Guide

AI automation examples for UK small businesses

Harrowtech.ai · last checked 13 October 2026

A small London bakery was funded in January 2025 to test AI that forecasts how much of each product to bake each day, to cut waste. A gardening business in England turns a recorded voice note into an invoice in its accounts software. A Greater Manchester coaching business saved about 180 hours in a year with AI tools. And a commercial landlord saves 6 hours a week with a dashboard Harrowtech.ai built. These and other AI automation examples are below, each with what it does, what it needs and what the source actually says.

Where a figure comes from the company that sells the tool, the page says so. For the plain definition, see What is AI automation?

1. Forecasting what to bake each day

What it does. Miel Bakery, a small bakery in London, is working with Oxford Cognitive Labs on AI that predicts sales and forecasts how much of each product needs to be made daily, to cut food waste.

What it needs. Historic sales and other data. UKRI says the project will assess “the technical feasibility of using historic sales and other data sources to predict future sales”, and explore the business case for wider use.

What the source says. It is one of 120 projects sharing £7 million in UK Government funding, announced on 14 January 2025 and delivered by Innovate UK’s BridgeAI programme. It is a funded trial, and neither source reports a result.

2. A spoken note that becomes an invoice

What it does. Zapier describes a small garden business in England. The owner records a voice note and submits it on a form. The system transcribes it, decides whether it is an invoice, a quote or a to-do, pulls out the details, creates the invoice in Zoho Books and sends a confirmation with the invoice’s details.

What it needs. A form tool (Jotform), Zapier, OpenAI’s Whisper for transcription, ChatGPT 3.5 for sorting and extracting, and accounting software.

What the source says. The builder says each request “costs less than a cent”. The walkthrough has no approval step before the invoice goes out, and notes that one can be added. The page is undated and names ChatGPT 3.5, so check which model you would use today. For a client-facing document, add the approval step.

3. AI tools that cut admin time in Greater Manchester

What it does. A case study published by Gatsby describes a business innovation adviser at Hopwood Hall College who has delivered AI training to more than 600 businesses since 2024, under a programme funded by Innovate UK’s Further Education Innovation Fund. C The Solution, a professional coaching business with “minimal knowledge and no prior use of AI”, was shown ChatGPT, Claude, Otter AI and Gamma for admin, research and client-facing materials. BAS Recruitment, a financial recruitment consultancy, was trained to use ChatGPT for job adverts and marketing, and other tools for outreach.

What it needs. Off-the-shelf AI tools, a few hours of hands-on training, and follow-up support. The case describes people using the tools, not automations that run on their own.

What the source says. “Over a 12-month period, C The Solution saved approximately 180 hours of time, equivalent to around 23 working days”. At BAS Recruitment, “work that previously took one to two hours can now be completed in 10–15 minutes”. The case study file is dated 30 April 2026. The figures come from the case study, which does not say how they were measured.

4. Support tickets handled by fixed steps

What it does. Koralplay, a gaming software company of around 85 people, automated blocked-payment support tickets in one market. The workflow logs into the back office, looks up the transaction, checks regulatory status, updates the ticket in Notion and tells the customer.

What it needs. n8n, and access to the back-office systems. The case page describes no AI step in this workflow.

What the source says. n8n’s page reports 70% of those tickets automated. It quotes the company’s COO: “It now saves us 616 hours per week”, and “What took 10-15 minutes per ticket now happens in about 70 seconds.” The page puts the saving as “616 hours (equivalent 14 FTE)”. That is a larger company than most readers run, and the numbers are the vendor’s. It is here because it shows a job done on fixed rules with no guesswork.

5. Proposals built in under a minute

What it does. BeGlobal, a Dutch seller of branded corporate gifts, automated its commercial proposals. A chat asks for product type, budget per unit, occasion and client, then searches a database of about 100,000 products, builds themed visuals and fills a Google Slides template.

What it needs. A product database, CRM data and past offers, a GPT step, n8n and a slide template. The system was built by an outside firm, PromptGorillas.

What the source says. Offers rose from about 50 a year to nearly 500, and the workflow takes “around 52 seconds” end to end. The chat agent “summarizes the inputs and validates them with the user before proceeding.” The page does not describe anyone checking the finished presentation, and it does not state the company’s size.

6. Examples from Harrowtech.ai

These are Harrowtech’s own client jobs, with the clients’ reviews from its home page.

All the numbers in one view: 6 hours a week saved. Michael Jones, a commercial landlord, needed “a combined BI view of all of our operational spreadsheets”. He wrote on 20 April 2026 that Harrowtech built an analytics dashboard “that was both interactive and displayed all of the linked data visually”. It saves him 6 hours a week, Harrowtech’s best saving so far.
A daily lead finder. Ovais Khan of saifcarpets.com wrote on 4 June 2026 that it “runs every day, finds leads, qualifies them, adds as much info as possible and creates a spreadsheet”, with a pre-written outreach email that he reviews before setting the status to send.
Also on the service list: quotes and invoices, paperwork, inbox and diary, and content. Each is checked by you before anything leaves, and where a job can run on fixed rules, fixed rules are used.

The first automation starts from £950, after a free 30-minute first session and a fixed written quote. Prices on the site are starting points, not caps. See AI automation for UK businesses for how it works.

What the examples have in common

Example 4 uses no AI to do the main work. Examples 2 and 5 use AI to read or write, and put fixed steps around it. Example 3 is people using AI tools directly, with no automation joining them. Example 1 is still a trial. That spread fits the advice in the guide on how to automate your business with AI: one job, fixed rules where they will do, approval in front of anything that goes out, and a check on personal data.

What this guide does not cover

It is not a survey of how many UK firms use AI; the government report guide covers adoption. It does not prove any saving for your business. The vendor and programme figures cannot be checked from the pages, and the bakery is confirmed by government sources only as a trial. Client reviews are on Harrowtech’s own site.

Sources, all read 13 October 2026

Government puts AI to work for bakers, road workers and more, GOV.UK, 14 January 2025
Over £7 million awarded to help AI boost growth in the UK, UKRI, 14 January 2025
How a small gardening business automates invoicing and expenses with AI, Zapier (undated)
Helping local business adopt AI: Greater Manchester Colleges, Gatsby case study (PDF, file dated 30 April 2026)
Koralplay case study, n8n
BeGlobal case study, n8n
Harrowtech.ai home page, Work section and client reviews