AI for trade businesses: practical ways to reduce admin

The last job is finished, but there are still messages to answer, a quote to tidy up and tomorrow's visits to confirm. That is a useful place to start thinking about AI: one piece of admin that keeps following you home.

For a trade business, AI can help turn messages and rough notes into a first draft. Your existing software can handle the predictable steps around it. You keep control of prices, commitments and the work itself.

Start with the admin you already repeat

Pick a task you do often enough to notice the effort. Write down what arrives, what you do with it and what needs to happen next. For example: an enquiry arrives by email, you check the area and job type, ask for missing information and arrange a visit.

Before adding a tool, check what your current booking, accounting or job-management software can already do. A saved reply, a required form field or a reminder might solve the problem.

• Use ordinary automation when the rule is clear: copy a submitted form into a job list, send an agreed reminder or flag an overdue task.
• Consider AI assistance when the input varies: summarise a long enquiry, turn rough notes into a readable description or draft a reply in your usual tone.
• Keep a person responsible for decisions that affect the customer: job scope, availability, prices, technical advice and promises about completion.

A practical example: from enquiry to a checked reply

This is an illustrative workflow for a plumbing business, not a customer case study or a claim about a system we have delivered.

A message says: “Can you replace the kitchen tap? We have bought one already and are usually home on Fridays.” It does not say where the property is, what tap has been bought or whether Friday means this week.

1. Capture the enquiry. Save the message in the existing job system. A form or email integration can do this without AI.
2. Prepare a draft. Ask AI to summarise the request and list missing details. It can draft a short reply asking for the postcode, tap model and suitable dates.
3. Check it. The person handling enquiries compares the draft with the original message, removes assumptions and decides what to ask next.
4. Send and record. Send only the approved reply, save it against the enquiry and create a follow-up task.

The draft should say “date to confirm”, not invent a free slot. If it cannot work out what the customer means, it should flag that for you rather than quietly guessing.

Four other places to try it

Quote descriptions

Use checked site notes to draft the description of work and the exclusions. Keep rates, quantities, VAT treatment and totals in your quoting or accounting system. Confirm the scope and calculations before sending anything. A fluent paragraph is not evidence that the price or specification is correct.

Job notes after a visit

Turn your own rough notes into a structured record: work completed, materials used and follow-up needed. Compare the result with the notes before saving it. Do not let a summary invent a measurement, a completed check or a customer's agreement.

Customer updates

Draft a clear update from confirmed facts, such as an agreed new appointment time. A fixed reminder may only need a template. AI is more useful when the explanation changes, but somebody still checks that the wording matches what has actually been agreed.

Invoice follow-up

Use accounting rules to identify an overdue invoice. AI might help draft a polite message, but check that the payment has not arrived and that there is no unresolved query. Start with draft-only messages so a disputed invoice does not trigger an inappropriate chase.

Put checks around the draft

The National Cyber Security Centre's AI guidance explains that generative AI can present incorrect statements as facts. A confident answer still needs checking against the job record.

For a first trial, give the tool a narrow job and a clear boundary:

• Let it prepare drafts without permission to send messages, change prices or book visits.
• Keep the original enquiry beside the proposed reply so the checker can compare them.
• Mark missing information explicitly. Do not fill gaps with plausible details.
• Keep technical judgement, safety assessments and certification with the qualified person doing the work.
• Keep a manual route available if the tool fails or produces an unclear answer.

Start with invented sample enquiries. Before using real customer details, check the provider's handling of data, retention, access and model training. The ICO's guidance on generative AI and personal information covers lawful use, explaining that use to people and checking the accuracy of outputs. A paid subscription alone does not settle those questions.

Set a budget around one problem

Ask for separate figures for setup, subscriptions, usage charges, integrations and ongoing support. Also allow for your team's time reviewing drafts and correcting errors. Set a usage limit and agree what happens if a connection stops working.

At Appening Now, a shared workspace setup starts at £1,500 + VAT, and bespoke business tools start at £7,500 + VAT. These are different scopes: the workspace setup is not a complete AI assistant, and the business-tool starting price is not a quote for every workflow above. Platform fees and any AI usage need to be scoped separately. See current pricing and what each option includes.

If your existing software can do the job with a template or a setting, start there. A bespoke build becomes worth considering when a valuable workflow cannot be handled well by the tools you already have, and the benefit justifies the build and running costs.

Run a small trial before expanding

Choose one workflow, one person to review it and a fixed trial period. For example, measure a week of enquiry handling as it works today, then try draft assistance for two weeks. Treat that as a practical starting point, not a promise that two weeks is enough evidence for every business.

Record these measures for each enquiry:

• Time spent preparing and checking the reply.
• Corrections needed, including missing or invented information.
• Time until a useful response reaches the customer.
• Whether the next step happened without extra chasing.

Compare similar jobs and include checking time. A low enquiry count or an unusually quiet week can make the result misleading. If checking the drafts takes as long as writing them, change the approach or stop. If the result is useful and reliable, extend it to the next related step.

Choose your first workflow

Bring one recurring admin task, the tools you use now and a few invented or properly anonymised examples. That is enough to start a conversation about whether better setup, ordinary automation or AI assistance makes sense.

Request a free AI audit to explore where to start. If you already have a workflow in mind, plan your project. You can also read how we approach AI engineering before deciding.