Automation for the work nobody should be doing
Somebody in your business spends Thursday moving information from one screen to another. That is the work worth automating first.
The process works. It works because three people hold it together with spreadsheets, copy and paste, and a reminder in somebody's calendar.
What changes
Measured in hours, not in features.
Hours back, counted
We time the task before and after. If the saving is not worth the build, that is a finding and we say it rather than building anyway.
Fewer things typed twice
Every re-entry is a chance to introduce an error that somebody finds three weeks later, usually in front of a customer.
A process that survives a holiday
Written down and running whether or not the one person who understands it is at their desk.
How it runs
Four steps, and one of them is doing nothing.
The third step is the one that saves clients the most money, and it is the one an automation supplier is least likely to offer.
Watch the work happen
We sit with whoever does it and time it. Not a workshop about the process, the actual process, including the parts people are slightly embarrassed by.
Count what it costs
Hours a month, error rate, and what a mistake costs when it reaches a customer. Now there is a number to beat.
Delete before automating
A surprising share of these tasks exist because of a report nobody reads or an approval nobody uses. Removing a step is cheaper and more reliable than automating it.
Automate what is left
The smallest thing that works, in front of the people who do the job, with a way to intervene when it gets something wrong.
What is included
Where this usually lands.
- Documents in, structured data out: invoices, orders, forms, applications
- Systems that should talk to each other and do not
- Reports assembled by hand every month by somebody senior
- Triage and routing of incoming enquiries
- Checks that only happen when a person remembers to do them
- A record of every automated action, so an error can be found and explained
Questions people ask
The ones that actually arrive.
Is this about replacing staff?
Not in the work we take. The clients who get value from this are short of people and drowning in admin. If the brief is to cut a team, we are the wrong studio and it will not go well with us.
What if it makes a mistake?
It will, so we build for it: a log of every action, a threshold below which it asks a person, and a way to undo. Automation without a visible audit trail is not something we will ship.
Does it need AI?
Often not, and we will say so. Plenty of this is ordinary integration work, which is cheaper, faster and far more reliable. AI earns its place where the input is genuinely messy, like documents and free text.
How do we know it worked?
The same way we decided to build it. Hours timed before, hours timed after, error rate compared. If the number does not move, we have not finished.
Your turn
Bring one question you cannot currently answer.
30 minutes. We will tell you whether it is measurable, roughly what it would take, and whether we are the right people for it.