Skip to main content
The Human Bit
Ask Bit

GuidesChoosing a modelAnyone handling personal, medical, financial or confidential information

Matching your choice of tool to how sensitive the data is

For Anyone handling personal, medical, financial or confidential information.

Not all data belongs in every tool. Thinking about sensitivity before you paste something in is a habit worth building, not an afterthought.

When to use this

You are about to paste something into an AI tool, a client record, a medical note, financial detail or an internal document, and you have not stopped to think about whether it should go there.

  • Can work with anonymised or redacted versions of sensitive material just as well for many tasks
  • Helps you draft a redaction pass, stripping names and identifiers before the real analysis
  • Speeds up work on the non sensitive parts of a task so you can focus care where it counts

Workflow

  1. Pause before pasting and ask whether this data identifies a real person or is confidential

  2. If it does, remove or replace names, dates of birth, account numbers and other identifiers first

  3. Check whether your organisation has an approved tool or setting for sensitive work

  4. Use the anonymised version for drafting, analysis or summarising

  5. Re-add the real details yourself afterward, outside the tool, if the task needs them

Prompts to try

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

I am going to share a version of this with identifying details removed. Please [summarise, analyse, or draft] based on this redacted version, and do not try to guess or fill in who the real people are.
Ask Bit to adapt it →

Human lens

  • You decide what counts as sensitive in your context, since rules differ by field and by employer
  • You choose to remove or replace identifying details before sharing anything with a general tool
  • You take responsibility for checking your organisation's policy rather than assuming a tool's terms cover it

What to avoid

  • Assuming a tool is private without checking its actual data handling terms is a common mistake
  • Redacting badly, leaving an unusual name or a rare detail in, can still identify someone
  • Some fields, health, legal, government, have rules about AI tools that go beyond general good practice

Related guides

View all →

When to spend more thinking, and when not to

Most tools now let you choose a faster answer or a slower, more careful one. Matching the effort to the task saves time and avoids polished but wrong answers.

Feeding in a whole document versus asking a short question

Some tools are built to hold a large document in view and reason across all of it. Others are better suited to a short, focused question. Picking the wrong one either wastes time or loses important detail.

When you need a tool that can actually see or hear

Text only tools cannot read a photo, listen to a voice note or watch a clip. Choosing a tool that handles the kind of file you actually have saves a step and avoids a flat refusal.

Using a light tool for volume, and a stronger one for the hard cases

Not every task needs the strongest available model. Saving the most capable option for the tasks that truly need it keeps things fast and sustainable for everything else.

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