AI Agents Put Creative Control Back On The Table

InnovationAIAI Agents Put Creative Control Back On The TableByRon Schmelzer, Contributor. Forbes contributors publish independent expert analyses and insights. Ron Schmelzer covers AI and data best practices at Forbes since 2018Follow AuthorJun 03, 2026, 06:35pm EDTJun 03, 2026, 06:49pm EDT--:-- / --:--This voice experience is generated by AI. Learn more.This voice experience is generated by AI. Learn more.Magnific CEO Joaquín Cuenca Abela at the Upscale Conference in San Francisco 2026Ron Schmelzer at the Upscale Conference SF 2026At the Upscale Conference in San Francisco, creatives focused on the emerging AI creative ecosystem are indicating that use of AI in the industry is becoming less a fight over AI taking over creative roles and more over wrangling AI to enable creative outputs and repeatable processes that realize people’s individual taste. In marketing, enterprise software, education and customer operations, the same question keeps surfacing: when machines can make more, faster, what remains unmistakably human? For creatives that are making the most intense use of AI, the effort is shifting from producing outputs to processes that ensure quality outputs that represent creative visions. Magnific CEO Joaquín Cuenca Abela shared that, “the goal is to help people “show others what is in your head. Greater control converts fear into curiosity.” That captured the general mood of the conference, which was neither overly enthusiastic or pessimistic about AI. For working creatives, AI is rapidly becoming a production necessity, but with challenges that extend beyond prompt outputs. The most irritating part of many AI tools is that outputs are a black box that wander, lack controllable reproducibility and often reinterpret creative direction. This is where the human gets reinjected into the AI process. Directors, designs and editors see the steps, catch the bad choices and steers that system back toward their creative intent. Shifting from Outputs To ProcessCuenca criticized the current crop of AI editors for taking too much creative license. Ask for one change and the tool may remake the whole piece. The output may look impressive, yet still miss the assignment. He likened it to asking AI for an answer to an exam problem when what you really wanted was help thinking through the problem. While AI generation image and video models continue to advance in capabilities, model outputs are only a small part of the total process. Cuenca described a new generation of agents that can work in loops, generate images and videos, review what they have made, then expose their process so a user can change any step. The company is tying those agents to its workflow “spaces” canvas, and reusable “flows” workflow processes that can be shared with teams or called through an API. In many ways this is moving the creative industry towards an AI creative OS. MORE FOR YOUThe focus is increasingly on process versus result. Adobe has moved in the same direction. At Adobe MAX 2025, the company introduced AI assistants for Express, Firefly and Photoshop, described as conversational, agentic experiences that let users create and refine work through language inside Adobe’s tools. Adobe has since pushed GenStudio toward what it calls an agentic content supply chain, connecting brand context, planning, creation, delivery and reporting. Global advertising, marketing and communications services giant WPP is making a parallel bet in advertising. In January 2026, it launched Agent Hub on WPP Open, an internal agent library built to package agency knowledge into reusable AI agents for client work. WPP describes Open as an AI platform for marketing that spans planning, media, production and commerce. The pattern is clear that AI is moving from “make me a thing” to “help me run the system that makes things.” In that model, a creative director’s value shifts from manual execution to judgment over sequence, brand fit, timing, image grammar, emotion and restraint. MCP Makes Creative Tools Part Of A Wider MachineOne technical choice that’s helping to simplify and enable more complex creative processes with AI is the Model Context Protocol (MCP), that provides an open standard for secure, two way connections between data sources and AI powered tools. MCP is enabling agentic tools such as Anthropic’s Claude Code, OpenAI’s Codex, and other tools to interact with tools such as Magnific’s flows and Adobe tools to incorporate creative and operational processes. If creative tools become callable from many AI interfaces, the front door to creative work changes. A designer may start in a chat interface, jump into a node based visual workflow, return to a team workspace, then expose the finished process through an API. The creative suite becomes less like a set of separate apps and more like a factory floor with shared machinery. Outside the creative sector, that same architecture is spreading fast. Gartner predicted that 40% of enterprise applications will include task specific AI agents by the end of 2026, up from less than 5% in 2025. McKinsey has argued that companies will need to rebuild workflows around agentic AI rather than paste agents onto old processes. The payoff comes from redesigning the workflow, not merely adding a chatbot to the corner of a screen. The Labor Question Has Not Gone AwayWith AI moving into more of the creative process outside of just outputs, there are anxieties in the creative economy. Brookings research on online freelance work found that freelancers in occupations more exposed to generative AI saw a 2% decline in contracts and a 5% drop in earnings after new AI software arrived in 2022. The World Economic Forum’s Future of Jobs reports that employers expect 39% of workers’ key skills to change by 2030. So impacts on the creative economy are already here. The risk is that companies treat creative AI agents as a headcount reduction tool, then discover too late that faster production without sharper judgment makes weaker work at higher volume. Anyone who has seen a brand flood social channels with lifeless variations knows the trap. More assets do not equal more meaning. At the same time, high powered creative executives from Netflix, Amazon, Apple, and other companies present at the event are showing that the human element is being injected increasingly into AI-integrated processes. For these firms, prompt writing is becoming table stakes. The higher skill is workflow design. Creatives will need to understand how a concept moves from brief to references, assets, variants, approval, localization and delivery. The person who can encode that path into a reusable flow will have leverage. In addition, creativity and taste will become more encoded into AI processes. Teams are increasingly building brand agents, style bibles, model preferences and organization-specific workflows. This also means that collaboration and process control will matter more. The future of creative AI may hinge on whether teams can see what agents are doing, correct them and reuse the parts that work. Emmy-nominated filmmaker Eliza McNitt at Upscale Conference 2026Ron SchmelzerThe strongest idea from Upscale is that AI will expose the difference between output and intent. A machine can now easily create a plausible image or video in seconds. It can extend a scene, age a character, combine video clips, add music and build a campaign variant. The harder question is why that image should exist, what memory it carries, whose story it serves and what should be cut. Cuenca introduced a short film produced by his company called “Candela” focused on his personal reflections. The point of showing the movie was not to show the quality of the outputs, but to show that a specific creative vision with editorial discernment and thousands of creative curation decisions can produce a compelling result. That is where the creative fight is headed. Not toward a future where every artist becomes a machine operator and also not a rejection of AI and return to pre AI craft. The new role focuses on the core of creativity and the intent of the creator. Editorial StandardsReprints & PermissionsLOADING VIDEO PLAYER...FORBES’ FEATURED Video