Almost everything you read about AI is written by someone with something to sell you — a tool, a course, a subscription, a future. I don't. I run an agency, and I have to make this stuff actually pay inside real businesses with real margins. That's a very different vantage point, and it changes what you say.
So let's start with what AI can't do, because that's the part the vendors skip. It can't run your business. Anything sold to you as 'fully autonomous' — an agent that books your meetings, answers your customers and grows your revenue while you sleep — is, today, mostly a demo. It works in the controlled clip and falls over in the messy reality of your actual inbox. It can't hold the context that lives in your head: why this client is sensitive, why that supplier gets a phone call not an email, what your brand would never say.
What it can do is far less glamorous, and far more useful. It is the best first-draft machine ever built. Proposals, briefs, job specs, FAQ pages, the meeting notes nobody wants to write up — anything where a competent starting point saves you the worst twenty minutes of the job. It's a tireless researcher: summarise this report, compare these three options, pull the key points out of an hour of transcript. And it quietly eats the admin that clogs up a small team's week.
Notice the pattern: the wins are all in tasks that are repetitive, low-stakes and easy to check. The failures are all in tasks that need judgement, carry risk, and are hard to verify. That line — between 'draft it and I'll check it' and 'do it and I'll trust it' — is the whole game. Most of the value is on the safe side of it, and most of the disappointment comes from dragging it across to the wrong side.
Which brings me to the thing nobody is selling, because you can't put it in a box: judgement. The businesses winning with AI right now are not the ones with the most tools. They're the ones with someone who knows which problems are worth pointing it at, and which to leave well alone. Buy ten tools without that person and you've bought ten more things to manage. Have that person and you barely need the tools.
If you're a small or medium business wondering where to start, my honest advice is boring on purpose: find the dullest, most repeatable thing your team does every week, and see if AI can get it to a checkable first draft. Don't start with the moonshot. Start with the chore. Get one unglamorous thing working, build the instinct for where the line sits, and let that earn you the right to the bigger bets.
That's the whole thesis behind how we work at Mynd Palace, really. The technology is the easy part now — it's a commodity, available to everyone, including your competitors. The advantage was never the model. It's having someone in the room who has actually run a business and can tell the difference between a tool that saves you a day and a toy that costs you a week.