How Target's tech chief hopes to lead an AI-powered turnaround

Jason Del Rey: It’s been a while since I’ve made a trip out to Target HQ. Sad to say it was probably pre-pandemic. I came out, spent time with [former CEO] Brian Cornell and the team that was there back then. Since then, I’ve followed along mostly as a customer and as someone covering retail and e-commerce. It’s a pleasure to get to chat with you.

Prat Vemana: Thank you for being our guest. We're happy to host you sometime in Minneapolis. Our demo days are usually popular within the tech community, so would be happy to have you join.

Thanks for the invite. Why don’t we start at a high level. How are you all thinking about using large language models, both internally and externally?

PV: I look at it as a through line enabling our growth strategy. It’s very important that we apply AI in the right places to enable the strategy. [New Target CEO] Michael [Fiddelke] talks about re-establishing merchandising authority as well as the guest experience. When you look at merchandising authority, it’s all the way from dreaming up the next new style and design we need to bring in, to reliably making sure the product is at the right place at the right time for the guest. That whole backbone is powered by AI.

When it comes to our guest experience, we set the bar for drive-up and in-store experiences. The store matters. So the question is how do we power a store with AI? We do a lot of things like checkout, where technology is already enabled. But how do I give intelligence into the hands of our team members so they can take care of the guest? Every minute I save them so they can be guest-facing makes the experience better. So powering our AI-enabled merchandising authority, as well as the guest experience, are both in play for us right now.

Beyond those, you’ve stuck your toe in the waters on some of the off-Target shopping experiences LLM companies are trying to build. I’m curious how you’re seeing both the opportunities and challenges of working with the AI labs?

PV: We are a discovery retailer. We do have essentials that you pick up, but discovery is at the heart of what we do. Then you think about how discovery is evolving. If I go back to search, you got a bunch of links, scrolled down, did self-discovery. Advertising played a role in bringing the right products to the top. That paradigm shifted with social media. Discovery meant scrolling through stories and themes. It’s curated and endless.

Now another paradigm shift is happening with conversational discovery. What’s more important for us is how we show up more relevantly, not just through product but also through the story, the moment, the event, the occasion. For us that is the shift we are seeing. We want to be with the frontier LLM providers, learning with them on that journey. That conscious decision opened doors for us to be one of the early partners for conversational discovery and eventually conversational commerce, and what it means to play in the ad space on these platforms. All of that is still in its early evolution.

Knowing it’s early days, how do you balance leaning in enough to learn without neglecting more critical things to fix or accelerate at the company level?

PV: Both can be true. I bucket this as foundational and innovation. We need to improve foundational capability. Even if everything were perfect, with this technology shift you would have to rewrite systems anyway. So it’s not a choice. We need to redo our tech stack for newer technology. So our biggest effort is in foundation. At the same time we leverage relationships with LLM companies so we can be part of discovery in this paradigm shift.

When you see a company like OpenAI reverse course quickly on something like Instant Checkout in ChatGPT, does that make you nervous about whether they know what they’re doing in retail or commerce?

PV: Whenever we have a technology breakthrough, you will see different ideas and tests. Some will work, some will not. I wish we had a crystal ball for the next three to five years. But I’m encouraged they are leaning in and testing. If it works, keep growing it. If not, kill it and move on. That shows they were very vibrantly leaning in. You know, we run hundreds and thousands of tests behind the scenes that no one ever notices. When OpenAI does it, though, it’s very visible. 

Historically Jason, and you've been in this industry so you’ve seen this, a lot of discovery shopping sites came in, deal sites came in, curation sites came in, and none of them exist today. And so in this breakthrough conversational commerce world, I'm glad that OpenAI is leaning in, and I'm glad we are part of that leaning in to see what we can learn from them.

Target is one of the only retailers that have built a specific app inside of ChatGPT, and OpenAI has said they are prioritizing these kinds of apps those going forward. How are you thinking about what the ideal experience is for that and why it should live there versus on one of your own properties?

PV: The conversations happening outside our properties are much broader. Someone on ChatGPT might say they’re planning a trip and need help with travel, shopping, and possibly some tours and events. When they come into our own Target site, they are not asking about travel tickets. But we also want to be part of the upper funnel on places like ChatGPT so we can curate the right basket when someone asks about, for example, a cozy movie night. So what does that look like? From showing snacks to like blankets…anything that makes it fun and exciting and a friends or family event. We’ll be there.

But what is the actual goal of the experience with your dedicated Target app?

PV: Within the ChatGPT app experience, you might start a conversation about a cozy movie night. We then curate product choices as a teaser. As soon as you click in, you enter the expanded Target world or app within ChatGPT. Eventually you could get a similar experience on Target properties. And within our own app, we will have more signals like your location and will be able to get a lot more personalized fairly quickly. But in the end, the idea of a curated assortment for you, that doesn't go away. The concept is the same.

Let’s talk about the Target experience itself. What key changes from LLMs would you like customers to feel over the next 12 to 18 months?

PV: Early signals show searches on Target.com and the Target app are becoming more conversational with more clarity, like looking for a dress for a certain event or specific weather. So for example, give me a jacket for extreme cold. People are moving from keyword-based search to conversational as we speak.

Are those still small numbers though overall?

PV: It’s growing very fast. 

The second thing you'll also see is that when we’re curating the online experience for customers, we now have the power of using LLMs to take their worldwide knowledge and combine this with our own Target-specific knowledge. So, for example, if you're buying an area rug, we used to only be able to see people who bought an area rug on Target and then what else they bought at the same time. Those would have been really our only signals. 

But now, with LLMs, I can also basically query the world and say, Okay, for people who bought area rugs outside of Target, what's trending and what else are they buying. And when I combine these two datasets, the power of our recommendations becomes even more effective. 

We have made good progress in the last 18 months, where 60% of our app’s code has been completely rewritten to be AI-ready. So we have rewritten our search engine, we've rewritten our recommendation system. It’s a fundamental change.

Do you work with LLM companies on custom models or build on top of existing ones?

PV: We do both. We use open-source models for specific intents internally and frontier models for ongoing enhancements. For example, our customer service chat experience online uses a good blend of custom models we built ourselves using open-source models. But we use a combination, using the right models for the right purpose.

Things are moving so quickly though. Is it always clear what the right model is?

PV: Yeah, it’s evolving. When the problem we’re trying to solve is clearly defined, and most of the information the AI needs is coming from our own data, then it’s more straightforward. In other areas, where we need to blend external and internal intelligence and move faster, we can be more open to using frontier models. The capabilities of those models are improving very quickly.

For example, in things like catalog work and image processing, what we could do last year in creating a room scene versus what we can do now is significantly different. This will keep evolving, but our teams are continuously experimenting with both open-source models and frontier models.

OK, one more. You mentioned Gemini in passing. So Target is also one of the early partners on the Universal Commerce Protocol with Google…are you convinced consumers will want to transact directly or check out on Google properties like Gemini or AI Mode in Google search?

PV: It’s early. We want to be there, learn with them, and help shape their thinking. Those learnings improve our own properties’ experiences as well. 

How so?

PV: So I told you we rebuilt our search engine to be conversational and be able to understand conversational queries. When we integrated with ChatGPT with our ChatGPT app, they are passing us the conversational queries. So not only do I have my own source of conversational queries but also from the conversations happening with our app on ChatGPT. So the volume helps us learn much faster. 

I think we’re out of time but I’d love to talk again in the future about internal productivity and tools like coding assistants.

PV: Yeah, when I look at 2023 when LLMs broke through, it was a lot of curiosity. In 2024, it became a capability, right? We had a ChatGPT rollout for the entire headquarters population of employees and Copilot for the larger workforce. What that gave the teams is familiarity and comfort that they can use it for day to day purposes. 

In 2025, we started moving into productivity. So if you look at 2025, like 85% of my software developers were using some form of coding companion. Now I look at 2026 as sort of an operational transformation. So now we can go into the core systems, core processes, and rearchitect these systems and workflows in a way that it becomes much more intelligent, much easier, while also bringing human creativity to another level. 

The example that we talk about is how intelligence will help the designers design their next new product. So I think about this in waves but the waves are happening, not in years or decades, but in shorter timeframes. But we feel like we are still in the beginning of this transformation. 

So I shouldn’t expect to see a Jack Dorsey-style announcement about 40 percent of employees going away because AI has solved our problems and we only now need half the staff?

PV: I'm a technologist, so this is an exciting time to be in, right? The technology is changing, we have a CEO with a clear mandate and focus on a growth agenda, and we have an opportunity to rewrite some of our foundational systems with this intelligence. So to me, what I lose sleep over is how fast can I flow through my roadmap? How fast can I actually launch the capabilities? And I can use every resource in my team, and we are actually continuing to hire.

Thoughts or tips on the topics in this newsletter or others? Just reply to this email or message me at [email protected] or confidentially on Signal.

And feel free to forward this email to someone who might find it useful. If someone forwarded this email to you, please consider subscribing to The Aisle here.