ABONVERO LEARN · AI AT WORK

Use AI at work
without getting burned.

Practise the habits that separate useful AI work from embarrassing AI work: prompts that say what you actually want, checking every number before it leaves your hands, keeping private data out of chat boxes, and setting up AI agents that can’t do more damage than their job requires. The first 2 lessons are free.

Start lesson 1 ↓

LESSON 1 · TASK, AUDIENCE, FORMAT, LIMITS

Writing a prompt that gets a usable answer

Most disappointing AI answers come from prompts that leave the important choices to the model. Say what you want done, who it’s for, what shape the answer should take, and any limits. The model can’t ask what you meant before it starts.

1/2 Your manager asked for a summary of last quarter’s customer complaints for the leadership meeting. Rewrite this weak prompt: “summarize the complaints”.

Weak prompt: summarize the complaints

This checklist looks for common wording. Compare your answer with the example to judge whether it meets the task.

  • ○ task
  • ○ audience
  • ○ format
  • ○ limit
  • ○ source

Practice, not a list of tips

There’s no shortage of “10 ChatGPT tips” articles. Reading them doesn’t change what you type on Monday morning. Each lesson here makes you do the thing: rewrite a weak prompt until the coach stops finding gaps, flag the sentences in an AI summary that the source doesn’t support, or choose what access an agent should get and see why the other options are risky. You get specific feedback straight away, the same way the spreadsheet lessons point at the exact mistake in a formula.

Prompts that get a usable answer first time

Most disappointing answers come from prompts that leave the important choices to the model. The prompt coach checks that you’ve said what the task is, who the result is for, what shape it should take, how long it may be, and what material to work from. When something’s missing it names it: “Who will read it?”, not just “try again”.

Checking AI output is now part of the job

AI writes confidently even when it’s wrong. In the checking lessons you compare an AI answer with its source and flag what doesn’t hold up: a swapped digit, a growth figure from data that was never provided, a person who wasn’t in the meeting. It’s the skill that protects your name when AI-drafted work goes out under it.

Managing AI agents

Agents don’t just answer; they act. They read inboxes, update spreadsheets, call systems. The agent lessons teach the management side: give each agent only the access its job needs, keep payments, deletions and outside emails behind a human approval, recognise when text inside an email is trying to give the agent orders, and write instructions with a clear goal, a defined scope and a condition that makes the agent stop and ask.

What you’ll learn

  1. Writing a prompt that gets a usable answerTask, audience, format, limitsFree
  2. Checking AI output before you use itVerifying facts and numbersFree
  3. Giving context and an exampleContext, examples, step-by-stepPro
  4. What not to paste into an AI toolData protection and company policyPro
  5. AI agents: tools and permissionsLeast privilege for agentsPro
  6. Managing agents: steps, checkpoints and stoppingDesigning and supervising agent workflowsPro

Want to run models on your own computer instead? The local AI and agents course covers hardware, model sizes and running agents safely, using numbers measured on our own machines.

Abonvero Learn is independent and not affiliated with OpenAI, Anthropic, Google or Microsoft. Product names are trademarks of their owners. Operated by Andronykus Technologies LLC · contact@abonvero.com