Welcome to the team. Here's the markdown. Photo by Brooke Cagle on Unsplash.AI: Think of them as a summer intern
Stop treating coding models as a tool on your machine. Encode the company and give always-on AI a job.
Twelve months ago, coding models were on par with a not-too-skilled intern. Today they are on par with an experienced senior.
Where I have gone wrong, is thinking of AI as a tool on my machine – rather than as a part of the online company infrastructure. In this case I’ve realised the potential of thinking of them as a newly hired summer intern.
- Tell them about us
Encode company knowledge and workflows in markdown. Show them where they can get more information if need be.
With clear boundaries and playbooks, both interns and AI can take a task end to end without you in the loop.
- Give them narrow tasks (less context)
You slow down an intern with a vague problem description and lots of considerations. Same with AI.
Define the different tasks involved in doing support, rather than access to the support inbox and tell it to solve tickets.
- Diff: You can hire a 1% or 2000% role
Interns should have context, responsibilities and expected to gain/knowledge that fit a ~100% role. Unlike AI, where you can “hire” a role that only does SMS translations – whether that would take one day per month, or 20 interns.
Remember: Less context is better.
- Diff: hire and fire without consideration.
A human intern needs hiring and contracts and weeks of onboarding. You cannot fire them with a keystroke.
Try one AI intern. If it works, hire a thousand. If it doesn’t, respecify the role. Or fire them. No worries!