AI in DAM for retail – from trends to tangible product

Welcome, everyone, and thank you for joining us. Today, we have a few interesting angles to cover in our webinar topic, AI and DAM for retail: from trends to tangible impact.

My name is Magdalena, and I'm the product marketing manager at WoodWing. I will be the moderator of the webinar today. Before we begin, I'd like to mention a couple of housekeeping rules. The session is recorded, and we will share the replay with everyone who registered for this webinar who couldn't make it, and of course, everyone who is joining us live today.

Throughout this session, you're welcome to ask any questions and submit them in the Q&A panel on the side. With the speakers, we have reserved some time later on to answer those questions.

Now let's introduce the speakers. We have Paul Walker, our product development manager at WoodWing, and we have Kristina Huddart, content operations and DAM specialist at Huddart Consulting and one of the leading voices of the DAM industry. Together, they'll guide you through the opportunities of AI in DAM for the retail industry.

Let's briefly go over the agenda for today. We'll set the scene first with Kristina and her expert insights from her State of AI in DAM 2025 research, where she uncovers how retailers are adopting AI and where they're seeing the most impact and potential.

Then we will shift gears from trends to real-world examples. This is where my colleague Paul is going to show what he and his team are working on directly with retail organizations that are using AI in their daily content workflows, from tagging product images and visual searches to accelerating content delivery. He will focus on those real-world use cases and show AI in action.

After these real-world use cases, we will reach our final segment, which is the implementation dos and don'ts. Here, our speakers are going to share their expertise and advice in a conversational format on what to do and what to absolutely avoid when embedding AI into your content operations.

In the meantime, I'd like to start a quick poll for everybody joining us today. That's just a brief information and feedback from you on where you are on your AI journey. That will feed our speakers with material to know where to concentrate and where to focus within their presentation and use case examples.

Without further ado, I'd like to ask Kristina to come on this virtual stage and share her research insights. Thank you for joining, Kristina.

Hi, Magdalena. Hi, everyone. Thank you so much for being here, and it's so nice to be here with you today. I really appreciate you all joining.

As you know, many of you are joining us from retailers. Retail is moving at a very fast pace right now. It's probably the fastest thing we've seen since e-commerce took off many years ago. The good news is that retailers are not standing still.

As we're seeing in your poll, as you're answering questions here, many of you are already using AI and looking at leveraging many of these tools. So let's take a quick look at what's accelerating in retail and why it's accelerating.

AI is exploding in retail because we see three different forces that are converging. We've got consumers who are expecting and demanding personalization now. Anytime we go shopping online, I think all of us do this. We're expecting that personalized experience. I think a few years ago, it might have been a little bit creepy, but now we've become used to it. We also realize that it's helping us get better recommendations.

Sixty-nine percent of us report that we like personalized recommendations. It's making our shopping easier and quicker, and we're getting better products for it. So it's the consumers who are demanding that we personalize at scale, and this is one of the reasons that AI is accelerating.

Then we have the pressure from staff. As those of us who are working in retail and creating content, launching campaigns, and going to market with new products know, we rely now on AI to help us do more faster.

A recent study from Nvidia said that nine out of ten companies are already using AI in their operations. We're seeing this as you're answering the question in the poll. Seventy-one percent of us say that we need AI to help us launch campaigns faster. A study by Cornell University in the US said that we can become 1.6 times more productive in our content creation workflows by using AI.

So there's a lot of data behind this, and there's a lot of reason why we are so quickly adopting AI.

The third piece of this puzzle is the tech partners. Our tech partners are building AI into every single platform that we use in our workflows. I just want to say a huge thank you to our SaaS vendors like WoodWing, who are developing our core technologies so quickly. The releases are coming hot and fast, and being able to have AI in our core systems is an amazing way for us to leverage AI.

We're seeing this across all of our systems, whether it's CMS, CDP, CRM, but of course also the DAM. So we see these three forces converging. We have demand from our customers, demand for AI from staff, and provisioning of AI from our tech partners. That's why we're moving really, really fast in retail with AI.

Now the question is, how is AI already being used in the retail space, and what is working well?

Hopefully, this is not a surprise to any of you. Let's start with digital twins. If you haven't heard about digital twins, absolutely go check out some amazing work that's being shared by all sorts of different companies who are using this. This is, I think, one of the most exciting shifts that we are making because of AI.

If you don't know, digital twins is the ability for retailers to create 3D renditions or mockups of our products. They call it twins because it's almost pixel perfect. It's like having the product in your hand. You can turn it in all directions, put it in a scene, and use these digital twins not only for campaign creation and packaging previews, but also for e-commerce photoshoots.

The brilliant part about this is that, because we can do this with AI at scale, we can now do it over hundreds and thousands of products. We could do this before AI as well, but now we can actually do it at scale.

This means that we can now start our campaign production much earlier. As marketers, we're not waiting for the factory to send us the products, waiting for them to be shipped across the world potentially, or waiting for the right photo scene or setup. We can do this all virtually using our digital twins.

It's saving us a lot of time, money, and travel as well. For a lot of my clients, and also brands with many SKUs like Nestlé, Coca-Cola, Unilever, Moët Hennessy, and Estée Lauder, the ability to have digital twins and use AI to create digital twins very quickly is removing weeks from our timelines.

I would say it's pretty green, right? It's a bit more sustainable than flying around the world to do shoots and to ship products. Although there is obviously quite a bit of environmental impact from AI as well, it's definitely more efficient, and it gives brands that head start before the product even exists. So if it's a TikTok trend today, we can start the campaign immediately.

A few other things that we're seeing happening in the retail space with AI are that we see e-commerce companies using AI for brand assets or checks. This is where we can now train our AI large language models on our brand specifically. That's your brand voice, your visual identity, your brand guidelines, basically your brand DNA.

This is what Unilever is doing. There's a great example from Unilever that they've shared widely, so go look it up. They've basically created a separate mini DAM system called DNAI, where they hold all of their absolutely perfect, verified, on-brand assets, as well as all of the metadata, the high-quality metadata that goes with those assets. They only train their AI tools on this DNA subset of assets.

We want to make sure we're not training AI with any data that's not on brand, completely finalized, and 100% correct with high-quality metadata. They're able to do this by putting it into a separate training subset.

Another thing that we can do is localization at scale. This is where AI really gets its superpower for retail. Retailers now want Amazon-specific crops. You want an Aldi-specific layout, a Carrefour positioning of the product on the image, regional compliance copy, different seasonal colors. There are so many variations that we need. This list is endless.

With AI, we can now scale much quicker than we ever could before. We can create hundreds, if not thousands, of retail-ready variations, which was just too time-consuming to do before. Previously, we may have put the same product pack shot onto all of our retail sites, but now we can really specialize and localize at scale.

Finally, I think this is kind of an under-the-radar use of AI. It's not as glitzy and glamorous as the generative AI production is, but AI for decision-making is, I think, the secret superpower of AI. This is allowing us to use AI to mine our big data.

Now that we're doing localization at scale, that means we have ten or a hundred times the amount of content that we had before. That's a lot more data coming in. How do we get the right insights from that data to really understand which of our marketing assets is performing well, which is getting the most engagement, and which is most effective, so that we can feed that information and those insights back into the beginning of the process and go back to strategic planning for next year, let's say, which many of you may be doing at this moment?

AI is allowing us to do this with big data, and we can now answer questions that we could never answer before. Which hero images are converting the best? Does yellow perform better in summer? Should we lead our retail page with a lifestyle shot or a pure pack shot for this retailer? AI is able to help us do that now.

These exciting AI-driven innovations are things we're all probably itching to do, if we're not doing them already. These innovations really only happen when the foundational technologies in our content tech stacks are actually mature enough to support them.

I want you to think about whether you could do any of these things with AI today, or whether you would need to reconsider if you have all the right tech in place, and whether you have the right data and brand assets in place to do any of these things.

Let's zoom in now on the AI adoption through the lens of what I think of as the heartbeat of our content: our DAM systems. This is how we organize, manage, structure, and govern content. This is a really clear indicator of how ready you are as an organization to actually leverage AI and use it properly.

The DAM is a great place to start with AI. What we're seeing in the market in 2025 is that AI adoption in DAM isn't really happening very evenly across all industries, but it's following this really familiar adoption curve that we see for all technologies and all changes.

On the right-hand side, you've got the companies that are adopting AI quickly, and you can see retailers and e-commerce right there in the middle. Alongside, we've got creative agencies, media and entertainment, and technology companies, which, as you might expect, are adopting very, very quickly.

This makes sense. These are the industries that have high volumes of content, short deadlines, and they need to deliver great customer experiences. Just a note: this data comes directly from my annual research initiative, The State of AI in DAM and Content Operations. I'll give you a download link in a moment.

One of the questions that we researched was: if AI could do anything for you in your DAM and how you manage content, what would you want it to do? This is what organizations like yourselves told us. These are the items on your AI and DAM 2025 wish list, a bit like your Christmas wish list, if only there were a content Santa who could help us bring all of our wishes come true this year.

Firstly, you've got automating tagging and metadata management, followed very closely by improved search and discovery. There's no surprise here. We all want to be able to tag our content more effectively and quicker. This is often a bottleneck in our processes, and we want to make sure it's not error-prone.

We also want to be able to find our content much quicker and better. Now with AI and computer vision, we can use natural language search and processing to really ask your DAM, “Help me find assets or that campaign that we did on that tropical island with the product in the blue bottle.” This was much harder to do before AI and natural language search. So love that for our DAM systems.

We also want to have automated workflows and resource management. Think about automated routing to the right person, AI first rounds of approval where AI checks whether the asset is on brand before it goes for final sign-off, and smart notifications and support to reduce all of the bottlenecks that we naturally have in many of our workflows.

Also in the top five of your wish list for this year is measuring content effectiveness and getting predictive analytics. Let's use AI and all of this data that we have to really predict which products are going to fly off the shelves and which creative campaigns are going to give us the best return on investment.

We also want to use AI for security compliance and reporting automation. This is where the biggest risks are, and it's also quite difficult. We know that rights management and tracking are incredibly difficult to do, especially manually now that we have so much content.

These are often repetitive and mundane tasks that AI is great at. We can send AI out to do our tracking, checking, and reporting, and then we, as humans in the loop, can manage what we do with that information.

I'm happy to say that Paul is here to be your content Santa today and to share how WoodWing is making this wish list come true for you. But before I hand over to Paul, I just wanted to share the research from this year, The State of AI in DAM and Content Operations 2025 report.

If you haven't seen it yet, do come and download the research. For all of you who download the report from this year, you will automatically get a notification or an invitation to be a contributor for next year. I'm going to be opening up contributions for the research in January. Please come and share your experience with AI, whether it's good or bad. It doesn't matter. We want to hear all aspects of it and get involved with the research.

Thank you all. I'm just going to pass over to Paul now.

Thank you, Kristina. That was some great insight and really good points on digital twinning. What I'm going to cover today is not digital twinning. It's a fantastic area to explore, and we've been helping retailers on that for many, many years now. I think it's worthy of a webinar all in its own right.

What I'm going to talk about today is AI in retail, and specifically the use of commissioned photography and how you can help shorten time to market, reduce that time, and make better choices. I'm really going to be leaning into some of the things that Kristina has already touched on, and I'll highlight those as we go.

First of all, let's look at the very typical retail asset workflow. I think most of you are probably familiar with this. We have ingest, enrichment, discovery, prepare, and apply.

First of all, ingest: the state of acquiring material from either commissioned works or your external partners. Enriching: AI lives and dies by data, so AI is not going to take that away. How you do that application and validation of data, the more you can do that with accuracy, the better your downstream results will be. Again, we'll be touching on those points.

Discover: this is where it's great having all of these assets. We deal with clients who have millions upon millions of images, and daily, they are acquiring more and processing them. If you can't find them, then there's a problem. Again, AI can help with this matter.

Preparing: this is one of the key parts that can burn a lot of time. Whatever image you get in, you have to slice and dice it and prepare it for consumption within your organization, and then apply it out the door to your many channels. It could be social media, your website, print, but also the application is for your other partners. You might be dealing with some international third-party partners who also want to utilize your material but want to lean on you to help prepare it.

What that means is there is a lot of pressure on your workflow at every step of the way. What you're trying to do with the people that you have: you could bring more people in. You have training to do, and bandwidth can be the problem.

Shortening that time to market as soon as you get it in is especially critical for retail. It doesn't matter how many boxes of the product you have in the warehouse. If the picture is not on the website, you can't sell it.

This is the place where, as soon as you get an image through, processing that and getting it out the door in your house style as quickly as possible can mean the difference of a lot of money on the bottom line.

What we're going to do is look at three core pieces today. The first one is enrichment. When we're looking at enrichment as images come in, it's adding that valuable metadata. The first one is just using simple tagging.

In a minute, I'll be showing you a little video that gives you a taster of tagging. This is where you can pass your image to one of many AI services out there that will look at your image, interpret what's there, and give you back some of that information.

I think we've all heard the term hallucination. Sometimes we have to be careful. We need to validate that information. Equally, we might be given a photograph by an external party or the photographer, and they say, “Yes, I've put the file name in there, and this says that it's a coat or a jacket,” when in actual fact, it might be a pair of trousers.

Validating that is important. We've had clients previously where time scales are short, people rush, and people make mistakes. You can use AI when you apply this tagging to validate: yes, I can see a jacket. Yes, it's a white spotty dress, and the spots are blue. This helps make sure that before you go downstream, at least you've had that check.

I'm just going to give you a quick show about the video with tagging before we move on to other areas. Let me just tie up that video, and I'll be back in 30 seconds.

There was a quick demonstration: throw images into the AI, and say one of many. The key thing to watch is that different AIs lean in different ways and see what tags they get back.

We can provide processes where, if there's a particular preference, Google, Amazon, Clarifai, and many others produce different AIs that we can hook into. Then there's also scoring. Only if you're absolutely sure you see this particular product, please let me know. If you're uncertain, 80% and lower, I'm not interested.

We can use information to use AI to get that tag, compare it to a taxonomy that we have, and then apply it to the assets on the way in.

We already talked about checking with product, but AI can do so much more at this early stage. The first one is checking resolution. We work with clients where they work with some fairly big partners, and the bigger the partner, the less willing they are to work with you because they work with so many people. Sometimes the images that are available aren't high enough.

AI can be used to up the resolution, and these days it's getting really, really good. Obviously, you do have to watch where it guesses, but for most systems where you get something in and you're looking for that kind of height, nudging the resolution up and doing it with AI can be really, really good.

Data: again, using this where you're matching data with your asset, being able to scrape data from PIMs and from other different sources, and being able to say, “I've got these images, I've got this data,” and then bringing those together. Many times, you can sit with humans trying to do this, but you can use AI to bring that closer, quicker, and more accurately.

One of the other things we have is relating. Relating assets together and having AI look at those pieces and say, “Actually, I believe they're from the same photoshoot.” I have another video to show, which will dovetail into searching. I'll show that in a moment.

You might notice there's a box underneath translation. Quite often, one of the things we get asked is: I have tags, I have metadata, but I have many different territories I'm servicing. Can we have tags in different languages? You can, but then there's a thing called overstuffing with tags.

I've put this in a dotted box because, with AI, we can search and do this differently. I think Kristina touched on this a little bit earlier. Rather than pre-building into your metadata all the translations, it's far better to stick with one language and then use AI as an expression of that out the door for different people to use.

How many different languages are there? Currently, you might be working with five territories and five different languages, and then there are another five. Do you go back over your assets and add more languages? The answer really should be no, because you're now doing work for no reason. AI can help with this, but this is more in a discovery phase.

If we skip on to the next piece, you've got your assets, brought them in, enriched them, brought your PIM data in, brought data from external sources, and used AI to help enrich. Now this is where you can discover what you have.

Your brand managers and your marketing people can come onto your DAM and ask questions and search in a natural language form. Previously, DAMs would rely on the tags and metadata that you've added. Now you can elevate that and start talking in a natural language form, using AI to discover your materials, just like you do with ChatGPT these days and the way you talk to it.

This is the next evolution of finding assets in DAM using AI.

We also have visual search. I'm just going to touch on this. We looked at relating assets together earlier. I was talking about trying to bind those together. There's another video which touches on this. So we'll just watch this couple-minute video.

Here's another example of AI in DAM. Here we have a whole load of product images, all seemingly uploaded at random as we scroll through. This is currently sorted by the date it was imported.

We find an image that we think is exactly the thing that I'm after. But there could be other images from the photoshoot that were uploaded. They could have been uploaded at different times, and essentially, you want to find if there are any similar.

This is where visual search comes in. It's not looking at metadata. It's actually looking at the image itself and looking for anything that might be similar.

Here, I've selected the image, and we click on visual search. It's a simple little interface. Percentage score: if I was going for 100%, it would find exact matches. Am I looking for duplicates? It could be a different resolution, but essentially an exact match.

How many do I want back? In this case, as low as possible. Fifty will do. I'm going to do a quick search.

Now that I've done a search, it's going to return me to a tab. As we see here, it's actually returned the similar images from that very same photoshoot. Now I can gather those together, select those, pass them down the workflow, and on to be part of a product.

This is a great example of how AI can help, especially with discovery. Thank you very much.

There's a good way in which you can find images. When you are looking at the next piece, part of discovery is also insights. You can use AI to scan your product and say, “Do I have all the necessary images to move on?”

As a buyer, I'm very conscious that if I put a singular image up, my sales might be this much. If I have five images up, my sales might be greatly increased because the consumer can see my product in more detail.

The fact is, I might have actually taken one image and cropped it several times: a close crop to show a little bit more detail, color chips, or a close crop on different pieces. Therefore, there are more images for the user to consume. Generally speaking, the more images you have for a consumer to look at, there is a direct correlation on bottom-line sales.

As a buyer, do we have the necessary images for me to release this? Is this something that I need to push to the repro house? Where are the other images? I can search, as we've seen. Are the images there? I'm sure we had five images.

This is really good insight that you have. Equally, you can have it inform: when we get anything that is of this style product, new jackets, please let me know. As a marketer, you might think, we're approaching winter. I'm after coats. Any coat image that comes in, please let me know about it. I'm interested in coats, long-length coats.

Again, you can use AI to say, “I'm interested in coats with a bit of space to the left-hand side, please, because I want to add some marketing text.” These are great ways, not just from the basic image search, but then more detailed questions that different departments within your organization can ask. They can say, “I'm after this. As soon as it appears, I don't have to go find it. Let the image come to me.”

That's a really great way of using AI to speed the process up and have the material automatically float to the people who need it most.

Let's look at process. This is the next stage in the process. Now we need to do something physically to the image. First of all, there's reproduction work, from color grading to recolor.

Do I have the white jacket? No, I don't. I've got the blue one. I've got the green one. I'm going to need to recolor that product because I'm not going to commission another photoshoot just for that one color. I can repurpose these images again.

Cropping is a really big one that pops up time and time again. That's very labor-intensive, and sometimes we've had clients take an easy route in cropping because it takes too long. Ideally, they would like a house style. I've got a little video to show you on that a little bit later.

We have informed selection. As well as AI helping color grade, recolor, and crop, it can also help your camera roll. It can analyze and pick. You might have 200 images from the photoshoot. Out of those, I've had AI trained on my own material. What are the likely choices I'm going to make as a user, as a buyer?

If I can have AI suggest that, and you still keep a human in the loop to then validate and go, “Yep, that's a good choice,” or, “You know what, the third one, I wouldn't have picked that. I'll pick this one instead.” This is where we're using AI as a tool. It shortens that time and essentially brings to the user a preset of material and says, “How about this? I'm suggesting this. Are you okay with that?”

As you grow, you've got buyers who really know their stuff and really have been out in the world, brought together the boots, the scarves, whatever product it might be. Having them make that final choice is always going to be important. But narrowing that down, you don't want to have them spending time looking through all these products. You want them to use this and say, “Here we go. Please help us make this choice.” Then you feed that back into the loop.

This is something where, from my part, the insight here is that we've got a final slide we will end on, which shows that AI is not a one-way process. You take your results, bring them back to the beginning again, and go again. You improve the cycle. We'll touch more on that.

AI is also useful for planning. We've found that some of our clients have photoshoots and are trying to work out the best way. Traditionally, you've got various tools, from Monday to Asana to a whole host. Using AI to say, “I need these photographed. These are the most important. This is what I've got in stock right now. Can we do all of our jumpers together because it reduces that time and saves a lot of cost in that organization?”

Using AI to take this data and the requirements for the business, for repro work, for photography work, for still work in the studio, and reduce this into: “Here's your plan. This is the most efficient way to go through.” That's something that AI has been really, really good for.

I'm just going to show you a little video now about AI cropping. This came from working with retail clients for well over a decade now. How can we crop the many, many product images that we have quicker? How can we make sure that we can have a house style? How can we then push this to our external clients where they're asking us to do all of the work?

We can't just push them some source images and say, “Hey, you do it.” They want it pre-prepared and ready to go themselves, but they want it in their own style. Then as the business changes, the marketing team says, “For 2026, we've now got this style that we want to introduce for cropping market sets.” So it's all about that process and how we do that.

It's a little five-minute video, and it just shows you a flavor of what we can do. A lot of this we actually automate via API. This is just an interface to explain. Let's take a look at that.

Everybody, I'm Paul Walker, and today I'm going to take you through AI cropping in WoodWing Assets.

First of all, you have a lot of images flowing into DAM of all sorts of shapes and sizes, and what you need to do is apply your house style into the cropping. It might not just be your house style. Also, as you outsource images to other brands, they might want to take your product but have their own style to that, and they might want you to provide that for them.

What I'm going to do today is take you through a simple view of the process of how we can use AI to achieve that.

First of all, though this could be done at mass via API, I'm going to show you through a simple little front end we have here. Here's our image. The first thing we're going to do is generate labels. We'll see here in a moment that it's going to give me a quick view of what it's found.

This is just a quick assist of what it's found. Now I'm just going to delete these images here because I want to show you this in real time.

What we have here is a set of options by which we are pushing the AI to make a choice based on the style. These are the presets where we say this is our preference. For example, the first thing I'm going to do is show you the boxes. Very simply, it's found two products: a handbag and a coat in this instance. We've also got the full body shot.

What we're going to do now is say, well, I could save that as a preset, but I want to start creating this house preset. The first thing that we could have done is say, yes, I'm after the full image, or no, I'm after the woman.

I'm looking for the woman. My shot is 2:3. I'm interested in a center-center. I want that piece to be in the center. Are there any focus offsets? We could go to town on loading these up. We're just going to take a simple view for now.

From there, I'm going to create my crop. What you see here now is that it's focused the woman in the center. The bag is off to the side, so she is the main focus. I've removed some of the redundant space that's required, and it's giving me that shot.

Had I applied this en masse to a whole heap of images, this is what we could have achieved. Equally, because it's found a handbag in there and through matching up metadata, you go, “Yes, we don't actually have a separate photograph of the bag by itself. Can we use this image to generate the bag?”

This is where we can come back and say, “I'm looking at the handbag. I want this to be a center focus again,” or I could want it aligned to the bottom because I always like bags aligned to the bottom. Again, I could be looking at a 2:3 because my main search image is always 2:3, for example.

I'm not going to do anything else. Create my crop for this. Here is my bag. This is very, very simply how we can use AI to identify product, match this with our PIM, and save it.

If I wanted to add padding, we could add padding to this. Essentially, what we're doing is using this to create a model for our house-branded style.

What we can then also do is reapply different styles to this by saying, “I need this to go to Zalando. I need this to go to another third-party partner, and they have a different requirement of style.” They might have square, and I can create my crop, and I get my square image.

There are lots of other things that we could do. If there is not enough image space on here, we can pass this through to other AI models that might extend the background. Or if it's just plain color, extend the plain color or change the plain color.

Here is where we can generate mass images based on the images that you've got in and reduce it to a normalized style so that you can remain on point and do this as quickly as possible. Then, if your brand house style changes, you can very quickly go back through your current estate and regenerate those specifically for your new requirements.

Thank you for watching.

Cool. What I'm going to finally end with is that you don't stop. You've gone through ingest to apply out the door. You can use AI to then harvest the results, see how the assets are being used, see what is used and what is not used, and use those to inform your AI models from enrichment through discovery and preparation.

Thank you very much. I'm going to hand back off to Kristina.

Thanks so much, Paul. That was a really great look at all of the exciting things that we can do with AI straight in our DAM systems and connected to that real workflow.

I was just looking back at the poll that all of you answered, and I can see that 56% of you are here today because you're researching potential use cases of AI and DAM. I hope you got some great inspiration from what Paul just showed us about all the different elements and problems that you could solve within your DAM directly on your content.

We wanted to talk a little bit now as well about what the dos and don'ts are that separate successful AI implementations from companies who get stuck in the experimentation and endless loop of AI pilots.

One of my biggest dos is to make sure that you start with business objectives and strategies before you run out and use new AI technologies and start experimenting, especially as your company. If you want to go off and experiment on your own, by all means, I do encourage that. The more we know, the better we're going to be at this.

But from a company perspective, you really want to take a step back and think about why we are doing this. From the research, we could see that the top three reasons that companies are exploring using AI and automation are to do things faster, to do things cheaper, and then to do things better.

I think that number three, doing things better, is really where we're going to see this shifting in 2026. All companies are now using AI for faster and cheaper. That's table stakes. Now, what we want to do with AI to shift our companies and be more competitive is to do things better and to do things differently.

Let's really think about your strategies internally and why you're using AI. Start with that why. For example, we need to improve our brand awareness for our European audience between the ages of 21 and 30. That is a business objective that's measurable and understandable. Then you can think about how we can use AI and automation to actually get there.

Really start with that why before you jump into, “Maybe we should use ChatGPT because everybody else is using it.”

We also wanted to share with you these six steps that we've created on how to get started if you're not sure. Firstly, if you're starting out with AI, definitely start with that why, like we just talked about, that strategy. Tie everything back to your strategic objectives.

If you're going out to use AI for reasons that don't link back to your business objectives, then you're going to be spending a lot of time and effort that will be wasted. Make sure to assess your organizational readiness, assemble that cross-functional team, run safe experiments, and everything that Paul just showed you are great examples of safe experiments.

If it's already baked into your systems where you have guardrails, security, and access controls, that's going to be a pretty safe place for you to get started. For example, you could test out that AI tagging in your test environments. Again, make sure it's in a safe space. A test environment is always better than a production environment.

You can test out those translations, and like Paul mentioned, just because we can do something like translations on all of our keywords doesn't mean that we necessarily should. This is why experimenting before you jump into that investment or implementation is really, really important.

Then make sure to go back and quickly evaluate what other AI tools are available on the marketplace. For example, last week we just had a really gigantic shift. A lot of companies are using OpenAI as their foundational large language model for AI work, but now suddenly Google Gemini, with their Nano Banana content generation tool, has really taken the lead. Now companies are thinking, “Should we look at Google Gemini because they're doing some things that we can't do right now with OpenAI?”

Make sure to always keep an eye out on the full marketplace and really understand what the alternatives are and what else you could be using.

Always make sure to operationalize and govern your AI. Before you say, “Okay, everybody, now we're using this tool,” make sure to think about change management. How is the AI going to be trained? What are the governance policies and guardrails to make sure that we keep the organization's information safe and our IP within our own organizations?

Just a quick one about fixing the foundations before automating. This is the idea that we're in this content volume paradox now. We can quickly create a lot of content, 50, 100, a thousand variants very quickly in seconds. But the reality is that AI is going to accelerate whatever you feed it. It's going to scale content even if it's bad content.

Think back to the example I gave right at the beginning of Unilever. They are hiving off a separate subset of assets that are their brand DNA, and they're building that into a separate DAM mini DAM area called DNAI. That is what they're using to train their AI.

That's a great example of using the right data. To do that, you want a single source of truth. We did see in the poll that 28% of you are not even using AI in DAM yet. Hopefully, all of you have DAM in your organizations. If you don't have a DAM system or that single source of truth, definitely start there.

I'm working with a lot of clients right now who are coming to me to say, “We want to use AI, but we realize that we don't have the right tools in place, our data is dirty, or we're not really sure what to do with our usage rights because they're not always clear.” That's where I want to start. Make sure we clean that up, get the foundations right, and have the right tools in place before we layer AI on top of that.

Remember, this is going to translate directly to the bottom line because 64% of consumers lack confidence in AI privacy and the content that we're putting out that is AI-generated. We need to make sure that we have proper governance and brand controls in place internally so that we can build that trust with our customers and we're not breaking that trust.

A good example is the Coca-Cola ad that just came out. For the second year in a row, they've generated it using AI, and there's been some real controversy there. They're an amazing brand, but there are a lot of questions around whether they're breaking some of that trust and confidence that their customers have in their products.

When we move to some of the don'ts, I want to bring Paul back on stage. Paul, are you still there?

I'm still here.

There we go. We wanted to talk a little bit about the don'ts. I know we only have a few minutes left, but I think you have some great examples here to touch on. Can you tell us a little bit about treating DAM as an archive or an island?

Yes. Quite often, we've had clients approach us and say, “I'm looking for an archive for DAM.” Really, DAM at its best is work in progress. From the moment something is thought of, an inception, even from a placeholder, all the way through to the end, DAM is that entire workflow.

It's really good for insights if you've got your material being gathered and worked on inside DAM, checked out, worked in the Adobe products or whatever products, and checked back in again. That helps collect the data in a single place and gives AI something to look at and use, and then also start suggesting.

If you only have it at the end of your workflow, you really lose out on a lot of good insights that AI can give you.

Absolutely. I think this is a really interesting one to stop on as well because we see a lot of AI implementations go wrong. There's a statistic that says that every change or transformational change initiative, out of all of them, 70% will fail.

These are just some of the reasons or some of the ways that we see organizations going wrong when they're looking at implementing AI in their DAM systems and applying it to their content.

Definitely think outside of the box and make sure to think holistically. We're not talking about DAM as this island anymore. We're really talking about it as the single source of truth for all of our content, not just the imagery.

It also has a role to play in tracking metrics. Even after content leaves the DAM and goes to our delivery channels, we want to make sure that the DAM is still there to help us track those metrics. Definitely a lot here to think about.

Finally, we just wanted to remind you all, we've talked a lot about the technology today, and we're seeing some really great innovation. But humans, all of you, are at the heart of this business transformation.

I always like to say that technology is only a tiny piece of the puzzle. A very important piece of the puzzle, but it's the people, the process, and the data that are needed to really make any of these changes successful.

For all of you who maybe are a little bit worried about your jobs, I honestly don't think that the machines are going to take over. But we need to make sure that all of us stay engaged, stay up to date on all of the changes that are happening, and stay enabled so that we can be the humans in the loop that we're talking about in all of these processes.

Absolutely. I second that. It's always human in the loop. Calculators came along. We no longer have to do pen-and-paper maths. AI is the same. It just gives us more time to validate what we're doing, helps make those choices, and AI is a tool. It's not going to get rid of people. I think it's just going to help the people who are there do more.

Fantastic. Thank you so much. It's been lovely to be with you all, and I'm going to hand back over to Magdalena.

Thank you. Big thanks to our speakers as well for sharing such tangible insights, especially, of course, Paul. We've gone from the big picture and the trends to hands-on use cases, and then real implementation guidance for all the retail organizations that are joining us today.

We have a minute left, but I'd like to ask one question to Kristina because the majority of the questions we've received, I believe, are already covered by Paul.

Kristina, if you could correct one misconception retailers have about AI, what would that be?

Yeah, I think it's a good question because there are a lot of misconceptions floating around. One of the things that I get asked a lot is, “Can we just sprinkle some AI on top of it, and it's going to be magic? It will work. It will transform the way that we work.”

I think that's a major misconception. Yes, we see everybody talking about AI everywhere we go, and our leadership are saying we need to use AI. But like we talked about, start with the why and the strategy, because just investing in some AI technology doesn't mean it's going to solve any of the problems that you have.

Also, remember to bring the people along in the process. We see companies do this all the time, not just with AI but with other tools. They start with the technology and say, “Okay, we're going to buy this tech. Here's a new thing. Please use it.”

But actually, if the people, the process, and the data are not aligned, that's not going to be the best return on investment. So definitely start with that why and think about your use cases first.

Yes, thank you. One last thing for Paul. What's the most surprising use case you've seen in retail, and something that teams really didn't expect that AI can handle so well?

I think it was the advent of tagging on images and it being able to correctly identify objects, and then from there, start cropping. Quite often, we've found image preparation has been the biggest consumer of time for retailers. To find that big piece can be reduced, I think, has been the biggest surprise for them.

Okay, perfect. Thank you both for your time and for your valuable insights shared with the community of WoodWing and everybody interested.

Again, this live recording is going to be available to everyone who registered on the sign-up form for the webinar, and of course live on our website in a couple of days as an on-demand webinar, because these use cases are actually something that I believe is really evergreen content, probably for the next couple of months, for the retail industry.

Thank you both, lovely speakers. Hope to see you again soon, and enjoy the evening.

Thank you very much.

Note: this transcript has been auto-corrected using AI, it may contain mistakes