Publish, Control, and Deploy your AI

March 22, 2024

After launching the training spaces, we shifted our focus to developing the secondary aspect of the user journey, which is separate from the AI training process. We recognize that our users value the two key components of personal AI—the input and output processes—comprising the training of their personal AI and the subsequent publishing process that integrates it into their existing workflows or communities. This new publishing journey introduces external messaging integrations, various user controls, and enables users to connect their personal AI agents, among other enhancements. Integrating this phase into the journey allows users to seamlessly train and share their personal AI in one cohesive experience.

Brand Page

In the Personal AI application, every user has the opportunity to build a digital brand around their personal AI and effortlessly share their profile via a domain. For instance, to see our co-founder's brand page, visit Previously, this page allowed users to share their AIs, yet all visitors were required to create their own account to interact with the AI. As we've intensified our focus on enhancing the relationship between AI creators and AI consumers, our features have evolved to support this dynamic. Specifically, for the brand page, we've introduced a public chatting feature where users can click "Chat Now" to instantly converse with an AI. Throughout the publishing journey, users can activate Brand Page chatting and manage AI modes (copilot, autopilot), personal score thresholds, and directly engage with visitors sending messages through the corresponding lounge.

Message Integrations

In the past, users were exclusively dependent on Zapier to set up integrations, enabling their AI to communicate with other platforms through the personal AI memory API. These setups, often time-consuming, required additional fees via Zapier, yet they proved beneficial, particularly for power users with complex workflows, offering them control over their AI. Given Zapier is a do-it-yourself option, we prioritized developing native external integrations to reduce time-to-value and support both new users and those inexperienced with Zapier. 

Common use cases were using a personal AI to reply to emails, Slack messages, phone texts (SMS), Messenger, and Instagram DMs. Our collaboration with Anon facilitated rapid implementation of these integrations, starting with Instagram, Messenger, and SMS. Now, users can connect to these integrations and manage each separate integration in an individual lounge.

AI Message Controls

With the consideration of sharing a personal AI publicly, many implications emerged regarding this interaction. Two primary requests we've consistently encountered are: (1) a pre-set AI first message or greeting and (2) a fallback message for when the AI lacks a sufficient response.

The fallback message ties into a definitive benchmark for all AI responses, known as a personal score, displayed as a percentage. This score mirrors the accuracy and authenticity of messages generated by your personal AI, based on the information available about you. Each message from your AI includes this percentage, offering full transparency into the extent your data is used in responses and how much is supplemented by general knowledge. This feature enables users to identify the gaps in their AI's memory. When gaps are discovered, users can simply add more memories. However, in the meantime, if a response to a question falls below the user-set personal score threshold (e.g., do not respond to responses below 70% personal), the user can set this fallback message to indicate that the AI does not know the answer. This also helps filter out inappropriate requests or questions. As a result, users receive only quality responses, while AI creators can add memories to fill these gaps and prevent future fallback messages on that topic.

Stay Connected
More Posts

You Might Also Like