The best hire you're going to make this year costs about $20 a month. It works seven days a week, doesn't take holidays, doesn't quit and do all your boring work without complaining.
You've probably guessed where this is going :) Your next employee is AI!
I know you keep hearing that AI is a great productivity tool and if you pick the right tools and follow the right gurus you’re going to save so much time and money.
But my guess is you’ve tried AI and been largely underwhelmed and have been left wondering what this “AI miracle” is that everyone keeps talking about.
So I want to try and reframe that for you.
You need to stop thinking about AI as a productivity tool and start thinking about it as a staff member who needs to be managed and delegated to properly.
How to manage your new “AI employee”
Every time you delegate something to AI you need five things to happen to get a good result.
1: Brief
A brief is everything you need to tell a new employee on how to do that task.
What you’re probably doing with AI at the moment is so unhelpful that your employee would have no idea what good would look like e.g. "Turn these site meeting notes into something I can send the client."
This is way too vague and that’s why you are getting a bad output. Here's what is better:
“Turn my site meeting notes into a written summary I can email to project stakeholders. The meeting was on site yesterday with the client and the builder. I'll paste in my notes. The summary needs to confirm decisions made, flag any items still requiring client sign off, note anything the builder has been asked to action and list anything I've been asked to follow up on. The first step is to review my notes and make sure you’re clear on everything. Ask me any clarifying questions before moving forward”
The reason this is better is because it’s giving context and constraints plus a clear first task.
2: Context
This is the where you train AI on what good looks like.
To do this take three or four examples of the good version of whatever you're asking it to do. Past these in with your prompt and explain why they're good.
For example, if you're getting AI to write client facing emails in your voice, you'd take five emails you've written that you’re happy with and paste them in and say: "These are emails in my voice. Notice that I write in full sentences, never start with 'Hope you're well', use specific details from our conversations and never use exclamation marks. I don’t use corporate tone or empty pleasantries. Do you understand my voice?"
You will be amazed at how AI's output improves once it has been given proper context.
3: Scope
At the moment you’re likely asking AI to do things that are too complex or too vague e.g. “Help me with my marketing" is not a task but “Draft a 200 word Instagram caption for the project I just photographed, in my voice, that opens with a specific design decision rather than a generic comment about how much we loved this project" is a task (you’ll get even better results if you add the images of the project you’re planning on posting as context).
The smaller and more specific the scope, the better AI will do. Think of AI as a high level graduate who needs the work broken into small clear chunks.
4: Review
Being a good manager means reviewing output and giving specific feedback e.g. “This draft is too formal. The opening sentence sounds like a press release. Rewrite it as if you were telling a friend what we did but maintain the structure of the rest of the draft." or another example "You used the word 'curated' three times. I never use that word in my writing so replace each instance with something more specific to what we actually did. Also add to your memory that I do not use the word curated”
Providing feedback is how you’ll teach your AI more about how you want it to perform.
5: Systematise
The final step is to capture the work you’ve done for next time. This can take many forms depending on your current proficiency with AI. Here are some examples:
✅ Level 1: Ask AI to write a prompt you can use next time to get the same output you ended up with this time. Save that prompt in a prompt database you keep in Notion for next time.
✅ Level 2: Ask AI to save the workflow you’ve just done as a skill. That way it knows how to run the workflow you’ve just done again next time without having to repeat yourself.
✅ Level 3: Load the whole thing into a permanent Project (in Claude). You give it the brief, voice, context and examples of good finished examples. From that point when you need to do this task you just open the Project, paste your raw site notes and get back a summary in exactly the style and format you want.
Whatever level you choose you now have a system you can run again next time this piece of work needs to be done (without doing the entire process from scratch again).
What changes when you start treating AI as an employee, not as a tool
Stop thinking about AI in terms of prompts and start thinking about it in terms of tasks, roles and systems. If you do this you’ll eventually have lots of “AI assistants” that know how to do workflows and tasks in your business e.g. this AI assistant writes my client emails, this one drafts my Instagram captions, this one turns my notes into email drafts. Each one has a clear task and set of instructions to follow.
This is when AI stops being a disappointment and you find ways for it to save you time and money.
To do this ask yourself “How do I build a team of AI assistants for my business?”
The easiest way to put this in practice today is to to pick one task that you do regularly and don’t enjoy and open your preferred AI (mine is Claude at the moment).
Give it the brief, context, constraints, teach it your voice and the structure of what you want and what to avoid. Take time and treat it like an investment in a new employee who you want doing this work for a long time.
The great news is that once you’ve built a whole range of workflows and systems you now have a team of staff members that cost almost nothing, don’t quit, don’t take holidays and are always happy to keep working on things for you. They can also all be linked together to work on larger projects (but that's a conversation for another time!).
Welcome to having AI employees :)
