AI. But what does it mean?

Will artificial intelligence replace the need for performance marketers?

Recently, I read the book Life 3.0 by Max Tegmark and it got me thinking about how exactly artificial intelligence may impact life as we know it, specifically from a digital advertising perspective.

AI is obviously a huge topic at the moment, with market leaders Google and Tesla making huge moves into the space.  AI is already usurping humans in tasks like calculations, video games, board games, investment strategies…even driving cars. With AI development a clear focus for Google, it has made me question whether performance specialists will still be behind the wheel in years to come.

The chart below, created by scientist Hans Moravec, depicts the landscape of human competence across a range of tasks and AI depicted as rising flood water in this landscape. Creative abilities such as art, science and programming sit significantly higher, out of AI’s grasp, while low-lying abilities such as arithmetic, chess, driving and translation are steadily being swallowed up the rising AI water levels.

Hans Moravec (Robotics Institute of Carnegie Mellon University) – Landscape of human competence

AI for digital marketing

Google has already implemented automated AI systems to optimise bidding, optimisation, and budget management. This innovation has been a real time-saver for performance specialists, with smart bidding algorithms now handling CPC bids across multiple levers such as locations, devices and audiences. This, in turn, frees up more time to work on strategic insights and creative tasks such as ensuring a robust search account structure, copywriting, analysis, and planning.

However, as predicted in the Landscape of Human Competence above, AI will continue to progress in capability and many, including myself, see tasks such as keyword building, copywriting, landing page development and optimisation becoming increasingly automated as capabilities continue to grow. In 2016, Google Translate began using Neural Network technology to invent its own language to better understand the context of full sentences. This technology has hugely improved Google Translate’s ability to reflect the nuances of different human languages. Earlier this year we saw Google Duplex, the AI Assistant successfully imitate a human voice, convince the recipient of the call that it was, in fact, a human and successfully book a haircut. These 2 examples show the true potential of AI and its increasing competency in interacting and communicating with humans.

According to Moore’s Law, the exponential growth of learning for these technologies puts the idea of AI actually writing ad copy in the near future in the realm of possibility. The current scope of the ability of AI is narrow. Systems that are able to exceed human-level competence exist only in specific and narrow corridors. They are still not capable of achieving increasingly complex problem-solving. AI will only become a comparable skillset to specialists if (and when), these systems are able to combine multiple skill sets while also analysing ad effectiveness and audience resonance.

This all sounds good for advertisers and bad for performance specialists, right?

Well, here’s the bit where I think we should take all of this AI buzz with a grain of salt. (And I’m not just saying this because I want to keep my job!)

Why we’ll always need digital advertising specialists

I see performance marketing moving towards and taking advantage of AI technology. Ultimately, I think the move will be less towards implementing or advising on the use of digital advertising platforms, but we will move toward becoming gatekeepers and controllers of AI technology. It will be our responsibility to monitor and ensure that the systems are acting as expected and aligning to the results that we are focused on achieving for our clients.

Increasingly, performance marketers are spending their time learning and understanding about the new emerging technologies and digging deeper into the specifics of how AI can deliver positive results to campaign conversion targets. For example, current smart bidding technologies have limitations such as a minimum volume of conversion data for learning and optimisation. This excludes a large chunk of smaller advertisers from utilising Smart Bidding AI right off the bat. Smart bidding has the potential to react differently and unpredictably depending on the category, search account structure and fluctuations to conversion rate driven by external influences. Unexpected problems with conversion tracking, sudden changes to the level of budget spend and adjustments to keyword structures can be disastrous for conversion performance, with AI grinding to a halt when unsure of how to react when faced with unpredictable situations. Although continuously improving, there is still no clear picture of how external variables are able to be taken into account by the algorithm.

An additional aspect to consider is the concept of ‘goal alignment’ between humans and AI.  When we set the AI a goal, such as a specific revenue or CPA target, to what extent will the AI to try to attain these goals? With increasing data segments and bidding levers provided to Smart Bidding, the more potential outcomes we can be faced with. I If given the freedom to adjust keyword structures and ads, would the algorithm begin using unfavourable keywords or deceptive copywriting to achieve these targets, to the detriment of a brand’s reputation? If given the ability to analyse competitor insights data, or scrape search results and competitor websites, would Smart Bidding provide an unfair advantage to achieve the perfectly optimised way to own search results? How will multiple advertisers using multiple versions of Smart Bidding, in the same product category, competing on the same keywords, react?

Of course, the last few questions are far-fetched. But, the truth is that we honestly just do not know what potential results we could face as we gradually hand over increasing control and resources to AI systems.

Ultimately, all of these questions make me realise that there will always be a need for humans to be present. To verify, to validate and to confirm that there is total alignment between AI technologies and our client goals.

Let’s work with AI

Columbus has wholeheartedly embraced AI technology on a number of levels. We have begun utilising Smart Bidding, reducing the time spent on optimisation, allowing us to work more strategically with our clients. This has allowed us to focus on better understanding their consumer, brand and category, resulting in better-planned campaigns, adding real business value.

On a more advanced level, Columbus has embraced Machine Learning (ML) and are increasingly creative in the ways in which we leverage these learnings. We are now using ML to build predictive and propensity models to find the most valuable customers for our clients. ML allows us to identify patterns on past behaviours of our clients’ consumers. The ability to apply these patterns to new data, in real time, allows us to identify more valuable customers and those most likely to convert, increasing ROI and reducing wastage.

AI has clear potential in benefiting efficiency and effectiveness within our day to day lives. There is true potential for performance marketers to get back to the creative and strategic work for clients, delivering quality insights and digging deeper into analysis. I don’t think that AI has the ability to supersede performance marketers. It will instead transform how the industry operates, in ways that humans probably haven’t even imagined yet.


0 Comment