The short answer
Building with AI means turning a task you repeat into something you set up once and reuse, instead of typing a fresh prompt each time. Start small: pick one task you do every week and write it down in four parts: what starts it, what happens, what gets checked, and who it goes to next. The five builds below need only tools many businesses already have, and each comes with a way to tell whether it is working.
Why it is worth fixing
Workday’s Global Workforce Report, based on job data from employers using Workday Recruiting, found that demand for basic AI skills such as simple prompting peaked and then fell, while demand for hands-on skills such as building AI tools and automating workflows rose. The data is global rather than UK-only, but the direction is clear: knowing how to ask is becoming the baseline, and knowing how to set something up is what stands out.
For a recruitment business that matters twice over: for the people you place, and for your own team’s week.
Step by step
Write the plan in four blocks
Before you open any tool, write one line for each block. When: what starts it, such as a call ending or Monday morning. Do: what happens, in order. Check: what a person confirms before it counts. Hand off: who or what receives the result. If you cannot fill in all four, the task is not ready to build yet.
Build one: meeting notes to actions
Take candidates’ and clients’ personal details out of your notes or transcript, then paste it into the AI assistant you already use, with the same instruction each time: list each action, who owns it and when it is due, then draft a short follow-up. Save that instruction so you never retype it. Check: the person who ran the meeting reads the list before anything goes out.
Build two: a weekly report from a spreadsheet
Export the same spreadsheet each week, such as open roles or activity, and give the assistant a fixed brief: the headings you want, the comparisons to make and the tone. Keep a good past report as the example to follow. Check: someone compares the totals with the spreadsheet before the report is sent.
Build three: a house-style skill
A skill is a saved set of instructions an assistant uses whenever a request matches. Claude offers skills on every plan. Gemini has added them to its chat for personal accounts of people over eighteen, with work and school accounts to follow. ChatGPT offers them on its Business, Enterprise and Edu plans. Write your house style once: how you lay out a candidate summary, the words you use and avoid, the length of a job advert. Check: compare a few results with your best examples written by hand.
Build four: a small tool on Google Sites
In the Gemini app, use Canvas to build a checklist or quiz, such as one that checks a client’s job brief, then export it to Google Sites as a private draft, test it and publish it. Check: run a good, a poor and a half-finished example through it, and confirm every answer before you share the link.
Build five: inbox sorting rules
Not every build needs AI. A rule in Outlook or a filter in Gmail can label, move or forward mail by sender, subject or words in the message, so job board replies, client requests and supplier mail each land in their own place. Check: at the end of the week, look in each folder for anything that landed in the wrong one.
Measure before and after
Before you start, write down how long the task takes in a normal week and how often something is missed or has to be redone. Measure the same things after a month of using the build. If it is not clearly better, change the instructions or stop: a build nobody trusts is not helping anyone.
Keep a person on what matters
Anything sent to a client or candidate, anything with names or dates in it, and anything that cannot be undone needs a person to read it first. Name that person when you write the Check block, so it never falls between two people.
Checklist
- The task is written in four blocks: When, Do, Check, Hand off
- The instruction is saved once, not retyped
- A named person checks anything that goes to a client or candidate
- You noted the time taken and the misses before you started
- You measured them again after a month
- Personal details are taken out, and you know what your AI provider does with what you paste
- One build is working before you start the next
When it is worth getting help
These five builds are within reach of anyone willing to spend an afternoon on them, and they are a good way to learn what building with AI involves. It is worth getting help when a build needs to read from or write to your own systems, when it should run on a schedule without anyone starting it, or when a mistake would reach a client or candidate.
Rec Int. builds automation for UK recruitment agencies, around the systems they already run.
Questions people ask
Do I need to learn to code?
No. All five builds use written instructions and settings in tools you may already have. Code helps later, when a build has to connect systems that do not talk to each other.
Which build should I start with?
The one tied to the task you repeat most often. Meeting notes to actions suits most teams as a first build, because the result is quick to check.
Sources
- Workday Global Workforce Report: AI Is Rewriting Jobs More Than It’s Cutting Them Workday Newsroom
- Use skills in Claude Claude Help Center
- Gemini Apps’ release updates and improvements Google
- Skills in ChatGPT OpenAI Help Center
- Publish interactive apps from the Gemini app to Google Sites Google Workspace Updates
- Create rules to filter your emails Gmail Help
- Manage email messages by using rules in Outlook Microsoft Support
If you would like to talk through which tasks in your agency are worth building first, book a discovery call.