I built the system our family firm runs on.

Hesketh & Son writes for firms that sell to councils, funders and awarding bodies, or want to. They are small and mid-sized, in the UK and Norway: care providers, cleaning firms, builders, specialist trades. The bid goes in under the firm’s own name, as its own work.

I am a copywriter by trade, and persuasion is the thread through everything I have done. So when I built a system that writes, the first person it wrote for was my father. For years, his own professional letters were written that way, under his own name. He was our first client.

Then I turned it to bids, because bids are scored. Councils publish how they scored: the real marks, the screen-outs and the reasons. I thought small firms lost marks by writing from the gut. Reading real firms’ records showed me the value is in the read, before any writing: what does the buyer score, and can the firm’s evidence carry it? So every job now starts with the read.

The craft is the family’s. The speed is new: AI does the structure and the first draft, and I check every line before it reaches anyone.

First it reads. For each firm, it reads every public record that carries the firm’s name: its certificates, its past work, who has been buying what it sells, who won last time, and what is open now. By hand, one firm’s record is most of a day’s work. The system does it in hours, and it keeps reading past the places a person would think to look first.

Before the machine writes a word about a firm, the system decides what it must know and what it must not guess. When there is a bid to write, I read the scoring scheme first, and the draft is written to that and nothing else. Every factual sentence in the draft is labelled with where it came from, the firm’s own papers or the notice itself, so anyone can check a claim without taking our word for it.

My name is on every job. Nothing reaches anyone until I have read every line. I check each draft against two things: what the buyer scores, and what the firm can actually deliver. Where the evidence is weak on a question, I say so and aim the answer at something the evidence can carry. I do not fill the gap with sentences a scorer cannot mark. The reading is wide and fast. The judgement stays with a person.

My father, Philip, is the firm’s chairman. He teaches the psychology of influence and persuasion to audiences around the world. He is a published author, and two of his books were Amazon No.1 bestsellers. I run Hesketh & Son, and I put my name on every job.

I always wanted to work with him. The easy choice is a made-up brand. We put the family name on the door instead. The rule since then: nothing goes out that we cannot trace, even when it would sell.

The values are fairness, truth and my father’s philosophy, as his books lay it out. One of his rules says most of it: do what you say you are going to do.

The firm keeps that rule in public. It publishes what it refuses to do. It will not invent a win rate. It will not write two competing bids for the same opportunity. It will not take a percentage of what a client wins. It will not pad a bid the evidence cannot carry; it turns the work down instead. And it does not guarantee an outcome, because no one honestly can.

A machine is only as good as what it has been given and what it has been told not to do. The system I built is the part that decides what the machine must know, and what it must never guess.

In UK central government buying, using AI does not move the responsibility for accuracy off the firm that bids. So every fact carries its source, and somebody has to own the end of it. In this firm, that is me.

• Inside a company

  I built one shared memory for a whole company, and it became the centre of its AI.

• Forge

  The company of AIs I run my businesses with. I built it with the methods I teach.

• Work with me

  I build with your people, I coach them, and I speak.