GuidesBy featureAnyone handing over a multi step task instead of a single question

Letting an assistant take multiple steps on its own

For Anyone handing over a multi step task instead of a single question.

Agentic modes plan a task, take a sequence of actions such as browsing, running code or editing files, and check their own progress along the way. They can save real time on chores with several steps, and they can also wander off course, so you set the boundaries and check the outcome.

When to use this

You have a task with several steps, like compiling a list from multiple pages, tidying a folder of files or working through a small coding change, and you would rather describe the goal than do every step by hand.

Breaks a goal into a sequence of steps and works through them in order
Takes actions such as searching, running commands or editing files, inside limits you set
Reports back on what it did and what it found along the way

Workflow

State the goal and the boundary: what it may do without asking, and what needs your sign off
Give it access to only what the task needs, nothing wider
Let it work, then read through the steps it took, not only the final answer
Check anything that changed a real file, account or message before it goes further
Stop and redirect early if the first few steps are off track, rather than letting it run

Prompts to try

Use it, adapt it, and make it your own.

I want you to [goal] by taking a sequence of steps yourself. You may [actions allowed] without asking me. Stop and ask before you [action that needs approval]. After each step, tell me plainly what you did and why. When you finish, summarise the steps taken so I can check them.

Ask Bit to adapt it →

Human lens

You decide what the assistant is allowed to touch and what needs your approval first
You set the goal clearly enough that a reasonable set of steps can satisfy it
You review the actual actions taken and the result, not just the summary of them

What to avoid

A plausible sounding sequence of steps can still drift from what you actually wanted
Broad permissions are riskier than narrow ones, so start narrow and widen only as trust builds
Always review actions that touch money, send messages on your behalf or change real files

Related guides

View all →

Using deep research without trusting it blindly

Deep research modes read many sources and return a longer, cited answer. They are powerful for a first pass, and they still need you to check the sources and the reasoning.

Using projects to keep AI in context, not starting over

Projects or workspaces let you gather files, instructions and past chats in one place, so the assistant carries the right context into every new conversation instead of you repeating it each time.

Talking to AI out loud instead of typing

Voice modes let you speak naturally and hear a spoken reply, which suits thinking out loud, practising a conversation or getting help while your hands are busy. It is a different medium from text, not a smarter version of it.

Generating and editing images without losing the plot

Image tools can create a picture from a description or edit an existing one on request, which is useful for drafts, ideas and simple visuals. They are not a replacement for a professional illustrator or photographer when accuracy, likeness or rights matter.

Tools that fit: ChatGPT, Claude, Gemini · Reviewed 2026-07-19