Responsible AI
How trōfik uses AI
This notice explains when people interact with AI, how AI assists project work, and the limits users should understand.
Key points
- Interactive AI features identify the experience as AI-generated where that may not otherwise be clear.
- AI assists people; qualified users remain responsible for decisions and important outputs.
- TAP is not intended to make autonomous decisions about a person's legal rights, employment, credit, housing, benefits, or access to essential services.
Where AI may assist
- Organizing, summarizing, searching, and drafting from customer-selected project material.
- Transcribing supported audio submitted through an enabled workflow.
- Supporting cited research from public sources.
- Running practice conversations with a synthetic, project-grounded composite rather than a real or cloned person.
Information AI may process
The information depends on the feature a customer enables and the material an authorized user selects. It may include project notes, meeting records, survey answers, uploaded documents, stakeholder details, submitted recordings or images, practice-session audio and transcripts, public-source material, and earlier project analysis.
Customers and users should avoid submitting personal or sensitive information that is not necessary for the requested task. Additional provider and data-flow information is available to customers through the applicable agreement and diligence process.
Human review
- Field-capture extraction creates information that an authorized user can review and correct.
- Governed workflows may require confirmation, approval, or a publication step.
- Interactive features may respond immediately, but their outputs are assistance rather than verified facts.
- High-impact or public uses require a qualified person to check the evidence, context, and limitations before action.
Composite stakeholder types
TAP may create evidence-grounded composite stakeholder types to help teams practice and plan. A composite is a planning tool, not a real person, a secret psychological profile, or a statistically representative poll. Platform instructions tell the AI not to invent missing facts or fabricate personal details.
Limits and corrections
AI can misunderstand context, miss information, or produce inaccurate material. Users should examine citations and original evidence where available. AI-generated or altered material intended for publication must be reviewed and labelled where applicable law requires it.
If you believe personal information about you is inaccurate or should be removed, email christian@trofikimpact.com. We may coordinate with the customer that controls the relevant project. See the Privacy Notice for more information.
Questions
Trofik Impact LLC
Brooklyn, NY 11201
christian@trofikimpact.com