Introducing our new AI Training Studio, designed to transform how you interact with, train, and customize your Personal AI. From the start of this design process, our goal hasn’t been to reinvent the wheel. Instead, we’ve focused on creating a training studio that combines the familiarity of traditional data-upload experiences (like Dropbox) with innovative AI features, driven by user feedback, based on functions that make the training process 10x faster and deliver better quality.
New Global Persona Feel and Feature Consolidation
Previously in the application, the ways users trained varied due to touchpoints and various places where we had uploading options. For uploads, users had to constantly specify the persona, and some features like integrations were hard to configure at the top level across personas, given the segmentation of data our users want (e.g., wanting to sync a personal Google Drive file to a Personal Life AI persona and a work folder to a Professional AI persona). This led to extra clicks and work during the training process, and on the back end, made tracking uploads and knowledge management more difficult since all uploads across all personas were being routed to one upload page and memory stack. Our users have communicated to us the value of each persona, with each filling key roles in their organization or use case, trained exclusively on specific data to boost efficiency, streamline workflows, and fuel collaborative intelligence that powers innovation. Therefore, to allow for this rich feature set that I will delve into in the remainder of the blog, and customization both in terms of training and configuration at the persona level, each persona now has its own AI training studio. All uploads, settings, integrations, and knowledge management functions tie to the persona that the user is working with, creating a new feel that a user is working within “a” persona at a time rather than their broader coalition of personas.
Streamlined Separation of Upload Methods and Upload Enhancements
Upon clicking “Memory Upload,” you'll see a secondary selection to switch between our supported upload methods: files, URLs, text, and media. This new interface gives more real estate to each method and simplifies the process. Unlike the previous system where users had to press upload, pick a method, then add content, users can now select one method and work within it. This is especially beneficial during initial training, as users tend to upload each type of content in batches.
We've made several improvements to streamline the process:
Multi-file drag and drop: Users can now upload multiple files simultaneously, significantly reducing training time. Just drag and drop :)
Automatic naming: We autopopulate upload names based on file names or URLs, eliminating the need for manual titling.
Persona-specific routing: Because of the new global persona feel, uploads are automatically siloed to the current persona, removing the need for manual selection.
Method-specific history: Each upload method has its own history, making it easier to track and find past uploads. You can still visit the Upload Library to see a collection of all uploads for the given persona.
These changes result in a much quicker interaction with fewer clicks, creating less friction and allowing users to seamlessly train.
Document Card Actions: Refine, Assign to Persona, Upload Status and More
We have enriched the functionality on the document card while reducing its size for easier navigation when looking for uploads.
After initiating an upload on any upload method, we now have a status bar to inform you how close the upload is to being completed. If the upload fails, you will be provided with a specific error state specifying the reason why.
Although we are routing uploads to the persona you are in by default, we are giving you control by allowing you to click "more," and then select additional personas to add the upload to.
You can now click "Clean" on an upload, which optimizes the memories you add, making them more compatible with our AI model based on the learnings from how our training team cleans uploads for our enterprise clients.
If a given upload has hashtags attached to it, simply press the arrow icon to expand the document card and view the hashtags.
Enhancements to Integrations
We have made memory integrations much more effective by giving you control to set up specific accounts and syncs for your individual personas rather than at the top level. Click on 'Memory Integrations' to view our Gmail, Google Drive, OneDrive, and Outlook integrations.
We've made several improvements to streamline the process:
Now that memory integrations are separate for each persona, you have the control to sync specific Google/OneDrive folders and Gmail/Outlook labels to individual personas. This allows you to create various syncs to different personas based on the desired use case and dataset.
You can connect multiple accounts, allowing you to access all of your desired emails or files across multiple accounts.
You can easily track and view individual emails or files that stem from a label, folder, or individual one-time upload in the upload history.
If you select a label for Gmail or Outlook, or a folder for OneDrive or Google Drive, you can click on the sync tab to track which folders or labels are being synced. Additionally, you can manually re-sync a folder or end a sync.
We are excited about how this new studio empowers the Personal AI community to train and manage their Personas more efficiently.
.svg)
