You simply can’t put the importance of AI into words — not without sounding like you’re selling something. But if you’ve ever used an LLM, or even a basic image generator, you know the feeling. You type an idea. You get an output. And it’s obvious the ground shifted.AI is everywhere now. Corporate tools, phones, creative apps, search. It’s not leaving. And it doesn’t feel like we get changes like this often. The last one that felt this widespread was the household computer becoming normal.But here’s the part that keeps getting repeated like it’s automatically true: AI is “making everything cheaper.”I don’t buy that.AI doesn’t make the work disappear. It makes the work move.
The fantasy version is pretty simple. You write faster, edit faster, produce more, and need fewer people. On paper, that sounds like savings.And in small ways, yeah — AI can help at the start. Draft a rough outline. Generate a first-pass script. Summarize a meeting. Transcribe an interview. Move faster through the messy beginning where you’re just trying to get something on the table.
But media work doesn’t get expensive because words are hard to type. Media gets expensive because you’re responsible for what the final thing does in the real world. You’re not paying for output. You’re paying for something that holds up.
In practice, “AI helping” often means you suddenly have more versions than you asked for. People generate options, then generate variations of the options, then change direction halfway through because the tool makes it feel easy to restart.It can also mean you get a rough draft that looks convincing enough to create false confidence. Everyone feels like the job is mostly done — until the first real review happens and you realize it’s not actually ready to be public.
Or it means the AI gets baked into workflow tasks: transcript, summary, tagging, routing, approvals. That part can genuinely be useful. But even there, it’s support work. It doesn’t replace judgment.The moment the output is public-facing, donor-facing, customer-facing, or tied to reputation, the entire project shifts from “can we generate it” to “can we defend it.”
The lazy narrative is layoffs. “We won’t need as many employees.”Sure — some tasks compress. Some roles get squeezed. But the cleaner truth is less dramatic and more annoying: AI increases speed early, which tends to increase volume, which increases review, which increases approvals, which increases time.
When you can produce ten drafts instead of two, you don’t magically pick the right one faster. You argue longer. More people weigh in. More stakeholders get involved. More “can we tweak this one line” moments appear. The project becomes a loop.Speed doesn’t remove cost. Speed multiplies decisions. Decisions multiply approvals. Approvals multiply time.
A lot of AI-driven “savings” are really just cost transfers.Instead of paying for careful creation, you pay for review and correction. Someone has to check accuracy. Someone has to make sure it’s on-message. Someone has to catch the parts that sound confident but aren’t grounded. Someone has to make sure it doesn’t quietly imply something you can’t stand behind.And then there’s brand risk. People can smell “cheap.” They might not be able to explain why, but they feel it. Once they feel it, trust drops. That’s a real cost — even if it doesn’t show up on an invoice.
A lot of AI-driven “savings” are really just cost transfers.Instead of paying for careful creation, you pay for review and correction. Someone has to check accuracy. Someone has to make sure it’s on-message. Someone has to catch the parts that sound confident but aren’t grounded. Someone has to make sure it doesn’t quietly imply something you can’t stand behind.And then there’s brand risk. People can smell “cheap.” They might not be able to explain why, but they feel it. Once they feel it, trust drops. That’s a real cost — even if it doesn’t show up on an invoice.
