Service 01 · AI Automation

AI automation that removes the job, not the click.

Old automation replayed your keystrokes and fell over the moment reality changed. AI automation reads the situation, makes the judgement call and does the work — including the awkward twenty per cent that used to land back on a human.

What It Actually Means

The difference is judgement.

Traditional automation is a set of rules. If the invoice always arrives in the same format, from the same supplier, with the same fields in the same places, a rule handles it beautifully. The trouble is that a business is not a set of rules — it is a set of exceptions with a rule-shaped average.

So the familiar thing happens. The tool handles the easy eighty per cent, everything unusual gets kicked to a person, and the person ends up doing the hard part plus babysitting the tool. Nobody got their week back.

AI automation closes that gap. The system reads the messy input the way a competent colleague would, works out what it is looking at, decides what should happen, and only escalates when it genuinely should. The rules still exist — but they sit around the judgement as guardrails, not in place of it.

The Usual Suspects

What businesses actually ask us to automate.

None of this is exotic. It is the ordinary work that quietly consumes a week and never appears on anyone’s job description.

Intake & triage

Enquiries, forms, emails and documents arriving in twelve shapes — read, classified, routed and answered without a person sorting the pile first.

Quoting & proposals

Turning a description of a job into a priced, formatted, sendable quote — consistently, and in your voice rather than a template’s.

Document handling

Reading contracts, invoices, statements and reports; pulling out what matters; filing the rest where it belongs.

Follow-up & chase

The revenue that leaks because nobody had time to chase it. Chased politely, on time, every time, and it stops when the customer replies.

Reporting

The Monday report that takes someone until Wednesday. Assembled, checked and explained in plain English.

System-to-system data

The retyping between two tools that were never introduced to each other, done properly instead of by copy and paste.

How One Gets Built

Four steps. No mystery.

1
Free · A few minutes

Map the real process

Not the documented one — the real one, including the spreadsheet somebody keeps privately because the official system cannot cope. That spreadsheet is usually where the actual business logic lives.

2
Priced, fixed, written

Decide what should be automated

Some of it should not be. Automating a broken process just produces the wrong answer faster, so anything that needs fixing first gets said out loud rather than quietly built around.

3
Working software early

Build it where the work happens

Inside the systems you already use, not a new portal nobody opens. If the team has to remember to visit it, it has already failed.

4
Yours at the end

Watch it, then hand it over

It runs under supervision while the edge cases surface, because they always do. Then it is yours — host it yourself or have us run it.

The Honest Part

Automation without guardrails is just faster mistakes.

A human signs anything that leaves the building. The machine drafts, a person is accountable. Speed never buys its way past that.

Every decision is logged. If it did something odd on a Tuesday, you can see what it saw and why it chose that.

It knows when to stop. Low confidence escalates to a person instead of guessing confidently, which is the failure mode that costs real money.

It fails safe. When a model or an API goes down, the process degrades to something a human can pick up — it does not silently drop work.

Questions

The ones we actually get asked.

How is AI automation different from Zapier or Power Automate?
Those tools connect things and follow rules, and for predictable steps they are genuinely good — we use them where they fit. The difference is judgement. A rule cannot read an unusual email and work out what the customer actually means. AI automation handles the variable, messy middle that would otherwise be escalated to a person, and it can sit on top of the tools you already have rather than replacing them.
What if the process changes after it is built?
Processes always change, which is why the system is built to be edited rather than sealed. Changes are handled as straightforward change requests, and if you host it yourself your own team can make them.
Will it replace my staff?
In practice it removes the part of the job nobody wanted — the retyping, the chasing, the sorting — and gives the hours back to the part that needs a person. If your honest goal is headcount reduction, say so at the scan and we will tell you plainly whether the numbers support it.
Is my data safe, and where does it go?
Data handling is designed before anything is built, and where a process involves personal or regulated data it is scoped to that standard from the start. You can host the whole thing on your own infrastructure if you would rather nothing left your walls.
How do we know it is working?
Because you can see it. Every run is logged and reportable, so the value is measurable rather than asserted, and the failure cases are visible rather than quietly absorbed.
The Rest of the Lab

These combine.

Most builds use two or three of them together. See all services.

Step One

Start with one honest sentence.

“Here’s what’s quietly eating my week.” Tell Ray in your own words. A clear, priced build path comes back, and a human signs off everything that leaves the lab.

Start the scan — free
Free scan · Fixed prices · You own everything