AI in Advertising and Marketing – Hype, Disillusionment, Back to Hype Again?

Remember when we thought all cars would soon be self-driving – it was just a matter of time – and we were all about to get all that time at the wheel back again for greater productivity, more downtime, you name it?

In the oft-copied, never bettered Gartner hype cycle, automotive AI ultimately went from high to low with self-crashing and ‘smart’ motorways both featuring on the hall of fame.

But what about AI in advertising and marketing? Since the arrival of programmatic, arguably, this is a sector which, even more than others, has bought into the promise of artificial intelligence. In this piece, I will set out the history so far, where we currently sit in the hype cycle and the promise of where automation might go next in this sector.

Advertising AI, the Early Years

Ever since 2007, when programmatic was first emerging, there was much promise of time saving and improved ROI through automation. The benefits would be felt across the industry, ‘all boats rising’ in the process. Of course, this wasn’t exactly how it worked out, at least when seen in the light of one early player, one who incidentally pushed especially hard its credentials in AI:

Though beyond their exuberant claims, nobody quite knew how the company’s cutting-edge artificial intelligence was informing its media buys.

The title of AdExchanger’s post-mortem on this particular business actually gives the simple answer how – ‘by mixing client data that wasn’t meant to be mixed.’ The piece is worth reading in full, for nothing else than a cautionary tale of the worst excesses and potential pitfalls players in this space may be subject to.

Given what one employee described as the “ethical quagmire” it sat upon – even preceding as this did GDPR – and the fact that the company made AI its unique selling point – the phrase even appearing on its strapline, this is difficult to unsee.

Read More: Modernizing the Retail Supply Chain with AI For Resilience and…

Even if not commonly known now, you must think this marked a low point for AI in advertising – at the very least, having an impact at the time on trust among advertisers, agencies, and those looking in from the outside around the veracity of the use of the term.

From AI Low, to New High? 

In the intervening years, both the terms artificial intelligence and machine learning have done a lot of heavy lifting in ad tech – most often referring to one specific use case. That being in the categorisation of cookies dropped across the web into different demographic groups.

August marked a change however, with Gartner setting out the four technologies now set to transform digital advertising: with AI seemingly on the rise again, against a changing backdrop with the decline of the cookie and the rise of privacy:

The suppression of personal data for marketing alongside the rise of AI to assess contextual response anonymously is altering the data foundations of advertising and content marketing.

It went on to highlight generative AI as one of three specific areas of focus:

In the face of third-party data deprecation, generative AI can help to identify the core characteristics of customers and target them with custom content in a privacy-compliant way. 

Top AI ML News: MATRIX Successfully Accelerates AI World Generation Rate by About 35%

All of which brings us to ChatGPT. If you work in advertising and unless you live under a rock, you will have seen some of the giddy excitement around this new tool from OpenAI over the past month. Media buying, creative, accounts payable… you name it, they say it’s going to replace them all.

If you think back far enough, these aren’t that different to the old ‘rise of the machines’ scare stories that originally accompanied the emergence of programmatic itself. And though ChatGPT fulfils the requirement that like all exciting new tech, it looks like magic, there are a few reasons as with the above AdExchanger story quoted, we need to move beyond the hype to fully understand what it is doing.

While it does ‘hallucinate’, and invariably get some things incorrect, even appearing like an over-confident human at times – that very observation – that it could pass for having human characteristics – also shows what a potential leap forward it marks in AI. As one observer put it, with ChatGPT and its ilk:

We can start getting to a world where computers work in a way that comes closer to the way we intuitively think they should.

Democratizing computing in this way is arguably a massive game changer, regardless of which industry you work in.

In terms of potential drawbacks, like its image generation sister technology DALL-E, it draws upon third-party web sources for its content. While in written form versus illustration, the copyright issue is less glaring, it’s still worth considering. ChatGPT’s usage of ‘third-party content’ therefore arguably has question marks around it if used creatively. As AI researcher Margaret Mitchell has also pointed out, ChatGPT doesn’t provide its sources – so in some ways, it’s comparable to the early days of search, pre-page rank – and how it responds to or avoids misinformation for example, is still not clear.

In summary, and on balance, this still feels like an important advance in tech. As long as we test, learn and understand quickly where it can and can’t provide objective benefits to us. Generative AI has the potential, as Gartner suggests, to complement a cookie and ID-free future, where respecting the user comes first, but not at the expense of business.