This is the first in a series of posts about AI that I will be writing. As opposed to many of my usual, more technical, or at least lower-level posts about programmatic advertising, here I will be addressing issues also at a higher level when it comes to the intersections of AI and programmatic advertising. I will use the term AI to refer to both the more true-to-the-words artificial intelligence neural networks and other complex algorithms since the end results are often the same for us users - automation, like magic.
I am not exactly sure how it all began since I only entered the programmatic advertising industry in 2016 after working in more general web consulting. As I look back, it’s clear that even at that point, AI was already outpacing us.
One of the first concrete tasks that I actually performed after entering the industry was trafficking a basic direct campaign to one of the still-used big ad servers. When it came time to decide how to deliver the campaign, it was the first time I came face to face with our future overlords. I had to choose whether the campaign would deliver evenly, ahead or ASAP. At that moment, I had yet to grasp the trajectory propelling us towards an industry governed by AI.
I did not realise it, even as I filled in all the line item details and pressed the button to check if there would be enough room in the inventory for the campaign I was managing. The algorithm fortunately determined that there should indeed be enough inventory for the campaign. It arrived at this conclusion by comparing recent historical ad unit traffic data to older data and possibly even seasonal data or other factors. Back then and even now, we were not yet at a point where AI or algorithms could predict the future with 100% accuracy. Consequently, we still had to manually verify that the campaigns were delivering as predicted.
Then, I moved on to doing programmatic buying. Although I started with fixed price bidding, I do remember looking at the click, conversion, and other bidding algorithms with interest.
Mistake. Little did I understand that the more advanced AI, replacing me and everyone else, would be lurking within those choices.
Of course, I took that leap at some point when I had a suitable campaign to try the conversion algorithm. This marked a step into using more advanced algorithms, most likely involving actual machine learning. In this case, the demand-side platform used its vast historical data across many buyers, websites, and ad units to select precisely the kind of impressions that would lead to conversions on my client’s website.
Was there anything left for me to do after that? Essentially, my tasks boiled down to assigning goals, setting the targeting parameters within which the AI would be able to operate, evaluating the optimisation performance, and engaging in discussions with the client about new goals, campaigns, and methods. In essence, not much else!
After those experiences, I came across algorithms for dynamic floor pricing, creative optimisation algorithms, various types of lookalike and other modelled data audiences, and many other AI-powered features in both the selling and buying sides of the programmatic industry.
Well, if I were to ask a junior ad ops person working in an agency whether they are becoming bored due to the lack of work, I doubt that they would agree with the sentiment. Similarly, senior heads of programmatic at publishers, who are trying to figure out how best to build and maintain their programmatic ad stacks amidst changing technological, legislative, and advertiser pressures, would likely not share that sentiment.
The point is that, yes, AI is becoming an increasingly prevalent part of our toolkits in many areas of the buy and sell side. However, while it may help automate A/B and multivariate testing, it still requires someone to choose to use it for a particular task, set its goals, and analyse whether it achieves them. Furthermore, it’s worth noting that the same task of AI optimisation can be accomplished using several different tools, each with its own advantages and disadvantages.
While AI may shift human focus within the same tasks, it may also create entirely new ones. Consider ChatGPT, a highly discussed AI, and how it is giving rise to an entirely new role, or at the very least, a new set of skills called prompt engineering. In addition to making headlines, it has prompted the creation of a number of courses teaching these skills, even at universities.
If you are still unsure about what will keep you scratching your head while receiving a paycheck in the coming years, take a look at Gartner’s Hype cycle for Digital Advertising or Emerging Technologies. You will probably find at least part of the answer there.
How Should Adtech Companies React to This AI Change
We have long recognized the necessity of constantly learning new skills in the programmatic industry to ensure success, as new initiatives and technologies change how we work. Looking back, it has been just six years since the introduction of ads.txt. In the past couple of years, we have also had to adapt to significant changes, including the introduction of TCF framework, sellers.json and the loss of third party cookies, to mention just a few.
The advent of AI-powered features will only accelerate the pace of change. A recent article ‘Reskilling in the Age of AI’ in Harvard Business Review states: “The average half-life of skills is now less than five years, and in some tech fields it’s as low as two and a half years”. I would argue that in our industry we are closer to that two and a half years.
This means that companies that want to stay ahead, or even on par with the competition, need to invest in upskilling their workforce. In some cases, they may even need to consider reskilling workers. The aforementioned HBR article emphasises that companies should make reskilling a strategic imperative, led by the leaders. It should also be integrated into the goals of every team manager, including those at the managerial level.
The HBR article further suggests that ideally, workers should be provided with sufficient time, separate from their regular tasks, to participate in the reskilling process. This is because it is likely to require both mental space and time commitment. Moreover, the company should clearly communicate the reasons behind its upskilling or reskilling efforts to ensure that participation is not limited to only those who would have participated anyway.
As individuals and companies within the industry, we should work to strengthen and foster industry-wide learning through organisations such as the IAB & Prebid. This will ensure that the necessary skills and skillsets required for success are spread throughout the industry, rather than being concentrated solely within the few giants with almost endless resources.