Give it direction.
Describe the outcome you want and the boundaries it should follow.
AI is moving beyond answers. Agentic systems can plan, use tools and take action toward a goal, with people setting the direction and staying in control.
Traditional AI helps you create or understand something. An AI agent can work through a task in stages, use available tools and adapt as it goes. The best systems make their actions visible and leave important decisions to people.
Describe the outcome you want and the boundaries it should follow.
The agent breaks the task into steps, gathers information and uses connected tools.
Review the result, check its reasoning and approve consequential actions.
Agentic AI is useful when work takes several steps, calls for judgment along the way or spans multiple tools.
Explore a topic, compare sources and turn findings into a useful starting point.
Coordinate routine tasks across documents, calendars and other connected systems.
Move from an early idea to a draft, prototype or working digital experience.
Choose a starting point, adapt the prompt to your situation and paste it into an AI assistant that can work through multiple steps. Review its sources and actions as you go.
Help me understand [topic] so I can explain it to [audience]. Break the work into steps. Find current, credible sources if web access is available, and distinguish established facts from opinions or predictions. Give me a plain-language summary, three key takeaways, open questions and links to the sources you used. If you cannot verify a claim, say so.
Check the original sources before sharing the result.
I need to choose between [options] for [purpose]. My priorities are [criteria], and my constraints are [budget, location, deadline]. First ask me for any missing details that would change the recommendation. Then compare the options in a table, show trade-offs, identify assumptions and suggest a choice with reasons. If you have web access, use current primary sources and link them. Do not invent missing prices or features.
Treat the recommendation as a starting point for your own decision.
Help me turn [idea] into a practical plan by [date]. I have [time, people and resources]. Ask up to three important questions if needed. Then propose milestones, the first five concrete actions, likely obstacles and a simple way to track progress. Mark assumptions clearly. Before sending messages, spending money, changing files or taking other external actions, show me exactly what you propose and wait for my approval.
You decide which actions the agent may take.
We’re working toward agents that can help with multi-step tasks, use connected tools and make progress with clear human oversight. Those agents are not available on this site yet. For now, you can explore the concept and copy practical prompts to use in your own AI assistant.
A space to explore what intelligent agents can do, and how we can use them thoughtfully.