# Ten AI-first principles

**URL:** https://aibility.org/en/docs/coaching/principles
**Meta description:** The principles behind coaching and AI Workouts. From each situation of yours, coaching picks at most two that will help you most right now.

---
Coaching is built on ten AI-first principles. From the task you describe, it picks at most two that will help your work most right now and turns them into one experiment. Here are all ten, each with an example from everyday work.

## The principles

### AI First

Try AI on a real task before you start working the old way. Give it the goal and the material, and let it make the first attempt. Instead of a blank page, have it build a first draft of a reply to a customer.

### Choose the tool for the work

When one AI chat is not enough, use a tool that can handle what the task requires. Chat is good for questions, text and a few files, while large data or whole folders need another approved tool. So analyze a big spreadsheet in a tool that can run calculations on the data.

### Brain vs muscle

For a new or unclear task, first use AI to think, compare options and challenge your assumptions. Once the direction is clear, use it for fast execution. Before a proposal, have AI compare approaches, and only then prepare a first version of the one you chose.

### Everything is a prompt

Instead of retelling something at length, show AI the real material: a meeting transcript, a screenshot, a table, an email or a voice note. Say what you need from it. From a meeting transcript, have it prepare notes and open questions.

### Context is king

AI answers better when it knows your situation, your goal and the rules that apply next time too. Save repeated context and keep it up to date. For a proposal, AI then knows the client, the constraints and an example that worked.

### Iterate

Treat the first result as a start and say specifically what to keep and what to change. In a monthly report, you keep the charts and have only the commentary for management rewritten.

### Verify

For important results, compare claims and numbers with the original material. Match the depth of checking to the cost of an error. Before you send a report, compare the totals with the data export.

### Work is a system

Build a larger task from connected steps that AI can help you design and carry out. For a report, AI first prepares the data, then an overview of changes, and finally the commentary.

### OHIO: Handle it once

When work with AI goes well, have it prepare something usable for next time right away. A second similar task is already a signal to save a prompt, template or checklist. After the monthly report, have AI write down the routine and the rules for next month.

### AI can program

When you are missing a tool, describe the result you need and have AI create a small piece of software. A spreadsheet can become a dashboard with a switch for periods and departments, and you do not need to know how to code.