Blis recently launched Blis AI, a new AI-powered planning tool that turns a brief into qualified audiences ready for omnichannel activation, without relying on identifiers.
Using natural language inputs, Blis AI translates a simple campaign brief into ready-to-activate audience recommendations, each based on a dynamically assembled range of audience inputs and backed by transparent rationale. Built into Blis’ existing Audience Explorer platform, the tool draws on the company’s location, lifestyle, purchase, and behavioural data to surface audience combinations that can often be overlooked in traditional workflows.
New Digital Age spoke to Amy Fox, Chief Product Officer at Blis, to find out more…
What motivated the launch of the new Blis AI tool?
AI is a hot topic in our industry right now. Lots of companies want to be involved and do something with it. We’ve had various technologies over the years, machine learning-led, that have leaned into AI, but not really in a user interface way. Our technology has always been about using the very best intelligence to decide whether a bid is likely to result in a win, a click, or a conversion.
What’s been lacking is thinking about how users actually interface with AI. We’re asking: if you’re a client who wants to use our platform, our products, our data – how easy is it for you to do that, especially if you don’t know a huge amount about our business or our data partnerships?
How does the new tool help advertisers?
This is us leveraging a different style of AI. We’re integrating large language model technology into an interface that already exists – our planning tool called Audience Explorer. It’s our flagship entry point into our product suite, and it’s what many clients, whether planners or traders, use to build audiences that eventually get activated in our DSP.
One of the problems we’re solving is that to know what audiences to build, you have to have pretty in-depth knowledge of the data sets available. I’ve been working in this space for almost 14 years and have personally helped build the platform, but even I don’t know every single data category we get from partners such as O2, Experian or Circana.
There’s this problem of sitting in front of a blank page thinking, ‘Where on earth do I start?’ I like to call it a ‘starter for ten.’ Large language models can quickly do what a human would take a long time to research and figure out.
The combinations of audiences you can build in our Audience Explorer are staggering. There are tens of thousands of individual filters, but when you consider how you can combine them, you’re looking at trillions of possible audience combinations. That’s our challenge – people come into the platform and think, ‘Okay, where do I start?
Instead of building everything from scratch, you can now use a prompt plugged into an LLM. For example, you might say: ‘I’m planning a campaign for Aldi. My objective is to drive footfall into stores. I’m targeting young families on a low-to-medium income.’ The AI can then generate suggested audiences, complete with rationales and filters, so you’re not starting from zero.
At no point is the AI saying, ‘Job done, take it or leave it.’ It’s a starting point that you can refine. You might decide the reach is too broad, or you want to add an income filter.
How does the new AI element enhance your platform?
It’s important to treat large language models with caution. We’re very aware of the potential hallucinations that can happen. For us, the LLM is a starting point—a way to help you scan through data sets and suggest a few to begin building your audience. But you, as a brand or planner, still know your client best and can refine it further.
An AI, especially a large language model, is only as good as the product it sits on top of. The Audience Explorer platform is still the same core product it’s always been. The difference is we’ve layered on a conversational interface. You can now say, ‘Hey, I’d like help doing this, that, or the other,’ and it gets you from A to B much faster.
The inputs and the speed from input to output have significantly increased. It’s about improving the usability of an existing product—it’s not a new standalone product we’re selling separately.
What sort of data does your platform have to work with?
We work with two types of data: personal, ID-based data, which is increasingly scarce in Europe, and geographic, aggregated data. Our third-party data partnerships tend to be direct-to-source relationships, for example with O2, Experian, and Circana, where we get directly observed data. Usually, this data is sold at a geographic level rather than linked to individual IDs.
Instead of looking at single-person data, we’re mapping data back to geographic areas, postcode districts or custom grids. For example, an area might over-index for Ben & Jerry’s ice cream sales or under-index for Netflix usage. However the data is cut, it’s tied to places, not people.
What has been the reaction to the new tool to date?
When we build new UI-based products, we start with mock-ups and interactive wireframes and take those to clients for feedback. We’ve shown MVPs and prototypes to clients, and there’s been a really strong appetite for this. Agencies, in particular, are very time-poor and juggling multiple clients and projects. Anything that helps them work faster without sacrificing quality is hugely valuable.
Media planning just keeps getting more complex. There are more things you can do, more channels, and more ways to approach them. The fragmentation of the CTV marketplace alone could be several full-time jobs to stay on top of.
One thing we’re exploring, maybe over the next 18 months, is extending this technology into channel-specific strategies. Once you’ve built your audience, we’d love to use LLMs to help plan which channels to use, because the complexity in the channel landscape means everyone has to be a generalist in areas that are very complex. There are pros and cons for every channel and even for the subsidiaries within those channels.
Are there any trends in the digital marketing landscape that you think are worth keeping an eye on?
I suspect we’ll see an evolution on the DSP side, with more and more automation around campaign creation. We’ve started by tackling the audience challenge because that’s already a big problem to solve. But there’s definitely a world where the technology extends into channel strategies and even into actually building and shaping campaigns. Some of the work traders do is very rinse-and-repeat and time-consuming. There’s a lot of potential for AI to drive operational efficiencies in setting up, trafficking, and optimizing campaigns.
Companies in the creative space are investing heavily in AI for creative solutions, which makes sense. I think businesses like ours will focus more on operational efficiencies, making sure machines can handle more of the heavy lifting so humans don’t have to.