What the client said. What you agreed. What is still missing.
Decisions, next steps and questions—in your usual format.
Read it once. Correct the details. Keep the version that works.
Using AI in a business means applying it to a specific job.
Drafting a client follow-up. Organising project notes. Preparing content.
Or turning written requirements into a small tool.
A useful first project has clear inputs, a result you can check, and a reason to do it again.
Want one ready to try? Get the free AI business starter.
What should you use AI for first?
Start with something familiar enough that you can recognise a good answer. You do not need a perfect prompt or a new subscription.
Pick a task that has been sitting in your inbox, a repeated bit of admin, or a document you recreate for every client.
- Client follow-ups: turn meeting notes into a draft with decisions and next steps.
- Content preparation: turn a voice-note transcript into three post outlines.
- Project organisation: turn your notes into a list of open decisions and missing information.
- Internal tools: describe the fields and steps in a tracker that matches your process.
Choose the task where you can spot a good result. If you write client recaps every week, you already know what belongs in one.
1. Describe the job before choosing more software
Write down what happens now. Where does the information come from? What do you create?
Who uses it? What do you keep correcting? This short description is more useful than a long list of AI tools.
After a client call, you read the notes, write a recap and decide what to ask next.
The first AI task is drafting that recap. It is not replacing your client relationship or rebuilding the entire business.
2. Give it the context you would give a colleague
Provide the source material, the audience, your preferred format and an example of a good result.
If there are details it must not invent, say so. Start with material you can share with the tool under your business's data rules.
Turn these meeting notes into a short client follow-up.
Include decisions, next steps, owners and dates only when the notes state them.
Put missing details under “Questions to confirm.” Keep the tone friendly and direct.
Here is a previous follow-up I like: [example]. Here are the notes: [notes].
For a content task, replace the meeting notes with a transcript and the follow-up format with a post outline.
The same structure still works: source, audience, output and example.
3. Test it on three different examples
Try a straightforward case, a messy case and a case with missing information.
Read each result alongside the source. Notice whether you are making the same corrections repeatedly.
- Did it include the important details?
- Did it invent a date, promise or decision?
- Does it sound like something you would send?
- How much editing did it need?
- Was the whole task easier once checking time was included?
When the same correction comes up twice, add it to the instructions. The messy example is often the one that teaches you what was missing.
4. Save the instructions that worked
Keep a small document containing the job description, the reusable instructions and one approved example.
Update it when your process changes. This gives you a repeatable starting point instead of rebuilding the prompt each time.
For the next layer, read how an AI workflow combines context, tools and feedback and how to write rules for an AI system.
5. Connect tools only when the manual version is useful
A chat can help with a task you start yourself. An automated workflow also needs a trigger, access to its inputs, a destination and a way to handle errors.
Add those pieces when the draft-and-review process already works.
In the client example, the next step might be bringing approved meeting notes into the drafting process automatically.
Sending the message is a separate decision. You can keep approval with you while automating preparation.
How do you know whether it is worth keeping?
Compare the entire process, not just the seconds it takes to generate text.
Record the time spent gathering inputs, running the task, checking the answer and making corrections.
Also notice whether it helps you start work you normally postpone. Use a weekly business review to decide what to keep, change or stop.
- Task: ___
- Inputs: ___
- Useful output: ___
- What I check: ___
- Time before: ___
- Time including review now: ___
- What I will change next: ___
Do you need coding skills?
Not for the draft-and-review examples in this guide. Connecting apps or creating a custom tool adds setup work.
Begin with the smallest useful version, then decide whether the next step is worth learning or getting help with.
What is the difference between an assistant and an automation?
An AI assistant helps with tasks through conversation and context. An automation runs a defined process when a trigger occurs.
They can work together: an assistant can draft the message inside a larger workflow that collects notes and asks you to approve the result.
New to the vocabulary? Our AI glossary explains assistants, workflows and context with everyday examples.
Build it with us
Switchboards runs practical AI workshops for founders, freelancers and small teams. Bring a real task and work through the setup with Dani.
Start with the free personal AI assistant guide, or join the free newsletter below for examples and workshop announcements.
Your AI Weekly Business Review
Give your assistant a repeatable job: compare this week with last week, check spending, find what is stuck and prepare one improvement.
Includes the same dashboard starter so you can try it immediately.
Get the free starter →Subscribe for instant access. AI tips and workshop invitations; unsubscribe anytime.