AI won't split companies into users and non-users
AI won't split companies into those that use it and those that don't. Everyone is using it. It will split them into people who can spot a wrong answer and people who can't.
That's the whole post. Everything below is supporting evidence, checked by a human, which is rather the point.
The problem nobody budgets for
AI has made producing content cheap. Reviewing it hasn't got any cheaper, and that's where the time now goes.
no one has a budget line for "reading what the robot wrote". it's there. it's just hidden inside everyone's salary.
Take an AI-written audit. It has headings. It has a tidy summary. It has a confident tone, which is the one thing AI never runs out of. Then you read it. Some of it misses your context completely. Some of it is just wrong. A fair bit of it isn't even best practice, it only looks like it.
the company it describes sounds great. shame it's not yours.
You can't skim it either. Reading it is like proofreading a contract written by someone who's never met you. Every clause sounds reasonable. One of them sells your house.
The time saving is real. It just happened to someone else. The cost lands on whoever has to review it, which is usually your most senior and most expensive person. so friday afternoon is now fact checking a robot. strategic.
Why it all starts to look the same
Most companies are using the same few tools with similar prompts. The tools hand back the average of what's already out there. So every brand gets the same answer in the same tone.
It's a photocopy of a photocopy. The first one is fine. By the tenth, the edges have gone soft and nobody can find the original point. much like the last all hands.
Customers notice. They don't complain, they just stop reading, which is worse.
If your content could be swapped with a competitor's and nobody would notice, it isn't doing a job. it's just taking up room on the internet.
every brand is now excited to share something game changing. the game is exactly as it was.
Who loses
People who are good at prompts and have nothing behind them. Knowing how to ask the question isn't the same as knowing the answer. If you can't tell when the output is wrong, you can't use the tool. You can only pass it on.
level 1 character, level 80 sword. looks incredible in the screenshot. then the first goblin turns up.
Brands that hand strategy to automation. Positioning, the core message and what you stand for are the things that make you different. Giving them to a tool trained on the average is a very efficient way to become average. at least it's quick. you'll be mediocre by lunchtime.
Who wins
Real experts using AI for leverage. They read the output the way a mechanic listens to an engine. 10 seconds and they know what's wrong. The tool saves them time instead of costing it.
a cheat code is great if you know the level. otherwise you just fall through the floor faster.
Brands that are willing to hold an opinion. A view someone can disagree with is worth more than a summary everyone nods along to. an opinion is risky. so is being forgettable, but that one never shows up on the weekly report.
Organisations that publish what nobody else has. Their own data, their own results, real customers saying real things. A model can't average that, because it was never in the training set.
rare loot. doesn't drop from a prompt.
The paradox
When everyone can produce standard best practice content for next to nothing, it's worth next to nothing. Free things are commodities. so is any blog post that opens with "in today's fast-paced world".
What's left is the expensive stuff. Earned expertise, a point of view, and proof that something actually happened. The more automated the average gets, the more valuable the original becomes.
One warning for CEOs
Cheap to produce, expensive to trust.
If your team can't tell good AI output from plausible AI output, you haven't saved time. You've borrowed it, and the bill arrives when someone builds a decision on the part that was wrong.
Before anything goes out with your name on it, make sure someone who knows the answer has read it. there's no respawn, just a very long meeting.