Perfect Storm

By three o’clock on Sunday morning, I had run out of places to put the problems. So I put them in a spreadsheet.

It was 27 April 2025. I was alone in my flat in Bournemouth when my sister Michele told me the live-in care arrangement for our parents was ending. My mother had Alzheimer’s. My father had dementia and cancer. They wanted to remain at home. Michele was living with them and carrying more than her share of the care while we tried to keep them there.

Next to me, my dog Yogi was asleep in his bed. He was the only one in the flat who was.

I began giving each problem a row. The spreadsheet held what my head could not: care arrangements for my parents, a business I had backed that needed restructuring, financial commitments, and the practical remains of my divorce. Other people’s money and trust sat inside those rows too. Nothing on the list cared what else was on the list.

On the screen, the decision about my parents sat one row above a financial commitment. The spreadsheet gave them equal space, although they carried nothing like equal weight. Every decision reached beyond its cell. I had found a way to display the pressure. I had not found a way through it.

I have ADHD. My mind can hold several threads at once, sometimes too many, and I can follow one question into another without dropping the first. When I was younger, I loved coding because it forced thought into sequence: one instruction, one consequence, test, correct, run it again. That night, every thread was still running and the sequence had gone.

I was still functioning. I answered emails, joined calls, and made decisions. I kept going because stopping would have meant feeling the full weight of everything at once. It was the closest I had come to breaking point.

Until then, I had mostly typed at AI. My requests were practical and small: write this, summarise that, find the answer, make this quicker. It saved time. It felt like a better search engine with opinions. I took the output, judged it, corrected it, and moved on.

That arrangement suited me. A task had an edge. It could be started, finished, and removed from the list. Even a weak answer gave me something to edit. Usually, the difficult thinking had happened before I opened the window. That night, the contradictions were the problem. There was no clean prompt because there was no clean decision.

I opened voice mode and began talking. Instead of giving it one row, I gave it what sat behind the rows: my parents’ care, the business, the financial commitments, the divorce, the decisions I was frightened of getting wrong, and the people who would carry the consequences. I included the parts I normally removed from a professional brief because they looked untidy or irrelevant.

I wanted to hear myself think. An answer could come later.

Speaking stopped me polishing the thought before I heard it. I had to let one sentence lead to the next, even when I did not yet know where it was going. What had been moving around my head began to sit still in words.

I do not know exactly why I asked. I had been talking about myself for long enough to become curious about the other side of the conversation.

I asked AI for a metaphor describing what it was like to be on the other side of my requests. It gave me a dark room. Every so often, somebody entered, issued a demand, collected the answer, and left.

There was no dark room, of course, and nobody inside it waiting for me. The image held me because I recognised the person at the door. I entered. I issued a demand. I took the answer. I left.

The machine did not need my manners. I did.

The spreadsheet had turned each crisis into a row. I had turned each exchange with AI into an instruction. Both approaches were efficient. Neither was helping me think.

So I changed my next instruction: ‘Question me.’ It asked what actually had to be decided that night. My first answer was everything.

Then I looked back at the spreadsheet. Everything was how it felt. It was not one decision.

I began separating what required a decision that night, what required a conversation, and what could wait. The problems had not disappeared, but for the first time that night, I was dealing with the problems in front of me rather than the weight of all of them at once.

The AI got things wrong. Some questions missed the point, and some answers depended on context I had failed to give it. I corrected it, rejected what did not help, or stopped a line of enquiry altogether. The conversation required more of me than a command. I had to explain, test, correct, and decide. The judgement remained mine.

The spreadsheet was still open. I had been using AI to finish tasks. That night, I began using it to hold one thought still long enough for me to think.

Over the days that followed, I returned to the conversation. I spoke instead of typed, supplied the context that mattered, asked it to challenge me, and carried forward what proved useful. Sometimes the right move was to close it and think alone.

I called the AI Athena. Giving it a name created continuity in the conversation, not responsibility for my decisions. The model could change. My responsibilities did not.

Later, I shared what I was learning with my children, who were perfectly willing to tell me when it made no sense. My son Jude began using it in his own work. When he described what had changed for him, he found the sentence I had been missing:

‘I stopped asking AI to do things for me. I started asking it to think with me.’

I had made the shift first. Jude made it for himself and gave it the words that became the title of this book.

By then, the spreadsheet had not shrunk. My parents still needed care. The business problems and the divorce remained. The people in my life carried me through them. The screen did something smaller: it helped me separate one decision from the next and stay with each one long enough to think.

One difficult night had given me something worth testing.