Meet the tech partners adland is relying on to further its AI plans

In plainer terms, the agency is getting preferential access to Oracle’s powerful cloud and AI infrastructure, enabling it to create sufficiently powerful AI solutions for clients (generative AI at scale uses a tremendous amount of computing power).

According to Jay Jackson, vice-president of AI and machine learning at Oracle, the collaboration will deliver “the best of both companies’ expertise to help customers across all industries bring their AI services to market and overcome the multitude of challenges they face”.

Code and Theory isn’t the first to follow this route. Most of the major agency groups have some form of partnership agreement with big tech companies, with the aim of sharpening their edge in AI or creative automation, for example.

Omnicom, which owns networks BBDO, DDB and TBWA, is a case in point. It recently began offering clients use of Content Supply Chain, a software solution developed with Adobe that provides “connecting tissue” across the entire advertising workflow, from planning and production through to execution.

Linked to the company’s proprietary data unit Omni, it is primarily an organizational aid to clients with integrated accounts across its agency businesses (‘partner’ is a sleight-of-hand since Omnicom is licensing the product as an enterprise customer).

“The Content Supply Chain solution from Adobe allows us to drive far greater efficiency, far greater effectiveness and ultimately better engagement with consumers,” Paolo Yuvienco, Omnicom’s chief technology officer (CTO), tells The Drum.

Content Supply Chain is intended for the prosaic business of cutting time and costs. But Yuvienco points out that Omincom is a long-standing partner of Adobe with access to early-stage versions of upcoming products – including its AI tools.

”The crux of the partnership we’ve forged with Adobe… [is] how we get a first mover advantage on things like generative AI capabilities from Firefly and Sensei [Adobe’s AI products], getting true integration with our operating systems and processes, and the work management layer that comes from Content Supply Chain,” he says.

“We sit on the Adobe partner advisory board, so we get early access to some of the solutions they’re prototyping. AI is going to transform our industry and, ultimately, everyone’s industry. Early access to the teams that are driving Firefly and Sensei and coupled with our partnership with Microsoft and some other vendors… means we’re bullish on our ability to execute that fast.”

Despite reports that BBDO’s leadership had asked staff to refrain from using ChatGPT or Midjourney in their work, Yuvienco says the firm is “running towards generative AI”.

“We do believe that this is something that is going to really change the way we work. And we want to be right there at the forefront on defining those new ways of working.”

Developing working AI solutions in partnership with Adobe and Microsoft-owned OpenAI, he says, provides a safer environment – legally and ethically speaking – than experimenting live. Yuvienco says Omnicom wants to take “a very cautious path” so that it avoids issues surrounding copyright or sample bias in the medium term.

“This is precisely why we forged the relationship with Microsoft, which is basically providing us access to the base models from OpenAI. These aren’t the models that scoured the internet and learned everything and anything that’s available, it’s effectively the intelligence layer. Our intention is to train that on our data and on our clients’ data so we can drive higher degrees of intelligence and automation.”

Yuvienco says that Firefly, which was trained on copyright-free media and assets, is considered a safer option than competitors. “We know that it is safe from the legal aspect of it. And human intervention is still required; it’s still in beta, frankly, and we have teams of people who are effectively running use cases against Microsoft, Salesforce and Adobe’s products.”

Omnicom’s digital ‘center of excellence’ is next looking to implement an Nvidia toolkit called NeMo-Guardrails that adds programmable guardrails to conversational AI, which could exclude certain types of data from a training set for example. “Nvidia is at the forefront of AI… the type of processing you need to actually do learning models requires GPUs, which is their core product; most of the large language models that get trained are using cards like Nvidia’s cards.”

Guardrails, Yuvienco adds, will “basically help facilitate a safe, and purposeful use for AI”.