Meet Your Customer: Personas & AI

SPEAKER_1 (0:01)

This is The IAB UK Ad Tech Podcast.

Pat (0:06)

Welcome to The Ad Tech Podcast, where it's 34 degrees outside, but as usual, we are cool, calm, and collected. My name is Pat, I am an Ad Tech Manager at The IAB UK, and I'm joined by, and I'm Joe, Ad Tech Executive at The IAB UK. And we will be very shortly talking to our guest of the week, who is Joe Mountford from Crimtan. We'll be talking all things persona. But first off, we'll start with a bit of news as per usual, Joe. It's pretty worth mentioning, actually. We are down one of our usual contributors, Alex, who is unfortunately on holiday. But I think we can hopefully steer the ship.

Joe (0:37)

I think hopefully we'll be able to make up for her. And I guess to get us rolling, you've been looking into this Stackadap story, is that right?

Pat (0:43)

Yep. So there's been a newly released product called Ivy Studio, which is quite alluring as a name. My understanding is that it's essentially kind of all-encompassing, LLM-inclusive product, where essentially all of the different MCP integrations, I don't know if you want to define that afterwards, are kind of already baked into the platform. So essentially, it's kind of a one-stop-shop for any sort of campaign running. So you're essentially using kind of natural language to do everything, including all of the data integrations you might have via Stackadapt, all of those types of things. What is an MCP kind of layer, just to start off?

Joe (1:17)

Yeah, I guess we should probably explain that, just in case any of our audience don't understand. I mean, MCP is often described as like the USP for an agent world. It's essentially the layer, the tool, the protocol, I guess I should say.

Pat (1:30)

It's finally later a lot, yeah.

Joe (1:31)

Yeah, lots of different terms for it. It's essentially the layer that exists between an agent and between a set of tools, which could be a myriad of different things that just allows that communication to take place. So it's very industry standard, released by Anthropic, I want to say a little over a year ago now, but definitely we're really starting to see it sort of coming to fruition now, particularly in our industry, as more and more people are experimenting with agents. And I'm particularly interested in this Stackadapth story because recently during our AI community, we were talking a lot about agents and somebody kind of mentioned the now seemingly quite obvious point about, what exactly are the benefits of this? We're moving into this future which is a gentrified but how exactly does that help us? And I think a big point that came out there was this idea of accessibility, this idea of, well, now you can approach things using the actual language, and that sounds like what StackAdapt are going for with this.

Pat (2:27)

The tech layman gets the keys. I don't know if that's always a good thing, but potentially it is. I guess the unique thing about this is the fact that a lot of the time we're seeing MCP integrations kind of with the purpose that people can do a lot of this querying through their LLMs themselves. It sounds like this is slightly different in the sense that it's kind of being baked into the platform itself. So it would be interesting to see, I think for StackAdapt, obviously, a pretty big DSP, maybe makes sense if they've got extremely loyal customer base. It would be interesting to see if other DSPs try to emulate this. My thoughts around this is that, well, MCP is obviously going to be important for anything, either way, if it's the kind of integration people are using for their Claude, for example, or if it's kind of baking into a platform. But what kind of wins the race in terms of how people actually use them for buying? Will be interesting to watch.

My thought is the LLM route, but I could be wrong.

Joe (3:15)

No, I'd love to go against you there, but I think it definitely seems that way. In a bizarre manner, I think we're going to end up with a very similar situation as far as sort of relationship between DSPs and even SSPs is concerned. It's just that they will all flow into that one sort of LLM interface. And so the big benefit will be not having to jump around quite as many dashboards and try to understand the different nuances of each, because it will just be relayed through your LLM of choice. And like you say, whether or not that's a good thing, maybe it loses some of the nuance when you're missing out on the various intricacies.