Beyond the hype: How to take generative AI from novelty to growth driver

Generative artificial intelligence has shaken up the adtech world, but marketers mustn't get distracted from the work that actually drives business goals, says Graham Wilkinson (chief innovation officer, Kinesso).

2023 will be remembered as the year the large language models (LLMs) landed. The adtech world – like the rest of the world – has been through the whole gamut of emotions, from denial and even fear about the risks to excitement and optimism about the impact it’s already making in our industry.

Personally, I find my mind blown pretty much daily when I see all the ways in which generative artificial intelligence (GenAI) is going to change every part of our lives. I really do believe GenAI is going to be so transformative; it’s more akin to the discovery of fire than it is to the invention of the smartphone or the internet.

But here’s the catch. While GenAI (coupled with human intelligence) is going to be amazing at creating stuff, marketers shouldn’t let all the hype around GenAI distract them from meeting business objectives and goals. GenAI will absolutely play its part, but we must not overlook another important tool that helps us with analytics, insight, and optimization: predictive AI.

Whether it’s segmentation, media buying, or campaign measurement and optimization, these are classification and prediction problems. To solve them, we need to rely on AI that classifies information to provide predictions on future outcomes. So, while all the buzz and investment gravitate towards GenAI, we can’t neglect this critical capability. Here are a few reasons why predictive AI will be hugely important in any marketer’s AI playbook, and why brands should continue to invest in it.

Complex predictions rely on more powerful tools

We’re constantly hearing that the impact of devices and digital touchpoints has made consumer engagement – with personalized or even just relevant and timely offers – extremely challenging. But the truth is, this complexity is nothing new.

The world that brands operate in, and the people they’re trying to reach, represent a massively complex, dynamic system that has always outpaced our ability to master it with data-driven predictions. Simplistic segmentation and targeting fails and often backfires. Two people who look very similar on paper can behave very differently in real life, and vice versa. The most popular approaches brands use to make predictions still involve traditional machine learning methods.

These are valuable methods, don’t get me wrong, and they’re great at performing narrow tasks. But to cope with the complexity and dynamism of this system, and to have any hope of predicting and optimizing within it, brands need much more powerful tools like neural networks and specifically deep learning.

Brands face a choice after cookie deprecation

As Google looks set to follow through on the long-anticipated deprecation of third-party cookies in just a matter of months now, marketers are asking themselves how they’ll fill the cookie-shaped hole in their operations.

One way to react is to focus instead on the multitude of other signals available to brands – about customer sentiment, their engagement with websites, and so on. And, of course, a great way to interpret all those signals and make sense of them is with deep learning.

As cookies and plenty of other, similarly narrow paths to targeting fall away, what’s revealed is the challenge that was there all along – we just didn’t have the technology to tackle it: How do you make predictions in a dynamic system? For me, all roads lead back to deep learning.

The tech is tricky, but behavioral change will be even tougher

Deep learning sounds difficult, and it’s certainly more sophisticated than its simpler ML counterparts, but any data scientist can access a deep learning model. And it needn’t be a homegrown initiative. Plenty of adtech vendors are embedding deep learning in their solutions, whether it’s to make decisions about media bids or about the composition of the creative that’s served to audiences.

So, building a deep learning team or capability isn’t beyond any brand’s reach. What’s going to be much more challenging is bringing the rest of the organization along with you. Budget holders and other corporate decision makers can get cagey if they see AI as a black box. The sooner the work of educating and partnering with stakeholders begins, the better.

Much of the fear around AI is misplaced

Whenever I’m asked about the risks involved in black box decisioning, I think about the opportunities for using deep learning instead.

Deep learning allows us to leverage signals from anywhere and diminish or erase the risk associated with the use of traditional audience attributes that are intrinsically linked to unfair decisions and outcomes. So, with some of these deep learning platforms, you’re actually much more likely to truly tackle issues like data bias, and reach more of the marginalized groups you couldn’t have reached before and provide them with better, more relevant and valuable offers.

Of course, we have to be cautious about how protected characteristics or other variables are factored into decisions – but these problems and solutions already exist outside of AI.

We can’t judge new tech with old logic

When we upgrade our tech, we have to update our logic too. Otherwise, we impose outdated rules that will stifle innovation and limit what we can achieve.

That’s why getting buy-in from decision makers is so important. It will take effort. There’s a lot of education to be done in explaining how this stuff works and how we can use it to our advantage – not just in adtech innovation labs but in society – so that we can take the teeth out of some of the unfounded fears.

We need both GenAI and predictive AI

It’s amazing how quickly GenAI, as seen in tools like ChatGPT, has taken hold in our lives and changed the way we work. Behavior change always lags behind technical innovation, but that gap is shrinking as the advancements accelerate.

So, it’s really exciting to see the adoption of LLMs, but we mustn’t lose sight of complementary technologies like deep learning that power critical capabilities in analytics and prediction – the stuff that really helps marketers meet their business goals.

GenAI is disrupting adtech in a very visible way, but only when it’s combined with predictive AI can it solve the underlying problems that come when you try to impose some order on a dynamic, chaotic system like consumer behavior.

The combination of GenAI and predictive AI, powered by deep learning, will take us to amazing places – if we can embrace the change.