How to Make ChatGPT Into a Jarvis-Style Assistant

ChatGPT can work as a Jarvis-style assistant — not an autonomous one that runs your home, but a capable research and planning partner that stays on task and picks up where it left off. The key ingredients are well-written instructions, a clear scope for each session, and a way to carry context forward. Tools like fenbs store your project rules and reusable prompts outside any one chat, so every session starts with the same briefing without retyping it.

The gap between a one-off chat and a Jarvis-style setup is continuity. This guide covers what to put in your instructions, which workflows fit best, and how to keep the setup working across sessions and tasks.

What the Jarvis idea actually means in practice

The fictional Jarvis is always on, knows everything in advance, and acts without asking. ChatGPT is different: it needs context supplied at the start of a session and works by responding to prompts rather than acting on its own. That is still genuinely useful. Think of it as a briefing partner and research assistant rather than an autonomous agent. It can draft a morning summary, research a question and return the key points, write a first version of a document, and track what was last agreed — as long as you give it the project record to work from.

• Research: gather information on a topic, synthesize it, and surface the parts that matter.
• Planning: draft a task list, outline a project, or write the steps toward a goal you describe.
• Drafting: write a first version of a document, email, or proposal in the tone and format you specify.
• Reviewing: check a draft or plan against criteria you set and flag what needs attention.
• Tracking: record decisions, save a progress note, and confirm what is waiting on a person.

Setting up ChatGPT as your briefing assistant

The starting point is a clear set of instructions. In ChatGPT, these live in Custom Instructions (account-wide) or in a Project's custom instructions (scoped to one area of work). A good instruction set covers who you are and what role the assistant fills, how you want responses formatted, and any standing rules — what to avoid, what to prioritize, and how specific to be. Keep it short: two or three paragraphs work better than a long list.

1. Open Custom Instructions in your ChatGPT account menu, or create a Project and add instructions there.
2. Write a short brief: who you are, what this assistant helps with, and how it should respond.
3. Add standing rules: "Always cite sources," "Keep responses under 200 words unless asked," or "Ask one clarifying question if the task is ambiguous."
4. Test with three tasks you do regularly, and refine the wording based on the results.

Building a research and planning workflow

A research workflow turns ChatGPT into a faster version of spending an hour reading. The pattern is: describe the question, give relevant context, ask for a summary with key points and gaps. Review the output and follow up on the parts that matter. For recurring tasks — competitive analysis, weekly topic checks — create a template prompt you can reuse.

• "Summarize the three main positions on [topic], with one sentence on where they disagree."
• "Find the key steps in [process] and flag any steps where standard advice differs between sources."
• "Compare [Option A] and [Option B] for [use case] and give a short conclusion on which fits better and why."
• "Draft a weekly plan for [goal], with one concrete task per day and a check-in question at the end."

Save the templates you reuse most. A notes app works, or you can keep them in a shared prompts library that a connected assistant reads directly. The goal is that opening a new session never means rebuilding the workflow from memory.

Carrying your context across sessions

ChatGPT does not retain the details of previous sessions by default. Project instructions and Custom Instructions persist, but the specific state of a task — what you decided, what remains — does not. The fix is a short progress note at the end of each session: the active task, what was completed, what was checked, and what comes next. Paste it at the start of the next session to resume without a recap.

A progress note is usually three to five sentences. It does not need to be a detailed log — just enough for the assistant, or for you a week later, to start useful work immediately. For team projects where someone else might open the next session, this note becomes the handoff document.

How to make ChatGPT into Jarvis with fenbs

fenbs handles the continuity layer: reusable prompts, project rules, and session handoffs, all in one place ChatGPT reads before it starts work.

1. Create a free account at fenbs and add a project for your Jarvis-style work.
2. Write your standing rules in Decisions and rules: what ChatGPT should always do, what it should avoid, and any constraints for this project.
3. Add your most-used prompt templates to Skills and prompts.
4. Connect ChatGPT from Integrations and ask it to read the project context before starting each session.
5. At the end of each session, ask ChatGPT to save progress to fenbs: what it completed, what it checked, and where the next session should start.

Once connected, a new ChatGPT session opens with your rules and current task already loaded. For more on building persistent assistant workflows, what an AI brain is explains the broader concept behind this setup.

When a Jarvis-style setup has limits

This approach works well for tasks that produce reviewable output: research, drafts, plans, summaries. It works less well for fully autonomous action — accessing external accounts, sending messages, or running background processes without your review. For multi-step autonomous execution, ChatGPT agent mode covers how ChatGPT Work handles those tasks. For prompt templates tailored to office work, ChatGPT prompts for work has patterns you can adapt directly.