What does an AI workflow look like inside a real business?
One enquiry, followed end to end: how the trigger, the data, the AI and the human review fit together in a process you already run.
Nathaniel Hart
28 January 2026
5 min read
“AI workflow” sounds technical. It is not. It simply means AI has been connected into a process your business already runs, so that the repetitive middle of that process happens without anyone doing it by hand.
The easiest way to understand it is to follow one job from start to finish.
The five parts
Almost every workflow we build has the same five parts, in the same order.
- Trigger — something happens. An enquiry arrives, a job is marked complete, an invoice is approved.
- Data — the workflow gathers what it needs: the customer record, the last three jobs, current pricing, stock on hand.
- AI — the model does the reading, sorting, drafting or summarising that a person would otherwise do.
- Human decision — a person reviews the output and approves, edits or rejects it.
- Action — the approved result is written back into your systems and sent on.
That fourth step is the one people miss, and it is the one that makes the whole thing safe to use.
A worked example
Take an enquiry landing in a trades business at 7pm on a Friday.
The trigger is the form submission. The workflow immediately pulls the data it needs: is this an existing customer, what work have we done for them, what is the site address, are we already booked in that area next week.
The AI reads the enquiry in plain English, works out what is actually being asked for, classifies it as urgent or standard, and drafts a reply with an indicative next step and two possible appointment windows.
On Monday morning, the operations manager sees one screen with the enquiry, the customer history and the drafted reply. They change a date, approve it, and move on. Total time: about twenty seconds.
The action step then sends the reply, creates the record in the CRM, and puts a provisional booking in the calendar.
What actually changed
Nobody in that business lost their job. What changed is where their attention goes.
Roughly ten percent human at the front, eighty percent AI through the middle, ten percent human at the end.
The manager no longer spends Monday morning reading, triaging and typing. They spend it deciding. The customer gets a considered response first thing instead of late Tuesday. The record is complete because it was written by the workflow, not remembered by a person.
Where these usually start
Good first candidates share a shape: they happen often, they follow a process people broadly agree on, and the information needed already exists somewhere in your systems.
- Enquiry triage and first response
- Quote and proposal drafting from standard components
- Job and site report summarisation
- Keeping CRM records complete without manual entry
None of that is futuristic. It is ordinary work, done faster, with the judgement left where it belongs.
Keep reading

