Artificial Intelligence is no longer just a buzzword for big tech; it is fundamentally changing how merchants—specifically in the re-commerce space—manage their day-to-day operations. In this discussion, we explore the transition from manual inventory management to a future driven by vision capabilities and autonomous agents.
The Power of Vision in Re-Commerce
One of the most immediate impacts of AI for sellers is in inventory registration. For secondhand stores and re-commerce platforms, cataloging unique items is time-consuming.
- Instant Grading: General Large Language Models (LLMs) equipped with vision capabilities can analyze a photo, identify the item, and assign a condition grade automatically.
- Data Enrichment: AI allows small store owners to create rich, structured data schemas simply by snapping a picture, making their inventory discoverable to search crawlers and AI agents alike.
Democratizing Data and Forecasting
Historically, complex demand forecasting and dynamic pricing were reserved for enterprise giants using tools like Relex. Today, AI is making these capabilities accessible to smaller merchants.
- Human-Readable Insights: Instead of complex spreadsheets, AI agents can provide conversational advice (e.g., "Double your stock of this jean model for Black Friday").
- Actionable Intelligence: The goal is to bridge the gap for merchants who may lack data science resources but need high-level optimization.
The Road to Autonomy: From Recommendation to Action
The conversation concludes with a look at the technical future of AI Agents. We are moving from a "human-in-the-loop" model—where the merchant approves every suggestion—to full automation. By utilizing Model Context Protocols (MCP) and robust API layers, agents will soon have the reasoning capabilities to trigger price changes, create bundles, and manage catalogs autonomously once trust is established.
Karri: We are talking about AI and how it's going to affect both commerce in general and especially re-commerce and we're gonna go through both buyer and seller or the merchant side. What about the seller or the merchant side?
Tuomo: For example, inventory registration not only enriching the content but actually specifically for re-commerce where we talked about that where vision capabilities can help a lot.
If you're handling product returns or maybe your whole business is a secondhand store, so if you register in the inventory it can be as easy as just snapping a picture of it. So even if you wouldn't have like a brand catalog against which that image would be analyzed, the general language models are pretty good at identifying what is this item and giving it a condition grading that can be used in your own inventory. So item registration is as simple as snapping a photo of that item as it comes in.
So that's to me like again it's kind of generative but there's also the element of vision capabilities there that allows the prompt to be born almost like, "yes please register with this data schema something for this image that we have here." So I think that's definitely a core capability that even small store owners should have better data, more enriched data and they can create that data more easily. And then if the technology partner that they're working with is on par with the modern standards, then that data is discoverable and indexed by AI agents in the same way as websites are indexed and crawled by you know Google crawlers.
Karri: How is AI then gonna influence like the components of this merchant world where you are for example working with inventory or pricing the items and kind of maybe not optimizing but making sure that the operations are working properly and why not of course optimize the business for you as a seller?
Tuomo: For example, in inventory, we could talk about like forecasting demand forecasting and supply forecasting or dynamic pricing like based on weather or whatever.
What's AI and what's just like machine learning and data models? Like now we tend to throw the term AI on top of everything. Everything intelligent. Okay, it's AI. So there have been you know companies like Relex and I know there's a lot of dynamic pricing companies like Pryce and similar that already before AI hyped it like a lot of data-driven optimization of dynamic pricing and demand forecasting and so on.
Maybe the role of AI there might be that for less professional merchants or smaller operations that do not have the resources or know-how or time to do like very complex data analysis on forecasting. These commerce platforms can embed these forecasting capabilities but deliver the results in a human-readable format kind of AI agent-driven relationship which might be more understandable for a less data-driven user.
Karri: So providing like recommendations and tips and tricks that you might want to implement.
Tuomo: Exactly, so rather than giving like a huge excel and whatnot maybe it can give a more simple that "hey you should probably double your I don't know this jean model for Black Friday because seems to be the one that is now hyped." I think it could be simple stuff where you know that the capabilities are the same but the way how it's delivered and packaged with an AI agent makes it more accessible for the merchants that have less resources to kind of analyze complex stuff.
Karri: Do you see that this is also like a journey that starts by AI kind of giving you these recommendations that you end up implementing yourself? And maybe once you start trusting it or just in some time when you are kind of overall maybe trusting the AI then all of this stuff can happen in the background automatically?
Tuomo: Yeah I think that's an interesting aspect and probably less visible of the merchant but it's an interesting gigi topic to discuss is like what does like agent automation look like under the hood. So if we've gotten used to kind of the idea that we do automated workflows where I do like we have even these visual drag and drop tools almost like Zapier or similar where "all right if this happens then do this."
Now the reasoning capabilities of many of the AI providers, the agent providers, are pretty good or even great in some cases. So instead of you having to map out all of these workflows you could hire a kind of AI agent to do the reasoning. "What should they do?" Then they just need access to these various configuration endpoints and that's kind of the role again where the commerce platforms would need to make sure that like all of the APIs to actually trigger things—like change the price or create a new catalog item or new bundle product for this campaign—that all of those actions are available for those reasoning agents.
That probably means having MCP layers or Model Context Protocols and so on on top of your API so that the AI agent can kind of make sense of what's available for them to play with. So that's kind of for the merchant like at the end of the day it is more of like you described that "all right what do you recommend," the reasoning part of the thing. And then they would need to kind of be able to access the data to do the reasoning well and then it would say that "and I would create a bundle of these two products with this price."
And then if it seems like a good idea for you you can actually do it yourself and then if it constantly delivers good results then you might say that "hey well why do I need to be here in between let's just like do your recommendations automatically like auto play." But under the hood, the difference is does the agent have access only to the data or does it also have access to kind of the service layer where they can trigger these?