Automation
Automation for the jobs you keep doing by hand.
Useful automation starts with a repeatable problem, not a tool looking for something to do.
When the same manual job keeps coming back
If a job keeps taking hours because somebody has to repeat the same checks, edits or imports, I can help turn the repeatable part into a faster, safer process.
Where I focus
I enjoy spotting a process that could work better. Sometimes the answer is a reliable spreadsheet. Sometimes it is a custom tool, an import or a small amount of code. AI can help with parts of the work, but it doesn't remove the need to understand the process.
Find the repeatable part
Which steps are the same each time? Which need judgement? Where do errors enter? Those questions matter before deciding what to automate.
A process with unclear inputs can become a faster way of producing the wrong answer. I'd rather define the matching rules and exceptions first.
Tools I've built or worked on
Examples include supplier stock checking, custom reporting calculations and the product image generator for my own business.
The scope differs. A specific stock-checking process is not a claim that I've connected every supplier to a fully automatic system.
Where AI is useful
I use AI to help with drafting, code and structuring information. For business data, the output needs checks against the source and a clear decision about what can be changed automatically.
My approach to AI-assisted workflows keeps that distinction between assistance and unattended decisions.
Which job keeps eating up your time?
Send me the site or a safe sample of the data and explain what is taking too long or not working. I will focus on the smallest useful job first.
Give exceptions somewhere to go
A missing identifier, a changed supplier page or an empty result should not quietly become a live update. An automation brief should say what happens when the normal path does not work.
That may mean a review list, keeping the existing value or stopping that record until it has been checked.
Agree the ongoing responsibility
Some tools are a one-off delivery. Others depend on sources that keep changing. Refresh schedules, maintenance and support should be part of the scope where they are needed.
Read import versus sync before treating a repeatable upload as an always-on service.
Questions
Can a one-off job become a repeatable process?
Often that is worth investigating. The match rules, exceptions, input changes and maintenance still need to be defined; a successful first run is not enough.
Will AI make the decisions automatically?
Not by default. I'd define where it can assist, what must remain source-backed and which decisions require a person before anything reaches a live system.
Explore this area
- A supplier stock-checking process
- Building a product image generator for MostlyPens
- Precision jigs for personalised products
- Making personalised text fit reliably
- Unknown stock is not the same as out of stock
- Where AI helps with product data, and where I would stop it
- What to send when you need a website or data problem fixed
- A one-off import is not a live sync
HAVE SOMETHING AWKWARD?
Which job keeps eating up your time?
Send me the URL, what is happening, what you have already tried and what a useful result would look like. A small anonymised sample is enough to start - do not send passwords, credentials or sensitive customer data.