Going AI-native

How to start going AI-native: a back-office checklist

General information, not legal advice. Last checked 31 Oct 2026. Examples use a fictional business.

You don't go AI-native in one go. You start with one routine job that eats your team's week, write down how you do it, and hand it over with clear rules on what needs a person. Then you do the next one.

The checklist

#StepDone when
1List your routine back-office jobs: logging, chasing, updates, portal entry, remindersYou have a list your team agrees with
2Pick one: frequent, follows a pattern, low risk, and annoyingOne job chosen, with one owner
3Write down how you do it (SOP guide) along with the templates it usesThe person who does it has checked it
4Write down what always goes to a personA short, explicit list
5Set the approval rule. Start with everything approved.Everyone knows who approves what
6Check your data rules: personal data only in tools set up for business useAgreed and written down
7Run it on real cases alongside the team, and fix the procedure where it goes wrongMistakes go back into the written procedure
8Review after an agreed period: keep, fix or stopDecision made; next job picked

Step 6 is general information, not legal or qualified advice. The ICO publishes guidance for UK organisations.

Signs of a good first job

Signs of a bad first job

Example: a letting agency's first job

For a letting agency, a natural first job is the repair report. The report gets logged, the landlord and contractor are chased, and the tenant is told what happens next, all following the agency's own procedure. Anything that needs a person comes back to the team. Letting agencies in Scotland are the first trade Lorework is built for. See Lorework for letting agents in Scotland.

Scottish agencies also have deadlines a routine must hit. See the Repairing Standard checklist and Awaab's Law in Scotland. General information, not legal or qualified advice.

Starting with Lorework

  1. Bring your procedures. Lorework learns your procedures and templates and shows its sources.
  2. Give it a task or a schedule. It takes on scheduled and recurring work and keeps a record of what was done.
  3. Your people approve. You can require approval before it writes or submits anything, and it works in the web systems you already use.

Questions

Where should a small business start with AI?

With one routine back-office job, written down and handed over with clear approval rules. Not with a company-wide strategy.

Do we need an AI policy first?

A short written rule on data and approval is enough to start, and you can build on it. General information, not legal or qualified advice.

How do we know if it's working?

Compare how long the job takes and how often it goes wrong before and after, using your own records.

What if it doesn't work?

Stop, keep what you learned (your written procedure is still useful), and pick a different job.