Tessera
Guide
Keeping work and personal context separate in AI
Convenience has an edge case.
The better an AI knows you, the more it draws on. That is the feature. It is also why the wrong context sometimes turns up in the wrong answer.
What goes wrong
Nothing dramatic. A work email drafted in the register you use with friends. A meal plan that quietly assumes the schedule from a client project. A suggestion that is perfectly sensible for one part of your life and slightly off for the part you were actually asking about.
It is a small failure each time, and it erodes trust faster than a clearly wrong answer does, because you have to check everything afterwards.
The instinct that does not work
The usual response is to keep two accounts, or to use one AI app for work and another for home. It does work, and it costs you the thing you wanted: neither one knows enough, and you end up explaining yourself twice as often rather than half as often.
Think in scope, not in walls
The better frame is not separation but relevance. Everything about you can live in one place, as long as what gets used is decided by what you asked, rather than by everything that happens to be known.
That is how a memory layer should behave: when you chat, only the pieces relevant to your question are shared with the AI model answering you. The rest is not withheld dramatically; it is simply not relevant to a question about dinner.
What to actually do. Keep the context in one readable place so you can see what is there. Label the things that belong to a specific part of your life clearly enough that relevance can do its job. Then check the memory occasionally, which you can only do if it is something you are allowed to read.
The part you should still keep out
Scope is not a substitute for judgment. Anything under an employer's confidentiality obligation, or anyone else's personal information, does not belong in a personal memory regardless of how well the relevance works. That is a rule about who the information belongs to, not about how the software behaves.
For everything that is genuinely yours, one memory you can read and correct beats two accounts that each know half of you.
Questions
Related questions
Should I use separate AI accounts for work and personal life?
It works, and it costs you the benefit: each account knows half of you, so you explain yourself more often rather than less. Scoping what gets used per question is usually a better answer than splitting the memory in two.
How do I stop AI mixing up work and personal context?
Keep the context somewhere you can read, so you can see what is there, and use a system that supplies only what is relevant to the question rather than everything it knows.
What should I never put in an AI memory?
Anything covered by an employer's confidentiality obligation, and other people's personal information. That is a question of whose information it is, and no amount of good scoping changes it.
Stop explaining yourself to AI.
Ten days free, no card required. iPhone and web.