Get Started With ChatGPT: Your First Reviewed Work Session

To get started with ChatGPT, open a chat, give it one small task using information you understand, and check the result against that information. Your first useful session should end with something you reviewed, such as an accurate meeting agenda or a practice exercise you answered yourself. Start small enough to recognize a mistake.

Tools like fenbs can keep the reusable instructions and project context for later connected assistants. You can add that continuity after your first exercise, once you know which instructions actually helped.

1. Open the official ChatGPT service and follow its sign-in instructions.
2. Choose a task with a source you can check directly.
3. Send a short request with the source and expected output.
4. Compare the answer with the source before making one focused revision.
5. Save the accepted result, reusable instruction, and next step.

What should you open for this first exercise?

Use ChatGPT in your browser if you want a straightforward starting point without installing an app. OpenAI’s quickstart↗ (opens in a new tab) describes the current web and desktop setup. Follow that page when the names of controls differ from what appears on your screen.

For this exercise, use a normal conversational chat and a short text input. You do not need to connect a work account, upload a folder, or explore every tool before learning the basic loop. Those choices can wait until a real task requires them.

If you need the broader account and orientation walkthrough, read getting started with ChatGPT. Here, the goal is to finish one reviewed piece of work and understand why it worked.

Which task gives you a result you can judge?

Choose a transformation of information you already know. Turning your own notes into an agenda is easier to verify than asking for advice on an unfamiliar legal or medical question. A study exercise can use a short passage you have read, provided your course allows AI for that activity.

Use the invented notes below if you do not have suitable material. They contain an agreed goal, a proposal, an unanswered question, and a next action. Those distinctions give you something more useful to inspect than whether the output sounds polished.

Project: a neighborhood reading event.
Agreed goal: prepare a simple event page.
Proposed: include a form for volunteer interest. No one has approved it yet.
Unknown: whether the library will provide a room.
Next action: Jordan will ask the library about availability.
Meeting purpose: agree what information the page needs before drafting it.

Use invented names and details while learning. For a real task, follow your organization’s rules about what you may share. Removing a name alone may leave identifying details, so a completely invented example is often the simpler first test.

What should your first message ask for?

Ask for one deliverable in a format that makes important distinctions visible. For the event notes, request a meeting agenda with a separate list of unresolved questions. Explain that proposals must remain proposals so the assistant has a clear boundary.

Turn these project notes into an agenda for a short planning meeting.
Audience: three volunteers who know the project.
Output: meeting goal, three discussion points, unresolved questions, and next action.
Use only the notes. Keep proposed ideas separate from agreed decisions.
Do not add dates, commitments, or event arrangements.
Notes: [paste the invented notes above].

OpenAI’s prompting guide↗ (opens in a new tab) recommends stating the result and adding relevant context and boundaries. The example does that without requiring technical language. The source and output format make the response easier to inspect.

For a study version, supply a passage and ask for one question at a time, with feedback only after you answer. Keep the answer key hidden until you have tried. Your goal is to expose a gap in understanding rather than collect a finished response to submit.

How do you decide whether the response is usable?

Compare the answer with the notes line by line. In the event example, a volunteer form is only proposed, and the room is unknown. An agenda that says the event will take place at the library has added a fact you did not supply.

• Does the meeting goal match preparing the event page?
• Does the volunteer form remain an idea awaiting a decision?
• Is room availability still an open question?
• Does Jordan’s next action remain asking the library, rather than booking it?
• Has the assistant invented a date, cost, audience size, or promise?

Mark each factual claim as supported, unsupported, or unclear. You do not need a score; you need to know whether the output can be used for its purpose. A concise answer that changes a commitment needs correction before a longer answer with an awkward phrase.

If you used real material, check the source itself rather than relying on the assistant’s explanation of where it found a fact. Asking it to review its answer may expose issues, but you still need to inspect consequential details.

What should you send when it gets something wrong?

Give a focused correction tied to the source. For example: “The room is unconfirmed. Move that into unresolved questions and remove wording that says we have booked it.” That tells the assistant what failed and what a successful revision must preserve.

Read the entire revised answer again, because a correction can introduce another change. Once the facts are right, refine presentation if needed: shorter sentences, a clearer order, or more explanation for a new reader. Keep the accepted facts stable while changing style.

For a study exercise, explain your own reasoning when you disagree with feedback and compare both answers with your course material. If you cannot resolve the difference, mark the question for an instructor or another authoritative source instead of accepting the most confident wording.

What is worth keeping after this session?

Keep the accepted agenda in the place you normally keep meeting documents. Save the reusable part of the prompt separately: output sections, the instruction to preserve uncertainty, and the check for unsupported commitments. Replace the event-specific notes before using it again.

Also save a short progress recap if the project will continue. For the example: the planning agenda is ready, the room is unconfirmed, and the next action is Jordan’s availability check. That recap describes the state of the work more precisely than “we worked on the event.”

In fenbs, Skills and prompts and project context give those two records suitable homes. The reusable instruction explains how to work; the current context explains what is true for this project.

How to do this with fenbs

1. Create a free account, free for one person with no credit card, and add your recurring work or study project.
2. Save the instruction you tested in Skills and prompts, with named inputs and a short acceptance checklist.
3. Add context notes for the project’s current goal and unresolved questions. Record decisions only when a person has made them.
4. Connect a supported assistant through Integrations and ask it to read the prompt and project context before continuing.
5. Review the next result and save progress with what was checked, what remains open, and where to begin next time.

fenbs does not import your ChatGPT conversations or check the answers for you. A one-time exercise can stay in a document; fenbs fits when you want reusable instructions and an ongoing work record across sessions or assistants.

Continue with the AI prompts hub for other task formats. ChatGPT for business covers workplace considerations, while Claude versus ChatGPT addresses choosing another assistant after you understand your task.