AI makes its mark at Havas Media Group

For instance, machine learning tools are currently being used to catch anomalous budget inputs within social campaigns to flag them for second-level approval.

So if a campaign is extended by a month, but the extra budget has been erroneously applied to the final two days of the current month, the tool can flag this error and prevent a month’s worth of budget being spent over a two-day period. 

“It’s sort of biddable operations,” said Mike Bregman, chief data officer at Havas Media Group. “We constantly have to think about the billions of dollars that we have in our treasury at any one time so the idea of using AI to help with that bookkeeping is key. We’re able to use AI to help keep track of all of that at a real scale.”

Elsewhere, machine learning is also being used to recommend how much to spend on a given day based on a week or month level budget, as well as the seasonality and past spend patterns for that specific brand. 

Both examples speak to the biggest benefit the agency has reaped from AI to date: speed and time savings — hours of human work reduced to minutes or seconds. 

Take the aforementioned case of catching anomalous spending decisions, for example. Not only did it generate upwards of one million dollars a year in savings, according to Havas Media Group, it also freed up hours — time execs at the agency could spend working with clients or with one another.  

“The agency model has to be futureproofed and in order to do that there are some fundamental changes to the way organizations are designed, the way they operate and the technology that helps them do that, which need to be addressed,” said Bregman. “AI has the ability to unlock all of that.”

Creative — or rather the way it is optmized — is another process AI has permeated through at Havas Media Group. Nowadays, any dynamic creative optimization-driven campaign deployed in and around the purchase funnel by the agency has AI in the background. 

“These AI platforms can come up with different combinations of messages in a few seconds when it would take a copywriter a few days,” said Bregman.

Copywriters should breathe easy. Havas Media Group isn’t planning to replace them outright. Rather, it’s using AI to handle certain tasks. 

As Bregman explained: “it’s still smart to have a human that’s reading all the iterations, because some of them might just be ridiculous or outside the norms.”

This is just the start. 

Over the next two to three years, Havas Media Group wants AI more deeply embedded into the fabric of the agency. So not just using AI to optimize all media buys but also to create custom algorithms within a programmatic bidder or using the technology to identify the right training modules for a planner as well as making manual tasks like filling out timesheets.

Some of these efforts are already well underway, like for lower funnel performance optimization. Havas Media Group’s execs are using AI to power custom algorithms in the demand-side platforms — or ad tech used to place programmatic bids — to get 25%+ improvements in the cost per acquisition charged when the ads it buys lead to a conversion.

The application of AI is expanding to broader metrics like attention, where it’s used to model consumer interactions with ads — which is important as the industry moves toward attention-based planning. 

Even from this limited vantage point it’s easy to see where this nets out in the short-to-medium term: campaigns planned and bought by AI. The tools are there already, it’s just whether marketers are willing to trust the technology with their money. The optimization, on the other hand — the bit that comes after the ad has been bought and the client wants to sweat those dollars harder — remains a work in progress. 

Bregman expanded on the point: “The idea of funding the fluidity [of campaign optimization] that comes as a result of AI is something that not a lot of advertisers are comfortable with.”

Eventually, however, this stance will change — whether marketers are comfortable with it or not. The emergence of AI solutions from Google (Performance Max) and Facebook (Advantage+) make that all too clear. Both solutions can be whittled down to this: advertisers share their data with the tech, upload the creative assets, set the financial parameters (think daily budget cap, price per conversion etc.) and sit back and wait for it to report back aggregate results. 

It makes sense then the marketers are taking a trust but verify approach to many of these solutions. 

“For advertisers, they see us a bit like the United Nations when it comes to being able to look at all the different options on the table, test out the use cases of them and then figure out how to get the most out of them,” said Bregman. 

Fulfilling a role like this is tricky to say the least. Not only are the likes of Havas Media Group having to stay on top of thinking about what AI means for their own businesses, they’re having to help marketers do the same thing. And all while making sure they avoid AI becoming another existential crisis. No wonder the group is trying to gather up as much knowledge on the matter as possible. It’s not just building out its own already substantial team of data scientists, it’s also working with academia, startups and platforms too. Oh, and there’s the cost.

“We’re still trying to figure out the numerator denominator of the tech ROI equation, because none of this is cheap,” said Bregman. “Data scientists aren’t cheap and the ones we have here are very busy these days.”

It’s akin to being in an arms race, in many ways. 

Whether it’s Havas Media Group or another agency, they’re all trying to rewire their businesses around AI. The challenge is figuring out how fast they should do it. Move too fast, and the impact the technology has on the structure and subsequent operations of a holding group like Havas Media Group could prove to be too disruptive in a very transitory period. Go too slow, though, and the business could struggle to keep pace with counterparts that are more efficient and adaptive thanks to AI.

“I would love to find a way to make this pivot quickly but with tens of thousands of people to think about across our holding company, it’s going to be a journey,” said Bregman. “Agencies function through really good handoffs because there are so many different teams. The idea of having data at the core would fundamentally change how that works.”

That’s a big concept for any business to get its head around — let alone one where the data team still only accounts for a quarter of the group. That expertise either needs to find ways to permeate further and faster into the business or it needs to grow. Either way, the costs are steep. But they’re also necessary. AI is nothing but a data hog, after all. Indeed, the technologies thrive on ingesting large amounts of data.

“There are no AI-driven agencies today,” said Bregman. “I’m trying as much as I can to push us toward that direction so that we can be an agency of the future but we’re not quite there yet.”

Still, Havas Media Group has time to stick the landing on AI. Senior marketers haven’t even got to the point where they know what they don’t know — the preservation of agencies — when it comes to AI. AI hasn’t even come up in pitches, said Bregman. Eventually, it will, and when it does he wants to be able to respond accordingly.

“AI inputs should be viewed as a creative partner that can be leveraged to frame angels from a messaging perspective,” said Marc Hardgrove, CEO of search agency The HOTH. “But it takes a little practice to hone in which inputs work best for your team. At this point, there shouldn’t be a question as to whether to use AI– that horse is out of the gate and frankly, we all have been using AI tools for years, even before it was the big buzzword.”