AI in practice
A tool, used where it pays.
AI doesn’t set the strategy; it does the slow parts. Work that took days takes minutes, and the time goes back into thinking. The systems below are in use today, not demos: a person signs off on each one, and each one answers to a number.
Studio
AI Content Studio: daily video, no filming
The done-for-you offer I took to market at Hook Point, for experts who should post every day and can’t spend the week filming. Our team researches, writes and edits the videos; an AI version of the client, cloned from a ten-second video and a minute of audio, delivers them. The client approves and posts. Anyone can buy an avatar; the work is deciding what it says.
- input
- 10 s of video, 1 min of audio
- our team
- research, script, edit
- client
- approves and posts
Ads
AI-presenter ads at real volume
The avatar is a constant across every ad: same face, same budget, same polish. The only variable is the script. Our strongest creative stops about one in four people who scroll past it; the weakest, closer to one in five.
- best
- stops 1 in 4
- weakest
- stops 1 in 5
- variable
- the first sentence only
Creative
Creative volume without a bigger team
One product, one angle, a hundred variations in minutes. In one week I launched 300 ad variations on my own. The tool did the making; my time went into the angles and into deciding what to cut.
- launched
- 300 in one week
- team
- one person
- my job
- the angle and the cut
Search
AI search visibility (GEO)
When buyers ask ChatGPT, Claude, Perplexity or Gemini for the best option, I measure whether a brand is named, where, and which sources the answer cites. Questions that already name the brand don’t count: repeating a name it was given isn’t a recommendation. Then I fix what the models can read.
- models
- ChatGPT, Claude, Perplexity, Gemini
- counts
- questions that don’t name the brand
- goal
- be the answer, not the link
Data
Data you can talk to
CRM, ad accounts and sales calls in one warehouse, asked in plain language. A weekly AI audit flags what moved and why before the Monday meeting, so a three-day report becomes a thirty-second question.
- sources
- CRM, ads, calls
- cadence
- weekly audit
- interface
- plain language
Agents
Agents that answer to a person
Each agent has one job and one owner. One grades every recorded sales call against the sales playbook, and gives no grade without a quote from the call. One works out who the reader is and how much they already know before it drafts a line of copy, and flags any claim with no proof behind it. Strategy, taste and the final call stay with a person.
- calls
- no grade without a quote
- copy
- no claim without proof
- people do
- judgement, sign-off