AI and automation are not new to the media industry. For decades, media professionals have relied on automated bidding, programmatic buying, dynamic creative optimization, audience segmentation, reporting dashboards, and machine-led optimization to improve planning and activation. What has changed in the past couple of years is the acceleration of agentic AI which are systems that do not just respond to prompts, but can plan, execute workflows, connect to tools, and automate multi-step tasks. This has created understandable excitement around cost and time efficiencies, especially in operational areas such as reporting, admin, research, competitive analysis, trafficking support, and data summarization.

The productivity potential is real. However, the risk begins when companies treat AI not as an enabler of better work, but as a replacement for people. Across industries, AI has increasingly been cited in restructuring conversations, while the U.S. media industry alone recorded over 25,000 job losses in the past year, with generative AI identified as one of several pressures affecting employment. This is where the real challenge begins.

The first challenge is that AI is often more expensive than companies expect. Many organizations assume AI will behave like traditional software: buy licenses, deploy seats, and scale usage without a major cost impact. In reality, AI – especially agentic AI – often scales on usage, tokens, compute, storage, orchestration, monitoring, and integration. A recent analysis on enterprise AI costs highlighted how companies such as Microsoft, Uber, Nvidia, Meta, and Amazon have had to rethink AI usage because productivity gains came with rapidly growing bills. The article argues that when AI tools move from light assistance to deep work – planning, testing, writing, researching, and iterating – the token meter accelerates quickly.

This is even more relevant for media agencies, where AI use cases may involve large volumes of campaign data, audience taxonomies, performance reports, creative variations, brand safety checks, and cross-platform analysis. Agentic systems can also run persistently, consuming compute, memory, logging, and orchestration resources even when they are not visibly “working,” which makes cost governance critical. In other words, AI is not automatically cheap labor; without controls, it becomes a new operational cost center.

The second challenge is more strategic: AI can produce black-and-white answers in one of the world’s most colorful industries. Media is not only about efficiency; it is about culture, context, human behavior, timing, creativity, and differentiation. If everyone uses the same tools, similar prompts, similar datasets, and similar optimization logic, the industry risks moving toward a “race to the mean,” where outputs become generic and campaigns lose distinctiveness. This is exactly why philosophies such as UM’s Full Color Media are valuable. UM describes Full Color Media as an approach designed to help brands “Stand Against Bland” in an AI-driven world, combining human ingenuity with data to understand each brand’s unique patterns of growth

The conclusion is not that AI should be avoided. Quite the opposite: AI can and should automate repetitive, time-consuming tasks so that smart people can spend more time doing smarter things. Reporting, research, data cleaning, summarization, and workflow automation are clear opportunities. But media remains one of the most innovative industries in the world. The goal should be to use AI wisely: improve efficiency without eliminating creativity, reduce manual effort without reducing strategic thinking, and build bespoke solutions for clients rather than defaulting to generic machine-made answers.