Assistants that know your business
Trained on your own documents, prices, policies and past answers, so staff (or customers) get the right answer, with the source.
AI for business
Sound
Sound
familiar?
- Someone bought a ChatGPT licence for everyone and nothing changed.
- Your team answers the same twenty questions every week.
- The knowledge is in two people's heads and one of them is retiring.
- Documents arrive, someone reads them, someone types the important bits into another system.
- You've been sold an "AI strategy" deck and still don't know what to do on Monday.
What I
actually do.
Specific jobs, done properly, with a person checking the output.
Inbox and document handling
Read the email or the PDF, pull out what matters, put it where it belongs, and flag anything odd to a person.
First drafts
Quotes, replies, reports and proposals drafted from your own templates and data. A person edits and sends.
Decision support
Ask a question of your own numbers in plain English and get an answer you can check.
Agents, carefully
Multi-step jobs a machine can run end to end, with limits on what it may touch and a human at the points that matter.
What stays human
Judgement, relationships, taste and anything you would be embarrassed to get wrong. AI drafts, sorts and summarises. People decide.
How it
starts.
1
You tell me what's broken
In your words. The thing that wastes time, loses money or drives you up the wall.
2
I find where the time goes
I sit with the people doing the work and map it as it really happens, not as the org chart says.
3
We split human from machine
Judgement, relationships and taste stay human. Copying, chasing and compiling don't.
4
Build it, then prove it
Built with a human checking the output, measured against the before, and handed over so your team can run it.
Related work.
All work →> visitor: dog-friendly lunch near the
waterfront, tomorrow?
< 3 places · open · 6 min walk
A tourism website that answers questions
Turned a tourism website into an intelligent visitor platform: ask it for a dog-friendly lunch near the waterfront tomorrow and it answers from what is actually open.
company.os · 214 processes mapped
An operating system for a whole company
Built an operating system designed to understand an entire company: its processes, who does what, and where the time goes.
Questions
people ask.
Is our data safe if we use AI?
It can be. Use business terms where your data isn't used to train models, keep personal data out where it isn't needed, and write down what goes where. If a job can't be done safely, I'll say so.
Do we need a big AI project?
Almost never. The useful work is usually one or two specific jobs that eat hours every week. Start there, prove it, then decide whether to do more.
Will it replace staff?
That isn't the aim. The aim is to stop good people spending their week copying, chasing and compiling, so they can do the part of the job that needs them.
Which AI do you use?
Whichever fits the job. I build a lot with Anthropic's Claude, and I'm not tied to any vendor.
Other services
Next step
Find out where
Find out where
your week goes.
Three minutes, no email needed to see the answer. The audit adds up the hours your team spends on jobs a machine could do, and shows what that costs.