Digiday+ Research: Marketing workflows benefit from AI, but trust is still a barrier to adoption

Overall, the marketers who are incorporating AI into their workflows are seeing the benefits of the tech, but issues with trust and complexity are proving to be barriers to widespread adoption among marketers.

This is according to a Digiday+ Research survey conducted in the fourth quarter of 2025.

How marketing workflows are benefitting from AI

When Digiday asked brands and agencies which specific types of AI technology they’re using in different workflows, the survey found that generative AI has higher rates of adoption than predictive AI.

Generative AI uses advanced machine learning techniques to produce new content like text, images, code and audio, while predictive AI uses data analytics to forecast upcoming events such as customer behavior, potential fraud and purchase probability. Predictive AI has existed longer than generative AI, but survey results suggest that generative AI is more applicable to marketers’ workflows.

Digging deeper, Digiday’s survey found that marketers use generative AI the most for creative production (82% of respondents said they use generative AI in this way), followed by marketing (at 81% of respondents), and external and internal communications (at 75% and 56%, respectively).

In one example of what this looks like, Unilever partnered with marketing services group Brandtech on its Beauty AI studio, an in-house AI system that creates assets for paid social, programmatic display and e-commerce. The tool has reduced creative production time, according to Selina Sykes, global vp and head of marketing transformation, beauty and wellbeing at Unilever.

“It’s a different way of working. We used to send briefs off and get content back. Now it’s this agile, iterative approach,” Selina Sykes, global vp and head of marketing transformation, beauty and wellbeing at Unilever, told Digiday. She estimated in the summer that, on average, Unilever was using the system to create 400 creative assets per product. “Before, we’d be doing 20 assets per campaign, and now we’re doing hundreds,” she said.

Digiday’s survey found that brands and agencies use predictive AI most for measurement and KPI analysis. Forty-eight percent of respondents said their company uses predictive AI for this purpose.

For example, Kroger Precision Marketing Powered by 84.51°, the combined division of Kroger that houses its data science and retail media divisions, uses predictive AI for KPI analysis. The company recently launched an AI-generated email digest for its suppliers that takes the form of customizable weekly reports that analyze both short- and long-term performance through several KPIs.

Marc Maleh, global CTO at design and technology agency Huge, said the company has expanded its use of AI tools to include both generative and predictive AI. “Design is using it for image generation. Within our creative department, we’ve gone from tools not just being used for pitch decks, but actually being used in production,” Maleh said. “Our data science and data engineering team built a product called Live and within Live we’re creating digital twins that learn, adapt and guide you about a brand. So, using it [AI] for rapid insight, strategy and persona development and concept testing.”

However, Maleh added that AI has yet to be used to its full potential. “As an industry, we have not fully embraced them [AI tools],” he said. “Whether it’s writing emails or similar activities, the human touch still goes a long way.”

Eric Lee, CTO at creative technology agency Left Field Labs, said that AI tools often overlap in their applications, and that, at the end of the day, brands are looking for results. “Increasingly, the lines are blurred,” Lee said. “The distinction between a predictive model versus what is run through a language model — there’s a lot of crossover. People largely don’t care what’s under the hood. They just want it to work. Whether it’s a language model or another model helping them get there, what’s the right tool for the job in the most effective way?”

Digiday’s survey found that marketers are using AI less frequently in workflows that include media buying and planning and financial analysis. About one-third of respondents said they use predictive AI for media buying and planning (35%) and financial analysis (34%). Meanwhile, only around a quarter said they use generative AI for media buying and planning (25%) and financial analysis (21%).

But some industry executives have said they think AI has potential to be used more for media buying and planning. At the Digiday Media Buying Summit in October, for instance, industry pros discussed ways in which AI could be applied for media buying and planning. “Just starting with an AI tool to analyze the [request for proposal] gets a good summary for the team,” said one agency executive who spoke with Digiday on the condition of anonymity. “In certain cases, with tight deadlines, it could be a very good first crack at what we might be able to do. And who knows, it might come back with something you didn’t even think about … I don’t think I’d rely on it to write the RFP, but it’s a great sort of first look.”

Adam Simon, former managing director at IPG Media Lab, said smaller brands and agencies might benefit from using AI tools in the media buying and planning process more so than their larger counterparts. “Meta and Google have great platforms for helping to automate A/B testing and optimize targeting, spend and creative,” Simon said. “You may not see that on a large agency or brand, but it’s transformative in small businesses and marketing teams where one person is trying to manage campaigns across multiple platforms, and they don’t have the time to commission and plan out elaborate A/B testing.”

Trust and complexity issues are barriers to widespread agentic AI adoption

Agentic AI has become a real buzzword in the marketing industry. But Digiday’s survey found that the technology has a ways to go to catch up to predictive and generative AI when it comes to marketer adoption. More than half of survey respondents (54%) said their companies don’t use agentic AI within their workflows.

Agentic AI and AI agents use the components of traditional and generative AI to anticipate needs autonomously and execute against them. They can learn from experiences, adjust future actions based on new information and take action on a user’s behalf, rather than relying purely on human commands like other forms of AI.

For example, digital agency Monks recently partnered with technology company Nvidia to produce an experimental 30-second film for athletic apparel brand Puma that was created entirely with agentic AI. Wesley ter Haar, co-founder and recently appointed chief AI officer of Monks, described the process to Digiday. “It’s written by AI agents. It’s mood-boarded by AI agents. The director of photography is an agent, and those agents have worked together to pull that script together, pull that mood board together,” ter Haar said.

Despite such use cases, there are still a number of barriers facing agentic AI when it comes to widespread adoption among marketing teams, especially compared with predictive and generative AI. One of those barriers is trust. Agentic AI runs without human feedback, which requires that users of the technology trust that AI agents are running tasks correctly. If agentic AI hallucinates in one step of a process, subsequent steps will be affected. (AI hallucination refers to when an AI application generates false information that it interprets as factual.)

Huge’s Maleh said marketers need to carefully weigh which tasks agentic AI is capable of completing correctly on their behalf and what types of data should be shared with the technology. “There is a governance concern, especially with agentic and data access,” Maleh said. “What do I want to give an agentic AI the rights to do on my behalf? … You may want an agentic experience, but what are your beliefs in regard to the data layer and access to that data? What do you want to allow the end consumer to take control of?”

Eric Lee, CTO at creative technology agency Left Field Labs, said using agentic AI for smaller tasks builds trust and comprehension over time. “Over the past year, we have seen more micromanaged agents, where you have an agent that can do a lot, but a layer of education and trust needs to be built before people are ready to say, ‘Yes, go manage my inbox, or execute this code for me,’” Lee said. “Building up that trust is a process, just like it would be with a human at the end of the day.”

Beyond building trust, a more critical barrier to furthering adoption of agentic AI among marketers is the complexity of the technology itself. Because agentic AI runs multiple tasks autonomously, rather than functioning as a single application, it often must be linked to multiple other tools like APIs and even LLMs.

“We’re also emerging from a mindset of collective extraction — let’s get as much information as we possibly can. That’s the old paradigm,” Left Field Labs CEO Sarah Mehler said. “Now it’s about what’s the right amount of information necessary to complete this for agentic AI to be valuable. Agentic AI is really hard to do, and not everybody can. There are very few examples of it actually working well in the world.”

Matt Maher, founder of independent research and development firm M7 Innovations, said with the fast-evolving nature of AI applications, marketers need to prepare their in-house teams now for future iterations of AI technology. “AI has moved from being an assistant to being more of an operator. We see this with ChatGPT’s Atlas, Perplexity’s Comet and Gemini in Chrome,” Maher said. “But how does a brand prepare its website for an agent. … Front of house, it’s making sure you’re agentically ready. Back of house, it’s how you can stack things [tools] in different ways to make you more efficient as a marketer.”

For more, read the full report.