om
What om does, which AI models it runs on, and why facts are not stored in a model.
om is AppThentic’s assistant. It reads your systems, asks owners, and answers anyone who asks, in chat or by phone. Its reasoning runs on leading AI models from enterprise providers, under zero data retention and terms that forbid training on your data. Your rules, owners and answers are stored only in Organization Memory, never in a model.
What om does
- Reads code and the conversations around it, and finds business rules.
- Decides which gaps are worth an owner’s attention.
- Writes short, specific questions, and understands the answers.
- Follows an answer to the person who actually decided.
- Answers people in chat and by phone, and calls the owners it needs to confirm an answer. See Calling om.
What om does not do
om does not store your facts. A model’s weights cannot cite a source, cannot be corrected one fact at a time, and can produce a confident sentence that nobody ever said. So every rule, answer and approval lives in Organization Memory, separately from om. When om needs a fact, it retrieves it and cites it.
| In om | In Organization Memory | |
|---|---|---|
| Finding the business rules in your code | ✓ | |
| Working out whom to ask about each rule, and what to ask | ✓ | |
| Your rules, reasons and approvals | ✓ | |
| Sources and history | ✓ | |
| Can be corrected or deleted one fact at a time | ✓ |
Which model om uses
The value is in Organization Memory, not in the model. om sends the model only the text of the request it is working on, and keeps everything it learns in Organization Memory. That means the model can be swapped without losing anything.
- On AppThentic cloud, om uses enterprise AI models under zero data retention: the provider keeps nothing and never trains on your data.
- On Enterprise, om can run on a model hosted inside your own network, so no request leaves it.
Where om runs
- In your infrastructure, with a self-hosted model. Nothing leaves your network. Choose this if your servers have no outbound internet access.
- Hosted by AppThentic. No hardware to run, and updates are applied for you. The requests om works on go to AppThentic’s servers and, under zero data retention, to the model provider.
See Deployment and Data handling.