By Peter Mason
AI in marketing has become synonymous with one area in particular – automation – and usually, as Andreessen Horowitz’ Benedict Evans describes it, around “one discrete task, at massive scale.”
By automating the trading of ads through real-time exchanges, programmatic tech has given buyers instant, always-on access to any number of publishers and broadcasters. Increasingly, that’s across every different screen and media type too, from audio, to outdoor and AR. In brief, marketers can do more with fewer resources, faster than ever before.
But when it comes to meeting marketing’s true potential in our connected world, two aspects of AI are still holding us back.
Fact and science fiction
So much excitement is attached to the phrase ‘artificial intelligence’. Amid the hype, the challenge is in separating fact from science fiction.
Brands are definitely building more technical knowledge, as they reportedly take more programmatic tasks in house. But equally, there is still a bit of a haze around AI right now, its limits and what exactly it can and cannot do. We are immersed in it on a day-to-day basis – making it easier to tell genuine innovation from the more mundane – in some cases, little more than narrow pattern recognition. But that isn’t necessarily the case for all corners of marketing and advertising.
As experience and knowledge keep evolving, our hope is that distinction becomes easier to make for all.
Conformity & evolution
The second factor holding us back is the flipside of the same coin – a lack of diversity in tactics.
Ever since the terms martech and adtech were born, we’ve seen an overwhelming focus across the market on one approach. Which is audience targeting, 3rd party data and retargeting. Though all bring great benefits, their overuse has arguably also spurred ad and browser blocking, as well as banner blindness. Founded as they are on personal data, these methods also face an increasingly uncertain future in a post-GDPR world.
Again, my hope here is that wider understanding of AI and machine learning will usher in more creative, diverse approaches – beyond what has effectively already become the ‘standard’, and as a result may already be starting to see limited effectiveness.
Advances in natural language processing, emotion AI and advanced contextual targeting are all promising alternatives. Worthy of an industry which, let’s not forget, still lives or dies on creativity.
But lessons from the past must be learned here too – even those new approaches require smart, differentiated implementation, as well as upholding the user experience if they are truly to succeed.
Beware AI self-driving mode
On a broader level, the tech giants that dominate marketing are still themselves learning what AI can and can’t do. See, for instance, how Google and Facebook have both massively ramped up their human moderator workforce in recent months. Even a short time back, that wasn’t part of the plan.
Where AI tends to go wrong, there’s a pattern: often we see an element of misunderstanding the tech’s limitations, then sitting back and effectively putting it in self-driving mode. Look at what happened with Facebook and fake news, which directly followed the removal of the social network’s entire human editor team. The same applies to ‘brand safety’ scares in programmatic advertising.
Automation is a huge benefit, allowing brands to do more with less, and faster too. But there’s still no substitute for humans monitoring, managing and optimising the process. For AI in marketing to truly flourish, our technical knowledge must evolve, while its application becomes more diverse. Our hope is the one will lead to the other.
It’s only then do I believe that AI can finally deliver on its original promise – to make marketing both more effective and customer-centric.
Peter Mason is the Chief Technology Officer at illuma.