A deeply moving story about asking a machine for clarity and receiving a confident paragraph-shaped fog.

A business autopsy, conducted by the person who made the decisions.

The machine was confident, so naturally I trusted it

When AI became the answer to everything, I had questions. Lots of them. What should I sell? Who is my audience? How do I market it? Why has nobody bought? Could you make that last answer sound less personal?

AI never looked tired. It never said, ‘I don’t know your customers.’ It gave me ten-step plans in under ten seconds, complete with market positioning, audience personas and a content calendar that assumed I had a design team, a video editor and the emotional stability to post three times a day.

The answers looked organised. That was the danger. A neat list can make a guess look like a strategy. Add a bold heading called Growth Levers and suddenly you are not confused—you are pre-revenue.

AI TRANSLATION SERVICE ‘Leverage omnichannel synergies’ often means ‘post everywhere and hope your personality survives.’

 

What AI knew about my business

AI knew whatever I told it, plus a large amount of internet-shaped probability. It did not know which product made someone pause, which part of my checkout felt suspicious, whether my audience had money, or why a person watched my reel twice and still chose peace over purchasing.

It could produce a customer persona called Priya, 28, urban professional, values convenience and sustainability. Priya had a salary, hobbies and pain points. What Priya did not have was proof of existence.

This is where generic advice becomes expensive. If the input is ‘help me grow my dropshipping store,’ the output can sound intelligent while remaining useful to approximately every store and therefore none of them. I kept asking for better words when I needed better evidence.

The AI advice I should have questioned sooner

·     ‘Post consistently.’ Useful only after deciding what the posts are meant to prove, change or sell.

·     ‘Create urgency.’ A timer cannot rescue an offer nobody wanted before lunch.

·     ‘Target a broad audience first.’ Broad audiences are excellent if your marketing budget belongs to a large corporation or a fictional person.

·     ‘Build brand awareness.’ Pleasant phrase. Difficult to deposit at the bank.

·     ‘Optimise your funnel.’ Before optimising it, confirm that an actual human wants to enter it.

What AI is genuinely good for

I’m not breaking up with AI. I’m just no longer letting it hold the company card unsupervised. Used properly, it is brilliant at speeding up work that already has a human point of view.

·     Turning one real idea into versions for a blog, carousel, pin and email—after I decide what I actually mean.

·     Finding gaps in a draft, challenging assumptions and suggesting questions I forgot to ask.

·     Organising messy notes, comparing options and helping me see patterns in real performance data.

·     Giving me a first draft to argue with. Apparently spite is also a workflow.

My new rule: evidence first, robot second

Now, before asking AI for a plan, I try to bring it something real: customer comments, search terms, actual clicks, failed posts, conversion numbers, competitor reviews or a clear description of the person I’m trying to help. The better the evidence, the less likely I am to receive a beautifully formatted horoscope.

I also ask it to disagree with me. ‘Tell me why this will fail’ is not as emotionally soothing as ‘write a launch plan,’ but it is significantly cheaper. I ask for assumptions, risks and the smallest test I can run before investing more time or money.

A PROMPT THAT IS ACTUALLY USEFUL Here is my idea, the evidence I have, what I’m assuming and what I can afford to test. Give me the three biggest reasons it may fail and the cheapest way to test each one.

 

The human part is still annoyingly compulsory

AI can draft. It can analyse. It can confidently suggest ‘authentic storytelling’ in a tone last heard during a quarterly earnings call. What it cannot do is care whether the story is true, notice when the joke sounds like a LinkedIn intern wrote it, or feel the tiny internal alarm that says, ‘Nobody I know would speak like this.’

That part is mine. So are the final decisions, the customer conversations and the consequences. AI is a useful assistant. It is not a co-founder, a customer or a small oracle living inside a blazer.

I asked AI for a business plan. It gave me leverage and synergy. I brought the common sense later.