Responsible AI Manifesto

Every session and every message with an AI Fellow is labeled as AI. A Fellow can have a name, a face and a voice so the conversation feels natural, and it never claims to be a person.

This holds wherever a Fellow appears: a live voice session, a text chat, a Learning Path session, or a public Fellow that someone opens from a link without an account.

If you are talking to an AI Fellow, you know it.

People in your organization decide what a Fellow does. An administrator writes its purpose, chooses the knowledge files it can consult and turns each capability on or off: screen sharing, camera, text chat, on-screen tools and Connectors to your own systems. Any of them can be switched off later, and screen sharing can be disabled for the whole organization.

Administrators can also see what happened. Within their own tenant they can read conversation transcripts, open a Fellow's configuration and its full edit history, and see which Knowledge Base it had access to. When the platform records which knowledge a Fellow retrieved while answering, that record is available for review. Not every generated sentence can be traced back to one source, and we say so.

What a Fellow was told to do, how that changed over time and what it actually said are all on record for the people responsible for it.

Frontierz is a multi-tenant service. Every request is scoped to your organization's tenant, and every action that touches a resource re-checks that the resource belongs to that tenant before it runs. Administrators of one organization cannot reach another organization's users, Fellows, conversations or analytics.

Customer data is never used to train, fine-tune or improve any AI model, ours or the provider's. Models are used for inference only. The platform and its AI inference run on Microsoft Azure in the European Union, and the Data Location and Export page names the regions. Monitoring providers that process technical telemetry are listed in the Privacy Policy and work under the data processing agreement.

Organizations that need more control can bring their own Azure AI Foundry resource. Inference then runs in an Azure subscription the customer owns and administers, subject to commercial agreement and technical validation.

Data is encrypted in transit and at rest. Authorized administrators can export their tenant's learning records, and can permanently delete the readable contents of a conversation while the activity record stays for reporting.

We use AI for the part of learning that is hard to get from a busy colleague: repetition, a patient listener and a safe place to get it wrong. A sales representative can rehearse a negotiation several times before the real meeting. A new hire can ask a Fellow the basic questions about company processes, then bring the harder ones to a teammate. A manager can practice a feedback conversation and try a different approach the second time.

The real conversation still happens between people, and the person keeps the judgment: what to say to the client, how to run the team, whether the recommendation stands. Practice feedback and scores are coaching input. Employment, certification and compliance decisions need a person to review the record, and Frontierz results are never meant to be the sole basis for them.

A Fellow prepares people for the moments that matter. The moments themselves stay human.

Frontierz keeps a small data footprint. For a learner it holds a name, a work email and learning records: progress, transcripts and the assessments a Fellow writes. The service processes no special categories of personal data and no financial or payment data.

Voice sessions use your microphone. Camera and screen sharing stay off until your organization turns them on for a Fellow or Learning Path, and even then nothing is shared until you choose to share it. You can stop at any time. The camera sends the Fellow snapshots at moments in the conversation, never a continuous video stream. Snapshots and screen frames are used during the live session only and are never stored. What you say or show can still appear in the transcript, like any other part of the conversation.

We do not sell personal data. Customer data is processed on the organization's instructions under a data processing agreement. You can exercise your data rights through the organization that runs your program or by writing to rgpd@founderz.com; the Privacy Policy explains how.