Saved replies and templates, and stop there

There is a clean line through this subject and it holds in every case. Automate the parts a buyer never sees, and leave alone every part they are paying for.

The parts a buyer never sees are the majority of your hours, so this is not a small permission. It is just not the permission people usually take.

Worth automating

The first reply. Acknowledgement, one clarifying question, price, turnaround. Thirty seconds instead of ten, on every enquiry, forever. It is the highest-leverage saved text in this trade, and the reason is in the post on replying fast without living online.

The scope confirmation. Two sentences, identical every time, sent before you start. This is the sentence that prevents most disputes, which is why it has its own post.

The delivery message. Same structure each time, with the specifics dropped in.

The record line. Date, buyer reference, fee, take-home, minutes. One row, automatically or in fifteen seconds, because you will want it a year later and cannot reconstruct it. In the UK it is becoming a software requirement: GOV.UK says sole traders with qualifying income over £50,000 should have started Making Tax Digital for Income Tax from 6 April 2026, with the threshold falling to £30,000 in 2027 and £20,000 in 2028.

Your own document skeleton. Headings, axes, house style, the boilerplate at the top and bottom. A skeleton is not a script, and the difference is worked through in the post on templates that speed you up without flattening the work.

Not worth automating

Anything that produces the assessment itself.

Generating the substance of a review with a model and lightly editing it is not efficiency; it is reselling a free product at a human price, and it is the single fastest way to lose the reason anyone commissioned you. An automated assessment costs a buyer nothing and arrives in seconds - the account of how automated scoring works is worth reading precisely so you know what you are not selling, and the consumer-facing tools that package it are what most buyers tried before they came looking for a person.

The commercial point is blunt. If a model could have produced your delivery, it will eventually be priced as though a model produced it, and your rate card cannot survive that.

Also leave alone: the clarifying question, which has to respond to the actual brief; anything sent after a complaint; and any part of a reply where a buyer would notice they had received a form letter, because that is what buyers report as the thing that made the whole purchase feel worthless.

The middle cases

Scheduling and reminders are fine and mildly useful.

Bulk messaging repeat buyers is usually against platform terms and frequently reads as spam even where it is permitted; the constraints are covered in the post on building a list of repeat buyers.

Auto-accepting orders is a bad trade. The screening question in front of an order is worth more than the minutes it costs, because the jobs you decline are the ones that would have produced the bad reviews.

Analytics tooling is almost always over-specified for a single earner. One sheet with a row per month does what a subscription does, and it is the whole of the monthly review.

The test

Before automating anything, ask whether a buyer who found out would feel cheated.

Nobody feels cheated by a fast standard reply or a consistent report layout. Everybody feels cheated when the judgement they paid for turns out to have been generated, and that reaction is not unreasonable, since judgement by a named person who actually looked is the only thing in this trade that is not already free.

The demand side's own version of that expectation - what a buyer believes they are commissioning - is set out at Rate Penis's judges hub.

Platform rules on this are not uniform and are worth reading rather than assuming; Rate Cock states what judges may and may not delegate on its judges page.