A few years ago, using AI meant typing a question and reading an answer. Today's assistants can read documents, look at images, browse the web, write and run code, and complete multi-step tasks. That shift changes what good prompting looks like.
From answers to agents
“Agents” are AI systems that can plan and carry out several steps toward a goal, using tools along the way: searching, filling in a spreadsheet, editing files. With agents, your prompt becomes more like a brief for a capable colleague: the goal, the constraints, what “done” looks like, and when to stop and ask.
Memory and context
Many assistants can now keep information across conversations or work with very long documents. This makes it worth investing in good context: a project description, your preferences, examples of past work. The better the background you provide, the less you need to repeat yourself.
Multimodal by default
You can increasingly mix text, images, screenshots, audio and files in a single request: “Here's a photo of my whiteboard, turn it into a project plan” or “Here's a chart, explain the trend.” Prompting skills now include choosing what to show, not just what to say.
What stays the same
The fundamentals of a good prompt haven't changed:
- Be clear about the goal and the audience.
- Give relevant context and examples.
- Specify the format of the result.
- Check the output, especially facts and actions with real consequences.
How to prepare
Practise writing briefs, not just questions. Keep a library of your best prompts and context documents. And keep a human in the loop for anything that sends messages, spends money or changes important data, since more capable tools make careful review more important.