The rise of the Media Centaur
The AI revolution won’t be the end of the world for people in media agencies but only if we can overcome the ‘them’ versus ‘us’ mindset says Alex Concannon.
As artificial intelligence (AI) becomes mainstream, it feels like we are careering past a point of no return. The data-driven approach to media buying we have proudly cultivated is going to render us all obsolete as we are replaced by computers. Or will it?
The impact of artificial intelligence on our workforce looks bleak. No industry is deemed ‘safe’. Two academics from Oxford University concluded nearly 50% of all jobs in the US were at risk from computerisation while a Forrester forecast predicted 7% of these jobs would be gone by 2025. When Elon Musk addressed the World Government Summit in Dubai, he posited that AI won’t just take our jobs but a jobless future would destroy our identity and sense of purpose. Time to bail, right?
I’m more optimistic about our prospects, however, we need to reframe the issue at hand. We will never win the zero-sum game of who does the job better, man or machine. Computers can do many things quicker, cheaper and more efficiently than us. We need to stop focusing on who does the job better and work out how to do it best.
‘Best’ comes from removing the dichotomy of ‘them’ or ‘us’. By opening ourselves up to symbiotic relationships between man and machine, we get better baseline performance supercharged with human ingenuity.
This isn’t just wishful thinking but a proven concept. After chess player Garry Kasparov lost to super computer Deep Blue, he hypothesised the machine didn’t beat him because it was a better player but because it had instantaneous access to a database with all the possible moves in chess. If Kasparov could tap into the same information, he believed he would win, a concept that gave birth to a new sport, Freestyle chess. These results offer a path into the future: today’s best chess player is not a super computer but a hybrid team of programs and people called Intagrand. While the programs provide informed recommendations regarding what move to play next, the human players assess new situations and overrule the machine when appropriate.
This fusion of human and machine is known as a ‘Centaur’, calling on the mythical half human, half horse beast.
The machines need us humans
A limitation of artificial intelligence is its ability to deal with unpredictability. Any data that underpins a machine is ultimately flawed because it is based on something that has already happened. In rapidly evolving environments like media, robots will still struggle to differentiate anomalous results from immediate shifts in the environment. Machines will be challenged to decipher whether outliers are lone events to be ignored or something they need to act upon.
Media Centaurs, on the other hand, can leverage automated consul from their AI partners and make the final decision on channel plans and the optimisation of campaigns. Again, this is not new behaviour as we already leverage recommendation engines to identify new products, music and movies without them becoming prescriptive. We’re simply applying the process to a new situation.
The best media agencies of the future will not be the ones that automate everything with algorithms and machine learning but the agencies that embrace this concept of Media Centaurs.
Limitless possibilities
If the above scenario is how we ‘know’ we will be able to work with robot partners, the unknown potential is even greater. The people who will thrive as Media Centaurs will be the ones who don’t have fixed notions about what human-machine collaboration looks like. They will develop new skills and adapt to the opportunities as they arise.
While still in relative infancy, marrying human understanding with the power of data is critical to how we approach our client’s business. Data must inform but not prescribe.
With IBM revealing earlier this year it has started to integrate Watson’s artificial intelligence capabilities into media buying, and brands already replacing agencies with AI, it’s easy to jump to the conclusion we’ll all soon be out of a job.
Crucially, the impact of AI on our industry is not certain so long as we continue to challenge how we view human-machine partnerships and identify how we can add value to AI instead of the other way around.
Alex Concannon is the national analytics director for Maxus Communications Australia
Alex, I love the article and the premise it delivers.
I especially agree with “removing the dichotomy of ‘them’ or ‘us’” to achieve the best results for our clients when referring to AI.
However—and I know you’re not the one to come up with this term—‘Centaur’ and these new ridiculous terms that are rising up need to be squashed.
If agencies that thrive are to become ‘media centaurs’ then that would assume that ‘agencies’ will die out.
Meaning all the ‘media centaurs’ will really just be media agencies.
So, why don’t we just call it like it is, and say agencies need to adapt to AI and use it’s power as leverage and leave it at that.
Your analogy to Intagrand kind of makes sense, but kind of doesn’t at the same time…
Chess is played for the pleasure of it—so both Intagrand and ‘normal’ chess can coexist completely fine.
This industry isn’t done for the pleasure of it, so the most effective forms will rise to the top and stay that way.
I guess what I’m saying is, why can’t we (as an industry) just provide information on trends without trying to position our advice as ‘sage’… aim to do it in more common speak.
i.e. use AI to leverage human ingenuity, rather than saying, let’s be half-human, half-robot amalgams called ‘media centaurs’ because that’s the only way we’ll survive. <– exagguration added.
Good article, for the most part though.
Cheers, James
User ID not verified.
We are so far from this reality. For the industry to get anywhere close we need –
Open access to all networks’ data (very few are “allowed”)
The ability to use that data how we wish (OZTAM, CRA, etc don’t let anyone fuse their data with others)
A decentralised marketplace of inventory (Yeah like that’s going to happen)
The ability to connect to traditional publisher playout systems (some of them are still DOS based)
The integration with everyone’s favourite (not) media billing system, that really only handles old hat TV spot buying.
And the integration with client systems for access to sales data. (Biggest LOL ever)
We can’t even get “programmatic” clearly defined for cross channel buying. So how can we apply AI? Sure there are exceptions in silos in some areas (mainly digital) but all this talk of algorithmic buying and automated cross channel schedule recommendation and optimisation coming in the near future is just bloody bollocks. Sure it is possible. And when it does happen, it will absolutely be the death of media agencies. But until then we are so far away from being Media Centaurs.
In fact, we are quite simply, still Media Neanderthals.
User ID not verified.
Amongst the jargon you miss the whole point that there will be no need for media buying agencies. Media buyers are glorified procurement departments (+insurance) .. imagine what an objective AI software could do for due diligence and risk mitigation when auditing their automated media placements..
User ID not verified.
My point is the opposite and the idea of “objective AI” is idealistic but unlikely. There is a lot of research at moment into AI bias. Programs have to learn from somewhere and often there a biases within data sets or even the way an AI program has been constructed. Researchers are increasingly finding AI programs inheriting these biases.
Ignoring structure, I believe we will still need media planners and buyers to work with automated programs to deliver the best results. The evidence is there that while machine might be better than people, people + machine tends to outperform the computer alone.
User ID not verified.
AI is a misnomer for what will be deployed. Its closer to ‘expert system’ but thats a misnomer too: its no expert, its the codification of what we understand to be the attributes of a system, optimized as we see them.
Even genetic algorithms in the end, are directed to what is ‘best’. And deep net learning systems, the training depends on what we put into them.
Codification of behaviour over ad placement, targetting, is going to turn out to be less productive than people want. It’s basically a cunning trick to downsize, and de-skill placement logic in a belief ‘the algorithm is always right’
Judging by the placement of ads for things I know the system as a whole knows I just bought, and the continual placement of ads for female intimate hygene product and bridal wear, which alas does not come in my size or gender.. I am not impressed with our new AI ad placement overlords.
here’s a question: if AI is so smart, and google ad placement so smart, how come I continue to get calls (ie real people phone calls) from my google account manager, to ‘help me improve placement’ ? (btw, I know exactly what my placements are, and why they are doing what they do: they simply don’t accord with googles model of what they THINK I want)
User ID not verified.