PriceCheck Podcast: AI, Agents, and the Future of Dynamic Pricing

Dr. Fabian Uhrich
Hey, I’m Fabian Urich, CPO of Quicklizard, and today I want to talk about AI and agents and what they have to do with pricing. For that reason, I’ve invited Dr. Markus Husemann Kopetzky to exchange some thoughts and reflections. Markus.

Dr. Markus Husemann-Kopetzky
Thank you so much, Fabian, for inviting me and I look forward to a very interesting conversation.

Dr. Fabian Uhrich
Amazing. Pricing is one of the business disciplines that leverages big data and statistical methods for the longest time. Most of the signals that explain demand, I would call them the explanatory variables to keep it simple, are already available as structured data like competitor prices, purchase prices, inventory levels, psychological price points, weather data, all that is codified as numbers. Therefore, big

Big data as a tool is very common for pricing. And typical methods are already being used since quite a long time, like logistic regression, random forest regression, just to name a few. And the these tools obviously work well in explaining historic price demand relationships. And they can also be used to explain and predict future demand reaction to price changes.

Now the question is, what will be the impact and on of AI and agents on all these methods? How do you see this, Markus

Dr. Markus Husemann-Kopetzky
It’s a very interesting question. Thank you for raising this. When I look at AI and agents and pricing, I look through the lens of a practitioner and the question is, what is in it for me? And I see AI as the best of two toolboxes. In the first toolbox is, as you described it, the statistical learning, the machine learning.

I know machine learning is not so super sexy because the term is around for some time. But this part of AI is super helpful to optimize for complex problems. In essence, when you look at one of the more famous machine learning algorithms like neural networks, the neural network mimics the brain. And that’s why we always talk about

artificial intelligence based on neural networks when we linked artificial intelligence to machine learning many years ago. So you have the statistical toolbox. And now we have something more. We have on top large language models. Large language models make sense out of all the data and the results and bring those to the attention of the pricing manager. So we don’t only have the…

analysis here, but we also have a mechanism to make sense out of this. So if I had to explain this to my mom, I would say the machine learning is the engine, but we don’t only have the engine, the machine room for pricing, we also have on top of this agents. And the agents are not only bring

these analysis to your attention, can also within degrees of freedom can act on these insights and do decisions for them. So if I would further explain also to my mom what agents I would say AI is the toolbox, agents are the workers and the pricing manager is still the boss.

Dr. Fabian Uhrich
I I definitely like your analogy, Markus. With AI that toolbox get just more powerful, better models can be chosen to explain the price-demand relationship, and of course we also see on our end accuracy increases of the predictions and forecasting models. However, this to me is still incremental improvements, given the fact that most of the explanatory variables in pricing are already available as numbers.

The bigger disruptions I see usually in areas and domains where images and patterns of text are the kind of the bottleneck to even work with algorithms. And therefore, in pricing, this is not the problem or not the challenge. Of course, adding using images for similar matches, interpreting text on social media and the latest search trends will help.

To be faster in the reaction to anticipate these trends also with pricing. But I think it’s not as pivotal as in other disciplines when I’m thinking about, I don’t know, installation work of a machinery. And with virtual glasses you can see the machine live, and only thanks to the images you see, the algorithm can recommend you what to do on that machine. For pricing, I see less of that disruption.

For me, I guess the the agents are the more pivotal part in pricing. How do you see this?

Dr. Markus Husemann-Kopetzky (05:14)
I totally agree that in principle, the data has been around for some time, even multimodal data, and that the technology, like the techniques, methodologies are also around there for a long time. However, what I see is that the application of these methodologies to make sense out of these data required a very specialized skill set in the past. And with large language models, we have a

created the capability that also less skilled people, so those that don’t have a doctorate degree in mathematics can apply those. So capabilities are not new. The opportunities are not new. But what is new that more companies can make use of this. And I totally agree with you that agents at like

steroids to this development. What do you think?

Dr. Fabian Uhrich
Totally.

To stay in in in your analogy, the agents are kind of the consultants that enable the boss or the advisors of the boss to take better decisions, to leverage pricing in a better way. They can help explain observe behavior, explain price recommendations in natural language, with a lot of reasoning around, but they can also challenge the boss.

maybe actually even better than a human would ever do because an agent is not scared of a demotion in his career ladder. So an agent can be totally blunt and tell the pricing manager what strategies might be set in the wrong way, what configuration parameters don’t really make sense. And also they can take over complete tasks like repetitive things, but also more strategic elements about setting the right goals for the right articles.

So AI actually can help to set the goals, achieve them, and agents can even challenge and let’s say change the goal also over the course of time. So for example, if the pricing manager would have said we optimize for profit for these items, the agent could tell him, No, actually, these should rather go for revenue maximization, but there’s even a few products in that segment that are wrongly classified, and you need to reclassify them. So that is

The level of change and benefit I I’m expecting we are seeing from the agents.

Dr. Markus Husemann-Kopetzky
How can I not agree? The agents will become the wingman or the worker for the pricing manager who still remains in the driver’s seat. And the agent will take over repetitive tasks. They will take over the number crunching, creating insights. But no, they also not only make their own decisions, but… they uncover those decisions that need to be made. So they bring these decisions to the attention of the pricing manager. And these are completely two different fields. The first one is about efficiency, doing the same with less time or doing more with the same time. The other one is being more effective. You just make better decisions than before.

Dr. Fabian Uhrich
I think this makes a lot of sense. and I I also see this and if if we look at our software, AI is improving efficiency on the one hand, choosing the right data science model for demand prediction, picking up new and additional explanatory variables like social media trends or image recognition. this is making the whole process of creating models just much more streamlined and faster.

but also when I think about the interaction with our pricing tool. Like you don’t have to click on different buttons to set a filter on the list of items. Instead, you can just type in natural language what you want to focus on, and the filters will be set automatically by the A by the AI features. Or asking about certain recommendations, getting explanations, all that can happen in natural language, and it’s just making the workflow faster, smoother, and more efficient, generating a report.

Just type in what you want to see, and the report comes out of the system. But the bigger and more interesting part, exactly, as you said, is also the effectiveness. So doing the wrong things faster doesn’t make the wrong things better. But being more effective, and this is really where AI is getting us to a step change. Challenging the strategy, reconsidering how products are grouped and clustered together, challenging actually the business logic behind.

This is really what can increase a step change in in the outcome. so to me, this is really the the big benefit. And to make it more tangible, like back in the days when I was responsible for the pricing at Zooplus, I used a very simple metrics where I was looking at growth and margin development. And there’s very obvious cases. you’re growing well, margin is shrinking, well, likely price decreases, fueled growth, amazing. if that’s a goal, keep it.

And you would also I also found cases where I was shrinking, but prices were also decreasing. So something is wrong here. Despite price decreases, growth did not come up. So that’s an area where I knew I need to dive in and I need to dig deeper. Now with Quicklizard Lizard, we are offering much more than just growth and margin development. There are many other KPIs. For example, looking at the average seasonality share in the different categories.

Understanding the out of stock situation as a potential explanation for bad growth. And maybe it was not pricing, was just the stock is not sufficient. Competitor sensitivity, understanding which categories are more sensitive to competitor price changes than others. The average price elasticity. Seasonality share. Are these items on promotion currently? Is there currently a seasonal peak? This easily gets up to seven, eight, nine different metrics. It’s not a two by two metrics anymore, unfortunately. It’s like a multidimensional space and it’s

Almost impossible, even for smart people like me, to derive the right conclusions out of all these different combinations. And this is really where the power of Agentic AI comes in and can come up with really precise recommendations and implications by interpreting all these different metrics at the same time and guiding the pricing manager where to dive deeper, what could be the hypothesis for change and improvement. This is really where I see the biggest value in our software from applying AI. Now also wanna hear like a more broader outside view, Markus as a renown pricing evangelist, where do you see the end game of AI in pricing?

Dr. Markus Husemann-Kopetzky
That’s also a very good question. I don’t want to get too philosophical here, but what I see in as the end game of pricing is the following. I strongly believe that pricing will become very efficient and all the decisions or input variables, all the insights that you just mentioned what to consider the contribute to this efficiency. So at the end, every customer

will pay in accordance to the value she or he receives. I’m not talking about personalized pricing. I’m just talking about pricing in general. So we have the right price given the value received. And this is good news because the algorithm will take care about the optimal price and the company takes care about the value delivered to the customer. And that is actually what’s at the core of

every business model. How do you create value for your customers that exceeds your cost of delivering those? And the ratio is ideally better than the second best offer. So we will focus more on adding value to customers and the algorithm that pick up the value difference that you create and adjust pricing accordingly. So pricing, I don’t want…

I don’t want to say that it will become a commodity. I just want to say this will become super crucial because if your pricing engine in general as a company is not efficient, you will likely overprice or underprice. So you cannot afford having an inefficient price if everybody else has an efficient price. That’s the end game that I think. What’s your view?

Dr. Fabian Uhrich
NowI can give the same answer as you did before. It’s hard to disagree to the statement. hundred percent agree. the future of AI in pricing, agents in pricing, they will free up time, they will support, give guidance and make sure all thinking of the humans is focused on value and how to get there is handled by by the software, the statistics, the AI and the agents.

And this is really I think where we are heading at. The only question is how we are getting there. That’s a that’s the next one.

Dr. Markus Husemann-Kopetzky
Yeah, that’s also a very interesting question. The core of the question is how do I create this capability of doing sophisticated professional pricing? And there are in essence two ways. If you are a very large e-commerce player with billions of revenue and it happened that you got founded in Seattle,

or at least have your headquarter there. You can invest heavily in these specialized skills to even optimize for very miniscule, very minute pricing problems because given the sheer size of your business, it always pays off because you just can scale your solution. But if you are not that big, you don’t have this scale and you’re not a special snowflake.

You have the decision to still build it yourself or buy it externally. And this special sophisticated pricing skills are very niche. So it’s definitely an investment in something that you cannot have beforehand. So you need to invest actually in this. The question is, how can you invest efficiently? And by my experience in this make or buy decision, the…

buy decision is usually the dominant strategy. Because if you look at the make decision, a software development team who has never seen a pricing software needs to build a pricing software. You need to build it, you need to maintain it, and you need to further develop it. What is your advantage here? You might think, okay, I have a kind of a secret pricing source that nobody else has, maybe, but then you just need to wait to apply your secret pricing source in a way that no other can.

But what is this technical backbone? This could be the same. If you buy the whole backbone somewhere else, the vendor who sets this pricing software, they spread their development costs across all the customer base, the maintenance costs across all the customer base. And they have a very specialized team of software developers who just

do pricing software, so their team is also usually better in this special use case. And I know the example sounds a bit off, but that’s also the reason why no company programs their word processing software like Word, and nobody builds their ERP system like SAP or Oracle. So there’s no reason not to start with standard software. And at one point, once you have…

a standard software, your company learned, then you can decide, okay, for this is a customized software, but still you have 5 % that requires individual development. Is it worth for these additional 5 percentage points to develop everything on your own? And you will see most likely not because it’s not only about development costs, but it’s also about opportunity costs. Your SDEs, even with vibe coding, are a limited resource. And the cost of not developing

what would have been developed instead of the pricing software is usually not worthwhile doing it on your own. That’s my point.

Dr. Fabian Uhrich
How could I not agree to that? I also as we used to frame it, it’s not a question of make or buy in our view. It’s a a question of buy and and build. So you can still put your ingredients and your magic sauce into the software and re use it to reflect your own very specific and unique pricing strategy and have all the agents and all the AI help you challenge it and support you implement it. So

I think we are aligned on this one at least. And very curious to see how this future of AI now really turns out over the coming months. I will I explicitly not say coming years, but coming months.

Dr. Markus Husemann-Kopetzky
Yeah, it’s just at the beginning. This is going to be a very good time for everyone, in particular for the companies. This will also accelerate the pricing career, by the way, within the company. These new tools will show the companies that pricing managers have very powerful tools at their hand. And the business case of adding more AI capabilities are

most likely to happen also in pricing because the return on investment pricing is hugely huge and with AI, it’s non-negotiable.

Dr. Fabian Uhrich
Perfect, Markus Thank you so much for taking the time. It was a pleasure talking to you and hope to see you again in another sequence of of such a conversation in the future.

Dr. Markus Husemann-Kopetzky
Thank you so much. It was a pleasure being on your show.