Why ‘you don’t have to use AI’ is the new AI pitch

Watching Adobe’s keynote at its MAX conference in Miami this week, I couldn’t help but think about the tightrope it’s walking with AI, and how similar it is to what the media industry is going through. In both, there’s a desire to be AI-forward and skate to where the puck is going. At the same time, they need to deal with a hostile crowd—meaning the growing backlash against AI among many of their customers.

At the last two annual MAX events, AI was a huge emphasis. The company touted AI features, including ways to generate content directly in their Creative Cloud apps, such as Photoshop and Premiere. Adobe hyped up Firefly, which was then mostly associated with the company’s image models. In 2024, Adobe told the audience that “2024 will go down as the year that creativity changed forever.” In 2025, the headliners were a new Firefly image model and a lineup of third-party models from Google, OpenAI, Runway, and others, and “choice” mostly meant choosing which AI model to use. The lone nod to skeptics was a passing aside: “Of course, you can ignore it all as well.”

This year, the tone shifted. In fact, one of the first things incoming CEO Anil Chakravarthy emphasized during the opening keynote was that users didn’t have to use AI if they didn’t want to. “Some of you don’t want to have anything to do with AI,” he said, promising to give those users “the precision and control that you’re looking for, and the option to not use AI.”

Instead of a new Firefly model number, the keynote spotlighted upgrades with no AI in them at all, like a rebuilt color mode in Premiere and Dynamic Text in Illustrator. Firefly is still an area of focus, but it’s evolved into a catch-all creative studio, more like Figma than an AI model. (Disclosure: Adobe paid for my travel to MAX, though it had no input on this column.)

AI-optional doesn’t mean AI-free

None of this is to say Adobe is walking away from AI. Far from it: during that same keynote, Chakravarthy announced Adobe would be providing a way for AI services to access Creative Cloud apps directly through the model context protocol (MCP) standard. That means AI users will be able to have Claude, ChatGPT, or Gemini act as an agent, effectively “driving” Photoshop et al. to edit projects, create content, and export files directly on their computers. The integrations go into private beta in early 2027.

Adobe already provides an AI connector that works similarly, but the MCPs would take things to a new level of precision and control. Jesus Diaz has a great breakdown of exactly how they would work and what it means for creators. In addition, Adobe is building out its own agents: The new Creative Agent is essentially an Adobe chatbot where you can describe what you want and the agent handles the execution in the appropriate app. It’s now in beta inside Premiere, Illustrator, InDesign, After Effects, and Frame.io.

All this makes it clear Adobe isn’t slowing down on AI. It has, however, recognized that a great bifurcation is taking place, one that’s affecting all industries to some extent, but especially the media. On the one hand, you have your AI skeptics, customers who are suspicious of AI generally and AI-generated or -assisted content specifically. And on the other, you have the AI enthusiasts, the ones who are using AI to enhance and speed up their work, including through tools from other companies.

This is the AI tightrope, and all businesses will need to navigate it over the next few years. Not just because AI represents “the future” in some vague sense, but because, if it does, it opens up a new kind of customer.

Now everyone knows kung-fu

My business is actually a pretty good example of this. I use various software, including Photoshop and Premiere, to help produce content like The Media Copilot podcast. But I’m not a creator per se, and I’m probably using only about 10% of the features that would help my use case. For many, that skill gap is the difference between subscribing and not. But an agent that bridges the gap could mean Adobe customers are no longer just enthusiasts and experts. Suddenly we’re all Neo in the Matrix: just plug an agent into your workstream, and you know kung-fu. In fact, one of the keynote demos was more or less my exact workflow: an outside AI agent turning a long-form podcast into a batch of short clips for social.

There’s also an urgency here, and again my use case is a good example. I’ve already begun offloading some of my production process to agents. Claude Code and OpenAI Codex are both excellent at analyzing workflows and recommending software solutions that they (not I) can operate. Often, it’s open-source software that gets the job done at virtually no cost other than the tokens I expend. And Adobe users can even employ computer use—where the AI uses your machine, just as you would—to operate the software. It’s messier than an MCP, but it gets the job done, and AI is getting better at this by the day.

In other words, the AI world is moving toward agentic work, with or without Adobe. Businesses of all kinds are beginning to demand this kind of functionality, and even if open-source software is cheaper, there are advantages to using commercial software. Security, reliability, and auditability are all very important to enterprises. And it’s mutually beneficial to Adobe, since the number of licenses those businesses buy will no longer be limited by the number of experts they have. Notably, the MCP integrations require the apps to be installed and licensed on the user’s machine. Even when an agent is the one driving Photoshop, someone is still paying for the seat.

Media’s version of the same customer

There’s a corollary here in media. As people turn more and more to AI answers for information, it means bots scrape all kinds of sites to find the best possible answer. A consequence of this is that niche sites—in particular B2B media—have become some of the most sought-after sources on the web. TollBit’s latest State of the Bots report ranks B2B and professional sites among the most heavily scraped categories it tracks, with scraping up 62% from Q2 2025 to Q4 2025. 

Their specialized knowledge might sometimes be impenetrable to mainstream audiences, but never to AI agents. They can quickly parse the bits that are important and relevant, then break them down in a way that’s understandable to anyone. Just like with those Photoshop agents, lack of expertise is no longer a barrier to adoption.

But what about those AI skeptics? They’re an important part of any company’s existing customer base, and they’ve made their views clear: They don’t want to see AI in the content they consume. I’ve written about the problem with AI tells in text, and the same goes for images and video: audiences consistently say they don’t like them. That doesn’t mean all AI is bad, and many Adobe AI-powered features garnered lots of applause at MAX. But it needs to be in the right context. 

A good rule of thumb: AI behind the scenes is generally OK, but generative content you need to be careful with. The audience research lines up almost exactly. In the Reuters Institute’s Generative AI and News Report 2025, only 12% of people said they were comfortable with news made entirely by AI, compared with 62% for news made entirely by humans. But 55% were fine with AI fixing spelling and grammar.

Most media companies have arrived at a similar place. There are tantalizing opportunities for content to reach new audiences through AI search engines and agent protocols like MCP. At the same time, they need to tread lightly when implementing AI in their own operations, lest they lose trust with their existing audience. That’s the AI tightrope, and staying balanced is obviously the trick to moving forward. 

Getting the agents to pay up

But what’s on the other end? At least for Adobe, the answer is clearer: a new and potentially huge customer base who don’t want to use tools but are happy for their agents to. For media, the other end of the rope is murkier. The agents are already here, reading trade publications and national news alike, but most of them are still taking what they find without paying for it. AI apps send publishers a sliver of the traffic Google does, and the deals that would make up the difference are stuck in courtrooms and licensing negotiations.

Adobe found a way to charge for the agents using its tools. Until the media figures out how to get agents to pay for the privilege of reading content, it’s walking the same tightrope, just without a net.

内容来自 Fast Company,版权归原作者所有。

阅读原文 ↗

用 HomepageTab 新标签页关注 RSS,打开浏览器就能看到更新。

安装 HomepageTab