Get started with AI. Start with your life.

If you’re new to this, open ChatGPT or Claude and start with a free account. Either is enough to try the exercise below. Use the default model setting for your first attempt; you can learn the finer controls when you have a reason to.

A chat assistant helps you think and produce work in a conversation. A personal agent can also use connected tools and keep working on tasks, depending on the product and permissions you give it. You can learn the basic habit before adding those capabilities.

If you’re ready for an agent, ChatGPT Dot and Grok Bot are the two I’d suggest exploring first, based on my own use. They understand my context and suggest useful next steps. That makes them feel more useful to me than starting over with a blank prompt each time.

There is an access difference worth checking before paying. As of October 27, 2026, Dot is rolling out to eligible ChatGPT Pro and workspace plans; an ordinary Plus subscription does not include it. Grok Bot requires an eligible plan; for new X subscriptions, that means Premium or above. Availability and limits can change.

Start free, then consider one paid subscription when you know what you need more of. A plan upgrade can give you access or capacity. It won’t replace a clear goal.

Context and useful suggestions are the common thread. Grok Bot also makes the experience approachable for me. I like the friendly Bot characters and the way specialized Bots can work together. I spend less time thinking about model and reasoning settings and more time explaining the result I want. In my use, it feels as though it’s improving quickly.

Its automatic model selection includes third-party models such as Claude Opus. I appreciate that willingness to use another company’s model. To me, it signals an emphasis on the result rather than insisting one provider has to do everything.

I’ve also used Hark and Meta Muse. Muse is capable, but my concerns about social media’s influence on young people make me more cautious about trusting Meta with my personal context. That’s a personal choice; you may feel differently. I’ll explore the wider question in a separate article.

You may make a different choice. Find an assistant you’re comfortable working with, then spend enough time with it to judge the work. You don’t need to subscribe to all of them.

Tell your assistant a little about your work and life, what matters to you, and what you’re trying to change. Include the constraints: limited time, a small budget, family responsibilities or a skill you’re still learning.

You can type this or talk it through if your tool supports voice. Keep it relevant. You don’t need to upload every document or share private details about clients, colleagues or family members.

Try this as a first message:

I’m working toward [one goal]. My experience is [a few relevant strengths]. What matters to me is [values or priorities]. I have [time and other constraints]. What keeps getting in the way is [one obstacle]. Ask me up to three questions, then suggest three useful things you could help me do this week. Explain why each would help, and wait for me to choose one.

Then ask it to summarize what it understood. Correct the parts that are wrong. Keep a short copy of that context brief so you can reuse and update it; memory works differently across products and isn’t a guarantee that every detail will carry forward.

The AI starting-point worksheet helps you do this and creates an editable prompt. Your answers stay in the open page until you choose to copy or download them.

Choose a task small enough to finish and useful enough to matter. A fictional example: a volunteer organizer has scattered instructions for a community event. Their first task could be turning those notes into a one-page welcome checklist, with missing details clearly marked.

For someone exploring a career change, it might be a comparison of two possible paths and the questions to ask someone already doing that work. For a small business owner, it might be a draft answering five common customer questions.

Give it a finish line: who the work is for, what format you want and what would make it useful. Ask it to point out missing information and assumptions. Review the result against your actual circumstances and check factual claims against their sources.

Andrew Warner recently shared a useful distinction on X: his agents became more purposeful when he gave them goals rather than isolated tasks. For a first attempt, I’d keep that goal small and the review close. You can give more responsibility as you learn what the assistant handles well.

Tell your assistant to challenge weak assumptions, ask when important details are missing and be clear about uncertainty. Useful help sometimes means telling you the plan needs work.

Keep external actions deliberate. Ask to see the exact message, purchase or change before it happens, and use the product’s permission settings too. A sentence in a prompt is guidance; it isn’t a technical access control.

You don’t need to connect email, a calendar or your files to complete this first exercise. Add a connection when a specific task calls for it and you understand what access you’re granting.

At the end, ask yourself: did this help me think more clearly or complete something worthwhile? What needed correcting? What would I ask differently next time?

Give yourself 20 minutes. Choose one assistant. Share a short context brief. Pick one task. Review what comes back.

That is enough for today. You’re learning how to explain, question, evaluate and build with a new kind of tool. Those habits matter even as the products change.

In the next article, we’ll turn one recurring task into a reliable routine: clear inputs, a useful output and a human review point. You can also read how to choose your first AI workflow now. After that, we’ll explore turning what you know into something other people can use.

I’ll share more of my own AI lab and building setup later, including the parts that require more time and care. You don’t need that setup to take this first step.

Credit to Alex Finn, Peter Diamandis and the Moonshots team for helping move me from interest to action. The Moonshots conversation that inspired my early OpenClaw experiment is part of that history, not my current setup recommendation.

For more recent walkthroughs, start with Alex’s October 1 video on getting useful results from ChatGPT Dot and his October 7 Grok Bot overview. These are demonstrations from his experience; my recommendations above come from my own. Check the current product guides before choosing a plan or granting access.