin trends and competitive exceptions flagged by agents, not manually approving thousands of quotes. And I think an important part of this process is really ensuring that you have mandatory human touch points. So, uh, maybe weekly reviews of the agent performance metrics and calibration sessions to adjust the parameterization that's feeding into their context, and regular strategic reviews assigning or assessing whether agent objectives still are aligning with the business goals. So I think, I think, you know, I'm going on a bit of a tangent here, but I think agents can start with some automation. But the more that you can give them context so that they have agency, the more value they're going to bring to humans.

Host: Right. So they can act independently, they have agency. It's a little bit more than just standard automation, but they still act upon the or, or towards the objectives that we set. They still act under our supervision. Right. And we decide what kind of tasks we delegate to them and we use them. We leverage them where they're most useful so that we can focus on things that are of interest to us. Uh, can we, if we go back to optimization, a lot of the times when we talk about AI, AI based price recommendations. You talked about star target floor pricing earlier. The science will help us with recommendations. It'll optimize the pricing for us. How do we know we can trust it?

Suzanne Valentine: Trust is obviously absolutely critical. Um, one thing that impressed me early on at Price Effects is the commitment to transparency. And this manifests in a number of ways. So first, we're very open to sharing our source code so the customers can both understand the underlying logic and make adjustments specific to their business. And this can be anything from changing the nomenclature and labeling to reflect their vocabulary in the UI to changes in the workflow or even changes in the underlying algorithms. Another flavor of explainability is helping users understand how the system arrived at a price recommendation. So this can be in the form of tables with interim results, such as which constraints were adhered to in an optimization. And ideally, there's some visualizations that help users understand why a particular recommendation is optimal. You know, this, this, as we've discussed, really matters because pricing and sales teams won't adopt, uh, recommendations they don't trust. If it's a black box that just spits out numbers, people are going to override it or ignore it. But if they can see that the model is considering demand elasticity, competitive positioning, cost structure, strategic objectives, and they can see how those factors are weighted in deciding what the price should be, they're much more likely to, to trust the recommendation and use it. And actually, on this topic, Price Effects has a very cool feature which we refer to as bring your own science, which allows customers to upload their own algorithms on our platform and run them off the pricing data that they're curating. We're definitely seeing a trend of collaborative AI and data science versus pricing vendors having one particular way of approaching the science and insisting that their customers accept that approach. Um, many larger and mature organizations have, not surprisingly, already invested in some form of data science, and they want to build on that investment, not start over. So bring your own science kind of gives you the best of both worlds. Where, you know, Price Effects has invested in, you know, many building blocks and algorithms and optimization routines so that, you know, customers can get started with that. They can also, you know, if they have a mature pricing organization that works within data science, they can bring those algorithms to the platform and perhaps better Scale their use of data and it gives you the single source of truth for data in the Price FX platform and then all of your cumulative investments in one place.

Host: That's really incredible. And I'm assuming that also this kind of co designing and everyone having some kind of input and kind of working out together what that solution looks like. I assume that also helps with change management and getting users on board, et cetera. Knowing that the company, company itself has got a significant input into shaping up the solution that its own science perhaps being used. It's not just some third party asking you to trust what they have built, which you maybe don't understand. So there's definitely a lot and change management. It's a beast of its own because. And we'll talk, I think a little bit later briefly about implementation challenges and then post implementation challenges where we have this educational piece, right. And getting users to use the platform because we can have all the great science there. But if they don't see the value, they don't trust it, then uh, that's not quite what we want to achieve. We want them to make the full use of it. Staying a little bit with this concept of flexibility. There's a lot of clients that we're talking to that they see what's happening in the market, they see the speed of um, developments and they see that there's new tech coming out and they're trying to make sense of it and how it can work for their business. But there's this sense of dread of getting started and this idea that I don't have a vision for the next five years or 10 years. And think about that when things are changing so quickly, like what if I implement something today and then maybe in two years that's no longer relevant, like how do I adapt and what if I don't have the right vision today?

Suzanne Valentine: Right.

Host: The more I learn, the more I need to adapt. So this concept of flexibility I've seen also clients asking for, you know, can we build something as we envisage it today and then scale that later in different directions depending on also our journey and how we grow as a business. Can you comment on that a little bit? Because I think that there's also a cost to that. Right, obviously. But what does that look like in practice? Like when can you get started with something small and then how can you flex in different directions afterwards to keep up with a technology?

Suzanne Valentine: I mean, you're touching on such an important point. You know, if we were to look back over the past even two years, the technology has matured so much. Like, you know, what you can do with LLMs, how LLMs can inform agents, you know, if you planned a two year implementation now, like, how are you to know how technology has advanced and you don't want to be locked into, you know, one particular approach and by the time you get to the end of that implementation, realize, oh my gosh, we're two, you know, we're a year and a half behind the market now as I mentioned, I think getting together the core data sets and start using some agents to understand the data quality and what might be missing. Putting together a data roadmap so separate from the technologies. It's like, what data do we want to curate? What work do we need to do in order to clean up some of these data sources and harmonize some of these data sources? That's something that every company can do now and that's not going to be wasted work. You're going to need this data foundation regardless of where the technology goes. You know, one, one thing we're seeing, I think there was a lot of hype maybe a year ago about how agents, oh, uh, agents can do, you know, it's magic, they can do anything for you. But now there's more realism around, you know, agents and AI is amazing in the pricing space, but you have to provide that semantic layer and you know, make sure you've got the right, I think the business rules are even part of that. You've got to have the right structural elements in place so that as technology evolves, you know, you've, you're, you're building that pragmatic foundation that can allow you to adopt more and more sophisticated technology.

Host: Right. And it's so important to, to be, uh, willing to put that mirror and, and start the discovery process.

Suzanne Valentine: Right.

Host: I think that's sometimes the hardest part. But you need to start discovery, you need to start understanding where you stand today. What are the limitations even if you don't have a clear picture of where you want to be in the next months?

Suzanne Valentine: Yeah, a lot of this is no regrets work, right? Regardless, uh, regardless of what happens in the market, if your business strategy changes, technology evolves, there's some of these things that are just so foundational and no regrets that you should, you know, put that front and center and then start experimenting with some of the different technologies that are available today, but not feel like this is it, this is the technology we're going to be relying on for the next five years or whatever. Recognize that it's going to be an evolution and if you're working with a software vendor like Price fx, we're helping you with that evolution. We're constantly keeping abreast of advances in the market and figuring out ways we can be bringing better, newer technology and features to our products so that customers don't have to do all of that investigation themselves.

Host: All right, so we've talked a little bit about some of the challenges, some of the work to be done pre implementation. We talked about data, some of the mindset when it comes to challenges, and also trade offs that we might have to make. So you know those difficult decisions that we have to sometimes make as we go through the implementation process. Can you share anything to help our listeners prepare for what that might be like? So what are some of the practical problems or practical challenges that appear that might require a certain uncomfortable trade off that they need to start thinking about?

Suzanne Valentine: In terms of the prerequisites, I could organize these into like four buckets. So there's data which we've talked about some people process and strategy. We've talked about how data is absolutely foundational. So I won't go more on that. Uh, in terms of people and mindset, this is huge. So you need executive sponsorship, not just, yeah, go ahead and implement pricing software, but you know, genuine buy in from the C suite that pricing is strategic and worth investing in. When things get hard, and they will, you'll need executives who are going to push through the resistance in their organization rather than retreat. You also need the right team. So at minimum you need somebody who owns pricing strategy, someone who understands the data and can be a bridge to the analytics, and someone who can manage change across sales, finance and operations. So if you're expecting your IT team to drive this alone or your finance team to do this as a side project, then you're probably setting yourself up for failure. Then there's the mindset. So this might sound kind of soft, but it's critical. Is your organization ready to let data and systems guide decisions or is there just so much gut feel culture that people are going to override the system whenever they disagree with it? If it's the latter, then you need to work on that culture shift as part of the um implementation, not assume that you'll do it after. So that's people and mindset. In terms of having clear commercial objectives, you've got to figure out what you're actually trying to achieve. Is it margin improvement, revenue growth, better price realization, faster quote, turnaround, reducing, discounting? We've seen companies who make the mistake of saying you know, we want to optimize pricing without really defining what success looks like and that's too big. You need to be specific and have measurable goals like improve margin by 100 to 150 basis points over 12 months or reduce pricing errors by 75% or decrease correct quote cash cycle time by 30%. These objectives should really guide where you focus your implementation. You know, if market competitiveness is the goal, maybe you start with firming up competitive data and instituting some basic rules based pricing as we've talked about and layering list price optimization in. But if sales effectiveness is the goal, then maybe cleaning up your pricing corridors with negotiation guidance analytics is your, is your entry point. The commercial technologies obviously can do a lot of things, but you need to be disciplined as an organization about priorities. And then I'd say the fourth, the fourth key thing is just process. You know, you need some basic pricing processes in place before you implement software. Like who approves uh, pricing changes? How do you handle exceptions? What's your price update cadence? How do you communicate price changes to customers? Software will make these processes more efficient and scalable, especially with the integration of agents. But software won't magically create processes where none exist. So if your current pricing process is the sales rep, ask their manager who asks someone in finance who eventually responds with a number. Then you need to fix that workflow before you, before you automate it. So those are the prerequisites. I think you also asked about trade offs, so I have, I have a couple of thoughts there. You know, every implementation requires making choices and understanding trade offs upfront helps you make better decisions. So you know, a uh, key one is flexibility versus simplicity. Price of X is incredibly flexible and modular. You can configure it to match almost any pricing logic. But that flexibility comes with complexity. The more customization you do, the more you need to maintain and the steeper the learning curve is going to be for your users. So my advice would be to start simple. You know, use out of the box functionality as much as possible in your initial implementation. And yeah, you might need to adjust some of your business processes to leverage the software rather than configuring the software to fit every current process. But for a lot of companies, a, uh, reexamination and reinvention of pricing processes can be really healthy. And then if you learn the platform and identify specific gaps, you can build some targeted customizations. But don't over engineer the whole thing on day one and insist that we're going to make the software do exactly what we do today. So that's the first one. Flexibility versus simplicity. I think there's a trade off between speed and perfection. So you can spend six months or more designing the perfect implementation that covers every edge case, or you can get something valuable into production in weeks and iterate from there. And I'm a big believer, as we've talked about, in starting small, learning quickly, picking a pilot scope, maybe one product category or region or sales channel where you can get to production fast, improve the value and learn, and then you can expand based on what you've learned. The companies that we've seen be the most successful are the ones that embrace iteration, you know, which we talked about a few minutes ago. They know the first version won't be perfect and that's okay. They'd rather have 80% of the value in three months versus 100% of the value in two years. So that's another trade off. Um, sophistication versus adoption I think is another one. So this ties back into our conversation around having something be a black box, uh, that you might have access to state of the art optimization algorithms and advanced AI, but if your users don't trust it and understand it, they probably won't use it. So you need to balance that technical sophistication with explainability and user experience. And then I think the last one is around standardization across your entire business versus local autonomy. I think, you know, with, with global companies in particular, uh, they have to decide whether to implement one global pricing approach or to allow for regional variations. And there's no universal right answer here. Standardization is going to give you consistency and easier governance and economies of scale. But you know, local autonomy allows you to adapt to regional market conditions and regulatory requirements and competitive dynamics. So what we typically recommend here is some sort of constrained flexibility where you define the core pricing principles and guardrails globally, like how you segment customers, your pricing objectives, your approval thresholds, but then allow regions to implement those principles in a way that makes sense for the individual markets. The software can handle this, but you need to think through the governance upfront. Right.

Host: That might sound like a lot, but also something we've promised is to talk about not just the potential of the technology and all the great things that it can do, but also prepare our listeners and our clients for what's coming. Having all this experience with implementation, uh, of pricing projects, some of the trade offs, some of the decisions that they'll have to make. And if all of that sounds very daunting, consulting companies like ours help clients navigate that help Them prepare for a conversation with Price effects. Right. For what is that moment when we can start having that conversation about software? Suzy, sometimes we have clients, usually in the pricing team or the sales team or product management. They reach out to us and say, I know we have to do this, I know this is important, but I need to get the organization to see it as well. I need to get my execs to see it. I need to get by my key stakeholders. And by the way, we also help them do that. And that's something that happens very often where everything that we've discussed here maybe speaks to two or three individuals in a small pricing team, but maybe doesn't speak to the whole organization. So we help make that much more. Yeah, translate that in a language that the whole organization can understand. So if, if, if someone out there is thinking that, don't worry, you're not alone. We know exactly, you know the, how that feels and we can help you with that.

Suzanne Valentine: Good.

Host: So this has been so incredible, so stimulating, such a great, refreshing conversation to have at the beginning of the year. And as we close, I wonder if there's any sort of highlights or if there's. We want to leave our listeners with two or three key messages, two or three key takeaways, including on how, uh, price Effects can help them. What would you highlight? What would you say?

Suzanne Valentine: How would I wrap this up? Yeah. So here's what I would want listeners to remember and think about. So pricing software such as Price Effects should be flexible enough to meet you where you are. Whether that's getting out of Excel and creating your MVP data foundation, or adding optimization and predictive analytics to mature pricing operations, agents can be a part of the entire process. So starting with agents can help you identify where the biggest value can be achieved. And then you can continue to add agents to identify and prioritize opportunities, monitor compliance with the, with the recommended prices, and automate a lot of the decisions that don't require human oversight. Transparency. We talked a lot about transparency. It's crucial in pricing systems so your teams trust the system and are comfortable to delegate more of the decisions to agents. Flexibility is also a key consideration. So both the ability to add additional capabilities as you curate more pricing data and to harmonize existing algorithms with that data. Price Effects, with its composable pricing AI and bring your own science, is a way that this can work in practice. And then finally, don't feel like you have to do this alone. There are a wide array of pricing specialists like Pricing Works who can help with every aspect of your pricing journey and can meet a variety of organizational budgets. So super valuable to start with an honest assessment of your situation and design the right journey and oftentimes it's great to have a third party helping with uh that.

Host: Susie, thank you very much. Thanks for making time for us.

Suzanne Valentine: We're hugely grateful.

Host: We'd love having you on and we will leave our contact details in the show notes so that if our listeners would like to reach out to you or us to follow up on anything that we've discussed here, then they can easily do so.

Suzanne Valentine: Great. Thanks a lot. Thank you.

Narrator: Many thanks as well to you our listeners for tuning in. If you've enjoyed this episode, please subscribe to the Sweet spot by pricing works on your favorite podcast app. If you have any questions or feedback for us, we'd love to hear from you. So please go to pricingworks IO and get in touch.

Host: See you next time.