We understand that a Digital Twin, with access to your calendar, email, and meeting history, requires an exceptional level of trust. That's why privacy and security are foundational to how Digital Twin operates.
Your Digital Twin will never share your information without your explicit confirmation. Every action it takes on your behalf is transparent, and you maintain full visibility and control. As part of our development process, we coined the concept of a “sidebar” to explain how the Digital Twin moves a conversation out of the public thread to one directly with you.
The sidebar is central to how the Digital Twin operates. Before sending any email that shares potentially sensitive information, the Digital Twin shows a draft and asks for approval. This creates a small amount of friction by design. The alternative—the Digital Twin sending messages without review—would be faster, but would break trust the first time something went wrong. The sidebar ensures users stay in control of what gets shared.
This unique process means that you get the benefits of intelligent automation without compromising the confidentiality that professional relationships require.
We've also designed the Digital Twin to operate within clear boundaries. It handles the tasks it's designed for and knows when to defer to human judgment. It will not take action without your clear consent. Your Digital Twin enhances your capabilities rather than overstepping them, with no risk to your data or enterprise infrastructure or intelligence.
Your Digital Twin acts automatically, but without complete autonomy, delivering the type of safety and security required of the largest enterprise organizations in the world, delivered free of charge to companies of any size. To protect against spoofing and unauthorized access, Ada verifies email identity using SPF, DKIM, and DMARC, and will only access data when you are an authenticated participant on the thread. Custom domains for certain customers that meet required thresholds provide additional security. A person can’t send an email pretending to be another person, and get Ada to respond.
Even so, in conversations, it will get things wrong. Our philosophy is to make mistakes visible, correctable, and instructive.
When you Digital Twin misunderstands a request or takes an unwanted action, users can pull it aside into a sidebar, just like pulling an assistant into your office for a private word. In that space, they can correct the error, provide guidance, and redirect. The Digital Twin incorporates the feedback immediately. The same pattern applies to factual errors. Users always review drafts before they go out, so mistakes can be caught and fixed before they reach anyone else, and it can only look at content from people who are already on the thread and have opted in.
More broadly, your Digital Twin is designed to be steerable. Users can change direction mid-task, override a decision, or roll back an action. If i created a calendar event that shouldn't exist, it can be deleted. If a meeting was scheduled for the wrong time, it can be moved.
Your Digital Twin doesn't lock users into outcomes, it stays responsive to course corrections throughout. It works with you to amplify your impact, combining automation with oversight. That’s the balance that makes trust possible.
Your Digital Twin acts automatically, but never without your approval.
%2C%20Color%3DOriginal.png)
