Digiday+ Research: Marketers’ AI use rises, but tech skills stall

Digiday’s survey — which has been conducted annually since 2022 — found that marketers’ adoption of AI technology has risen significantly. In 2022, 44% of brand and agency pros said their companies were investing in AI technology. That percentage rose to 57% in 2023 and 71% in 2024, before hitting 86% in 2025.

AI’s growing importance for marketers has also become evident in the number of companies that have created chief AI officer positions over the past two years. In 2024 and 2025, brands General Motors, Mastercard and ZocDoc appointed AI chiefs, as did agencies Golin, Luckie & Co. and Horizon Media.

“In every industry there’ll be a percentage of companies that figure [AI] out, then a large percentage of companies that don’t. And I think the economic upside to figuring it out makes for such a big competitive gap,” Wesley ter Haar, co-founder and recently appointed chief AI officer of digital agency Monks, told Digiday in April.

Consumer adoption of AI has seen significant growth, as well, and, as a result, many brands are now regularly using AI in consumer-facing applications. PetSmart, for example, relaunched its member program using AI to tailor deals for customers based on their past purchases, and Guitar Center launched a chatbot called Rig Advisor to help customers select the right products to suit their needs.

Matt Maher, founder of independent research and development firm M7 Innovations, said that, while individual users may be comfortable with AI tools, companies generally aren’t using the tools to their full capacity. “Users are definitely more knowledgeable at a baseline level, but there is a delta between understanding tools like ChatGPT, which has 800 million weekly active users, and Gemini, which has 400 million a month, and then using them to their utmost potential,” Maher said.

“When a company adopts [Anthropic’s] Claude and uses APIs for all of its internal software, and [Microsoft] Copilot for essentially everyone, it feels like a big machine,” Maher added. “It’s almost a failure of imagination of how much you can actually use these tools if you push them to their limits. … Big tech isn’t great at showing people the amazing things they can do. … There’s still a gap, even though, at a baseline, we’re all getting used to AI.”

When organizations — including marketing teams — make moves to adopt AI technology before they’ve fully developed a plan on how that technology should be used, a gap often results between how much is being invested in AI tools and the return on that investment, according to Marc Maleh, global CTO at design and technology agency, Huge.

“You have massive investments on AI and the basics of all of them is the infrastructure layer — TPUs [tensor processing units] from Google, GPUs [graphics processing units] at Nvidia — somebody has to pay for that,” Maleh said. “Every time an agency or brand wants to deploy a model, the Googles and the Amazons of the world need to find a way to monetize the infrastructure layer, and all of the layers within the AI ecosystem.” Paying to use that infrastructure is becoming a greater concern for brands, Maleh explained. “What if I want to turn on 500 more seats of Claude code? What does that look like financially? Am I going to get that money back if I’m only getting a 30% productivity increase?” Maleh asked.

“There’s a realization about the economics of GPUs and TPUs because money is going into those things,” he added. “All of a sudden, those models have to get monetized in a real way. AI was the shiny object and continues to be that. Brands thought, ‘We want the press release, so let’s worry about the GPU and TPU charges and the API calls later.’”

Digiday’s survey found that the majority of marketers continue to implement AI technology into their workflows by using out-of-the-box AI tools, rather than building tools in conjunction with existing large language models such as Google’s Gemini or OpenAI’s GPT, or building and training their own LLM in-house.

Eighty-five percent of survey respondents said their company is using out-of-the-box AI tools. Less than half of respondents (40%) said their company is building proprietary tools with an existing LLM, and only 19% said they’re building and training their own LLMs.

The expense of building customized AI tools through an existing LLM or building and training a proprietary LLM, along with the learning curve associated with implementing either of these options, are the likely reasons why most marketers are choosing to use out-of-the-box AI tools. Smaller companies also may not be able to afford an AI team dedicated to creating custom tools.

Huge’s Maleh noted that several new out-of-the-box AI tools have become available to marketing teams within the past year. “What’s actually happened is there’s more available out-of-the-box models now,” Maleh said. “Whether it’s an Adobe or a Google, you can start with a base model that somebody else has already invested time in creating, and then customize it to your needs. That’s a lot of what Google’s Cloud Platform has with out-of-the-box tools like Vertex.”

Google Vertex AI is an AI development platform that uses Google Cloud’s infrastructure to let users build their own custom AI or machine learning models. The platform offers pre-built models that serve as a base for users to build custom tools and capabilities.

Maleh said another change to the AI landscape that has taken shape over the past year is a democratization of AI models and collaboration among some of the big industry players, such as Adobe and Google’s recent partnership in which Google’s Gemini, Imagen and Veo models are integrated into Adobe’s creative tools. “Now, if I’m using Adobe’s Firefly but I want to use Google’s Nano Banana as my asset generation model, I can do it within the Firefly console,” Maleh explained. “A year ago that wasn’t the case. It was Firefly or nothing. … We went from a place where a lot of platforms were walled gardens to where it’s more opened up.”

M7 Innovations’ Maher said that some tech companies are lowering the barriers around their AI services and allowing brands to build on top of existing features. “What I’m starting to see now is a tech stack,” Maher said. “There’s not gonna be one to rule them all. I’ve seen brands say, ‘Copilot is our base, and we stack Claude on top with a bunch of really smart APIs.’ Or, ‘We use Adobe Firefly to create and we have Canva to complement.’”

This can result in significant savings for brands, Maher added. “I’m starting to see cost efficiency gains — partnering with the bigger companies, but creating their own version or sandbox. … And you’re saving a lot of money because you’re not having to build it from scratch,” he said.