At Adobe Summit, has agentic AI for marketers made its public debut?

The conference’s environs, in other words, are as larger-than-life as you’d expect from, perhaps, the maddest city on the planet. And though its themes are a more sober affair, they’re no more ambitious.

That ambition is simple: Adobe wants to cement itself as the leader in AI-enabled creativity across the entire marketing ecosystem.

Speaking at a pair of keynotes this week, the creative software company’s chief executive Shantanu Narayen declared a “golden age of design” enabled by AI, which has been rapidly “expanding the creative aperture” for marketers and designers over the last couple of years.

Adobe has been a major player in that rapid expansion of AI capabilities, most notably with its family of generative models, Firefly, which was announced two years ago and the focus of last year’s keynote. This year, the expected further developments of Firefly, including the launch of generative video, were nestled into a concept-heavy list of announcements as the San Jose tech giant sets out its stall for brands entering the AI revolution’s third phase.

Is agentic AI here?

The show’s splashiest announcement: 10 new ‘AI agents’ that the firm will roll out across its suite of products, underpinned by an ‘agent orchestrator’ (on which more later). That list of 10 includes an ‘Audience Agent’ designed to aid with campaign targeting, a ‘Data Insights Agent’ designed to connect up disparate data sources and a ‘Site Optimization Agent’ for website performance tuning.

Currently in beta testing with clients, the agents will be rolled out to customers in the coming months, with both how they’re commercialized and which products they’ll operate within varying across the range. Also announced: ‘Brand Concierge,’ which deploys that agent orchestration tech for customer service apps.

This hard launch into agentic AI is not unprecedented. In the rapid AI hype cycle, the tech has gone from speculative possibility to product catalogs in just a few months, with Adobe’s competitors across its myriad business areas announcing their own offerings, including B2B software giant Salesforce last September, data storage company Snowflake in November and Amazon Web Services (AWS) this year.

First, there was descriptive AI, then generative, now agentic. This third phase, in Adobe’s formulation at least, is characterized by significant upticks in AI models’ interactivity and reasoning capabilities. Most significantly, an AI agent is plugged in across systems to take action rather than making recommendations on a single plane.

Joe Stanhope, a principal analyst at Forrester, is not surprised to see things are moving so fast around agentic. “Two years ago, in the generative AI explosion, a lot of people got caught off guard,” he tells The Drum. “This time around, as agentic AI comes around, there’s a lot of people who have resolved to not get burned twice. So it feels like it has suddenly snapped into place.”

At this early stage, Stanhope says, approaches to and definitions of agentic AI remain varied. Salesforce’s approach (‘Agentforce’), for example, is more “platform-oriented,” offering a development environment for clients to create their own agents. What’s distinctive about the Adobe approach is its practicality: 10 agents aimed at marketers’ real problems.

In this way, Stanhope says, the Adobe approach is keeping one foot grounded in the existing AI reality: “These might not yet fundamentally change the way marketers work or what they do; it’s really about streamlining and making existing processes more effective and that should yield benefits. But as we build more of these agents and as you look at some [future] capabilities they’re laying out, like the ability to run multiple agents together and to connect to third-party agents and data – that portends a world where agents are far more automated and autonomous.”

Speaking with The Drum, Ely Greenfield, chief technical officer of Adobe’s digital media business, acknowledges that “a lot of people can’t even define what authentic is still.”

Personalization, personalization, personalization

But agentic and generative AI alike are just tools – tools that Adobe says it’s developing in pursuit of ‘personalization at scale.’ In his keynote address, Adobe head honcho Shantanu Narayen declared the company’s mission as “to change the world through personalized digital experiences.”

A further raft of announcements around the tech giant’s ‘content supply chain services’ aim at making marketing personalization available to a wide range of brands, both through AI-generated creative and through workflow and planning tools to enable the shift. It’s a shift that, according to the conference, is already well underway. In one keynote speech, Hilary Cook of the hotel group Marriott said that the firm has already multiplied instances of personalized content by a factor of 4,200, with content variations in the tens of millions.

Numbers like that, and the pace at which they’re achieved, are the sorts of stats that make some marketers nervous. Not least those working in reversioning, translation and marketing execution. This is a concern that a number of speakers at the conference are tackling head-on, developing the theme of AI as a copilot that will free up its human collaborators to think more strategically.

So, while the easiest ways to understand the capabilities of such tools are quantitative, Camila Nakagawa, global executive vice-president at Havas’s ‘content at scale’ agency, Prose on Pixels, argues that we should focus instead on the qualitative improvements it can bring to heavily iterated content. “It’s not about creating more but better content”, Nakagawa says. “Personalized content unlocks creativity. It’s a big opportunity for our teams to spend their time and effort on what’s really worth it… [We work with] retail brands that have to generate hundreds or thousands of different posts. I’m sure our copywriters would rather spend their time on something different. And when we talk about AI executing manual work on, say, resizing images, I think our art directors would rather spend their time and their knowledge working on something different.”

On the main stage, the chief executive of The Coca-Cola Company, James Quincey, spoke to the dual promise and trepidation around this dream of mass-scale personalization. “The cheaper it gets to make gen AI, the cheaper it gets for me to put you in the ad. The next stage is to say, ‘You looked at your phone but you didn’t buy the product’ and it goes back to a bunch of humans and they say, ‘What can we do?’… Well, if it’s cheap to redo the ad, I’ll just make a few changes, like an A/B test, and keep changing until you click ‘yes.’ So the lower the cost of the ad, the more I can set up gen AI plus an intelligent agent to chase you until you click yes. But at the other end of the spectrum will be a whole bunch of annoyed consumers going, ‘This is like Minority Report… I want to buy one of those apps that block everything!’”

Orchestrate this!

‘Orchestration’: not, perhaps, the sexiest word, but a candidate for the most oft-spoken word on the lips of Adobe reps this week.

If agentic AI is the firm’s big tech bet and personalization at scale is its promise to brands, this idea of orchestration is the gel between the two.

On the main stage earlier this week, the president of Adobe’s digital experience business Anil Chakravarthy said: “This is now the era of customer experience orchestration, where your content, customer data, your journeys, are all working hard with AI and so you anticipate what your customers want. You engage in natural language conversations across any channel. You respond in real-time and you continuously delight them.”

The vision is of agentic AI as a centralized intelligence between all of the channels, software and campaigns that make digital marketing so complicated, like a conductor in an orchestra. Speaking of all those software, what is perhaps most distinctive about this approach is the range of partners Adobe is bringing into the orchestra of which it hopes to sit in the middle. Announcements this week included a deepened partnership with Publicis through the latter’s digital consulting arm, Sapient, and its own AI suite, CoreAI, and with Microsoft’s AI ‘companion’ co-pilot. Even direct competitors are brought into the tent, with a “strategic collaboration” announced with AWS.

Longstanding partnerships across the ad business, including with behemoth WPP. So, while concerns in the ad world that the AI paradigm may shift ever more power to tech vendors like Adobe may well be exacerbated by seeing the latter’s AI agents in action, it is keen to strike a collaborative note. One such collaborator is agency Monks, whose global executive vice-president of martech platforms and innovation Remco Vroom says that agencies and brands still “have a big transformation ahead of them.”

The push for personalization and high-quality, high-volume content, Vroom says, may feel like it’s coming from inside the industry – but really, it’s being generated by real people’s sky-high expectations. “Everything needs to be these fast, super niche experiences. We’re getting all of these really great, faster-to-market, hyper-personalized services and we expect that from everything and everybody. So companies need to redesign what their role is in in the whole value chain of products and services.”

For Vroom, if the era we’re entering is one of potential marketing abundance – a hyper-personalized ad for every consumer, generated instantly, stored in an endless (if expensive) reservoir of data, iterated until ‘perfect – then the challenge for marketers comes down to exercising choice. Warning of one client that already uses only 20% of the assets it creates and the threat that unguided AI “makes things really vanilla – it gives you all shades of gray but it can’t give you pink,” Vroom says that marketers will have to think differently about how they engage and instruct AI.

“You shouldn’t, per se, do things just because you can.”