BAI · A guide to placing tools where they belong
Do more than learn how to use a new tool. Learn what it serves, what it takes and where it should stop.Learn the tool—and what it serves, takes and should never replace.
Tools for conversation, research, code, agents, images, video, audio, games and translation keep changing. Instead of teaching people to memorise brands, BAI offers a practical way to assess each new tool through human judgement, sources, data, age, privacy, error and responsibility for the outcome.
The challenge
New tools can change faster than we can think through why and how we use them.
A tool may give impressive answers, write code, create images or sound and automate work. That does not tell us what it knows, how it uses data, whether it is suitable for a child or who remains accountable for the result. A person should not be reduced to an operator who only enters commands, and a tool’s output should not become a judgement no one questions.
Knowing how to use a tool is a skill. Placing it well is a greater responsibility.
A basic distinction
A tool carries capability; a person carries amanah.
BA Internet Tools places each function with the person or system that should carry it.
The tool
Assists
- Searches and organises
- Finds patterns
- Creates drafts
- Makes a workflow easier
The person
Carries judgement and responsibility
- Sets the intention
- Checks the source
- Protects context and limits
- Makes release and representation decisions
Before using a tool
Ask the same six questions, whatever the brand.
These are not one-time approval questions. Revisit them whenever the tool, version, purpose or people affected change.
01 What real need does it serve?
Ask whether a task that does not need a machine is being handed over unnecessarily—and whose life the tool will genuinely make easier.
02 Which decision must remain human?
Make clear where the tool only drafts, recommends or assists with a calculation, and where a person accepts, rejects, releases and carries responsibility.
03 Which sources and data does it use?
Find out whether sources are visible, which data is uploaded, who owns it, what permission allows its use and how long it is retained.
04 Who may use it, and who needs protection?
The same tool cannot always be offered in the same way across ages, languages, accessibility needs, child and household privacy or professional competence.
05 What does it not know, and how does it fail?
Test confident errors, missing source context, discrimination, synthetic output and misuse with real examples.
06 How can it be stopped, corrected and left behind?
Plan human intervention, error reporting, rollback, account and data deletion, a move to an alternative and responsibility for ongoing care.
Tool families
Names and brands change; the kinds of work we need continue to evolve.
This is not a complete catalogue. One tool may belong to several families, and each use should be assessed again through its own purpose and risk.
Conversation, research and learning
Asking questions, finding and comparing sources, explaining, taking notes and preparing exercises.
Code, agents and automation
Drafting code, investigating errors, designing task flows, connecting tools and building automation with human approval.
Images, video, audio and media
Creating draft visuals, animation, video, voice, music and stories suited to different ages.
Games, simulation and interaction
Developing games, 3D scenes, interactive stories, learning experiments and field simulations.
Language, translation and accessibility
Supporting multilingual meetings, subtitles, speech-to-text, plain language and access for different needs.
Data, privacy and assessment
Data minimisation, local processing, permission and retention controls, together with reviews of sources, risk, error and human approval.
SEDD: while choosing, using and leaving
Safety and limits are not a stamp applied after the work is finished.
Human approval
Consequential decisions, actions outside the system and release require explicit human approval. Automation does not remove responsibility.
Data limits
Do not collect unnecessary data; protect children’s and household data separately.
Visible sources
Keep the sources behind an output, synthetic material, uncertainty and human changes visible.
A clear representation statement
Do not present the tool as if it speaks for a person, institution or body of work.
Rollback and exit
A wrong action can be stopped and reversed; an account, dataset, integration or automation can be closed safely.
When trying a tool for the first time
Begin with a small, reversible trial tied to a real need.
Not every tool needs to enter Labs, Build or WAQF. Some remain useful for personal learning, some belong in a limited household, school or professional setting, and some should not be used at all.
A shared resource in preparation
The BAI Tool Radar will follow tools through dated, correctable review records—not admiration or fear.
Each record will show what a tool does, the age group and setting reviewed, account and permission requirements, data use, privacy risks, human approval, open or local alternatives, the latest review date and example tasks. Because tools change quickly, no record will be treated as a final verdict. Guild practitioners, young people, households and the BAI Tool Review Circle will be able to correct it over time.
Staying current is not a one-time catalogue; it is a shared responsibility to observe, record and correct.
Share an interest or a tool to review
Would you like to examine a tool together or turn what you learn into a responsible digital contribution?
The BAI Tool Radar and regular review Circles are still in preparation. You can share a tool you use, a benefit or risk you have encountered, and the age group or setting you want to examine. This is not an active programme or product request.