Assistant

Assistant Behavior

LAST UPDATED: 27 MAY 2026

01. What the assistant is

The Triopia assistant is software that helps you prepare communication, summarize threads, draft replies, transcribe audio, and translate content inside your workspace. It is not a person.

02. The assistant identifies itself

When asked, the assistant will say that it is an automated assistant. It will not claim to be a human.

03. You stay in control

The assistant does not send messages, schedule meetings, sign documents, or take other external actions on your behalf without your explicit approval inside the workspace.

04. Working on your data

The assistant works on the content already in your workspace, such as messages, contacts, calendar entries, files, and notes. It does not have access to outside services unless you have connected them.

05. Source-grounded answers

When the assistant summarizes or briefs you on a thread, it draws from the workspace content you have. If the information is not in your workspace, the assistant should say so rather than guess.

06. Drafts, not commitments

Output from the assistant is a draft for you to review. Reviewing, editing, and approving the draft before it goes anywhere is part of the design, not a limitation.

07. Mistakes happen

AI models can make mistakes. Briefings can omit details. Drafts can misstate facts. Transcripts can mishear words. Always review assistant output before relying on it.

08. Not professional advice

The assistant does not provide legal, financial, medical, tax, or other professional advice. Consult a qualified professional before acting on information generated by the assistant.

09. Tone and style

The assistant aims for a clear, direct, professional tone. It should be useful and easy to read, not theatrical or salesy.

10. Refusal of harmful requests

The assistant will refuse requests that involve illegal activity, harassment, fraud, malware, or other harmful tasks.

11. What we do not do

We do not train AI models on your conversations. Requests to model providers are sent for inference only so they can generate a response in the moment.

12. Tier-based capabilities

Some assistant features are available only on specific subscription tiers. The capabilities included in each tier are listed on the pricing page.

13. Improvements over time

We will keep improving the assistant as model providers ship new capabilities and as customers tell us where it falls short. Material changes that affect what the assistant will or will not do will be communicated.

14. Telling us when it is wrong

If the assistant produces output that is unsafe, incorrect, or off the mark, report it through the support channel. We use the feedback to improve the prompts and safeguards inside Triopia.