Guide/Artificial intelligence
Vera, the AI assistant
Vera is the EterShield AI assistant (AI vCISO). It is the only conversational AI assistant on the platform: the former GRC Assistant has been retired, because Vera does everything it did and can see your organization's data. Anyone who opens the old GRC Assistant address is taken to the dashboard, with the Vera panel open.
Opening Vera
- Consultant view: floating button in the corner of the screen, or the ⌘J shortcut (Ctrl+J).
- Manager view: the Vera icon in the icon column (the floating button is hidden).
- The panel can be widened or narrowed. It shows the context Vera is looking at and the status Ready or Analyzing.
What Vera does
Vera answers using your organization's data, within the scope of your role and the active entity. Suggestions when you open it: summarize the most critical risks, check the status of assessments, create a task, generate a policy. In practice:
| Request | What happens |
|---|---|
| Look up risks and assessments | Summarizes the risks and the status of assessments |
| Search the knowledge base | Searches the organization's documents and approved information (only the sources your role and the visible modules allow) |
| Create or change a risk | Shows a risk proposal (or a proposed change to the risk): it is not saved yet. Confirm to save it; it is marked as AI-proposed and pending review, and enters the review queue. If you reject it, nothing is saved |
| Create a task | Shows a task proposal, with the same confirmation |
| Generate a policy | Drafts the policy text in the conversation; nothing is saved (to save it, use Policies, New Policy) |
| Register an AI agent | Opens a draft that only becomes a record after a person confirms it (see the AI Agents section) |
Each step is shown as a short summary ("Looked up the risks", "Prepared a risk", "Searched the knowledge base").
Limits and safeguards
- Vera sends the last 20 messages of the conversation to the AI model, with a maximum of 4,000 characters per message. If a message is cut, the answer ends with a note saying so.
- There is a limit on requests per minute, and each request spends your organization's AI credits (Limit reached appears when it runs out). With AI turned off, Vera is unavailable (see Settings, Data and AI).
- Suspicious content (attempts to manipulate the assistant) is detected and blocked, and personal data is masked before it goes to the model.
- Vera can make mistakes: always check what it proposes before saving. Possible error messages: No permission (your profile does not allow the request), No connection, and "Vera took too long to answer. Please try again in a moment."