Ignore the hype: There are no AI-driven media agencies
Despite all the talk surrounding artificial intelligence, media agencies still have a long way to go with it. Columnist Grace Kaye discusses where agencies are right now with AI and what we can expect to see in the future.
It has become commonplace for a media agency to boast that it’s “AI-driven,” when in practice this is hardly the case. Nor should it be: AI, as far as it has come, is still at least decades away from reaching the level of intelligence necessary to run a successful digital media campaign.
In the name of transparency, it’s time for any guilty media agencies out there to start telling it like it is.
How AI-driven is digital media at the moment?
Experts within the field of AI might dispute whether AI is used at all by media agencies. However, it’s probably accurate to describe the algorithms used in ad tech platforms such as AdWords and DoubleClick as being “narrow” AI — machine intelligence is used to rank search results, or in deciding how to serve programmatically bought media.
Optimizing bidding strategies and processing performance data might, in some cases, benefit from machine learning — though this is not to be confused with “deep learning,” which actually replicates the way the human brain learns. My employer, Brainlabs, has occasionally used machine learning to improve bidding, or to fill in the gaps in customer data using modeling. But rest assured, it really wasn’t that amazing, and most of the time, scripts can be written that do an equally good job.
Automated media buying? Nah.
As for replacing human staff with robots, this experiment has already taken place. Albert, the automated media buyer, received a lot of media interest, but obviously, it’s no threat to the flexibility, strategic power, cultural understanding and general intelligence of a human account manager.
How AI-driven might it become in the near future?
There are two possibly important developments on the horizon, though, of course, there are plenty of other possibilities beyond these.
Audience targeting in programmatic
Big data needs big analytics, which deep learning might be able to crack — one day. Supposing an AI was developed that could use deep learning to process data in such a way as to identify patterns in consumer behavior, and over time, learn from previous experiences to develop something akin to industry know-how — that could be a pretty big deal.
Using audience data to improve relevance is one of the major challenges of programmatic and one of the main incentives for advertisers using it. An AI that could crack big data analytics would enable its owner to target audiences with a precision beyond what humans are currently capable of — so watch out for that. And good luck to Baidu.
Google launched Smart display earlier this year. Targeting, ads, creative — everything automated based on machine learning. It’s obviously not that great, but who knows if it will eventually get better? And why not extend this to all digital media channels?
I seriously doubt, as stated above, that even Google with its intimidating, ever-growing supply of data and annoyingly good advertising prowess will be able to provide a better service than a human + ad tech combo, but you never know.
Automation, not AI
Instead of waiting for a future that might never happen, the focus should be on technology we can harness right now. And so much of the most useful ad tech out there is centered on automation, whether third-party or in-house.
Thousands of tasks, beyond the actual purchase of media inventory, can be automated within a media agency. Wherever possible, automation should be used not to replace humans, but to allow greater focus on the aspects of the job that require human brainpower.
An AdWords account manager, for example, needn’t spend hours manually writing the hundreds of thousands of possible keyword variations that some campaigns require. Instead, with a little bit of code, this sort of campaign can be generated in a few minutes.
Reporting, analyzing data, optimizing bids — these are all processes that can be formalized and, to a large extent, automated. You still need humans to operate these processes, to develop the means of automating them and to continue to optimize them. We have built countless scripts and pioneered a lot of software development for the purposes of efficiency and improving performance.
No AI out there can offer more than this — the strategic aspect of digital marketing is still very much the domain of humans. Anyone claiming otherwise would need to offer a one-size-fits-all approach, when experience has always shown that every client requires a unique approach.
There may come a time within this century when there’s AI sophisticated enough to outperform humans in media planning and buying, but until then, it’s the job of any good agency to apply human intelligence in their use of automation, machine learning and tech innovation to drive digital media forward.
It’s worth remembering that machine intelligence is always preceded by human intelligence: We will be the ones who lay the foundations for AI-driven digital media, and we still have a lot of work to do.