Skip to content

AI & automation

The repetitive half of the day, handed to software

Not a chatbot bolted onto a website. AI built into the system your business already runs on, doing named jobs with a person reviewing anything that leaves the building — and an honest list of what we will tell you not to automate.

What it actually does

Six jobs, named

Each of these is a task somebody on your team does today, by hand, at a predictable hour. That is the test for whether it is worth automating at all.

Reading

Documents into data

Invoices, purchase orders, forms, ID documents, statements and handwritten slips turned into records in your system. The saving is not the typing; it is that nothing waits three days in somebody’s pile first.

Sorting

Everything that arrives, triaged

Enquiries, complaints, applications and support messages classified, given an urgency and routed to the right person the moment they land, instead of at whatever hour someone opens the inbox.

Writing

First drafts, not final ones

Quotations, proposals, replies, descriptions, listings and reports drafted from your own records and your own past work, then edited and approved by a person. The draft is the work being removed, not the judgement.

Scoring

Which lead to call first

Ranked on what your own closed deals actually had in common, rather than on a consultant’s rule of thumb. It improves as the history grows, and it is explainable — a score nobody can interrogate is a score the sales team ignores.

Chasing

Follow-up that happens without being remembered

Sequenced messages on WhatsApp, email and push, that stop the moment the customer replies. Most lost deals are not lost arguments; they are a second call nobody made.

Reporting

Numbers that arrive

The weekly figures assembled and sent on Monday morning, with the movement explained in a sentence. A dashboard is something you have to remember to open; this is something you have to remember to stop.

What we will talk you out of

Four things not to automate

This list is on the sales page on purpose. Most failed AI projects were sold by somebody who did not have one.

Anything final that nobody checks

A model that is right 95% of the time is wrong on one letter in twenty, and the wrong one is not flagged. Anything that reaches a customer, a regulator or a bank has a person between the draft and the send.

A decision you cannot explain

If a rejection, a price or a refusal cannot be accounted for to the person affected, it should not be automated. That is a legal position in more places every year and, separately, it is how trust in the system is lost internally.

A process nobody has written down

Automating a process that only lives in one person’s head reproduces their bad days along with their good ones. The process gets documented first — which on its own is often the improvement being paid for.

A chatbot in front of a broken operation

If customers are writing in because something is unclear or unreliable, an assistant answering faster makes the problem cheaper to ignore. Fix the cause, then automate the remainder.

How we start

One task, measured, in a fortnight

Not a transformation programme. We pick the single most repetitive task you have, automate that one, and measure the time it actually saved against the time it used to take.
01

You name the task that eats the most hours.

02

We time it as it is done today — the baseline nobody usually records.

03

It is built, with the review step in from the first version.

04

It runs alongside the manual process for a week and the two are compared.

05

If it did not save what we said it would, we say so and stop.

What does your team do every single day that a machine could draft?

That question is usually enough to start. Answer it in a sentence and we will tell you whether it is worth automating — including when the answer is no.