Personal AI offers both a model and an app to help you train your AI model with your documents, files and memories. Unlike conventional machine learning techniques, you don’t need to know how to code, be an expert in neural networks or sort everything into data sets. You can simply upload your data and your Personal AI will start using it to generate replies to any question.

Personal AI offers high response accuracy, making it a great match for mission-critical workflows and high-risk industries. Its proprietary small language model (SLM), MODEL-2, is just flexible enough to represent your data faithfully. And if there’s not enough data to produce a good response, you can set it to blend the output with other leading LLMs from providers like OpenAI and Anthropic—but none of your data will actually be used to train any of these providers’ models.

As you train your Personal AI, it becomes increasingly accurate and offers more nuanced answers to complex questions. You can use it as a single-source-of-truth with perfect memory, a business consultant or an AI chief of staff, always accessible through dedicated chat windows.

And the same way you can chat with your Personal AI, you can also set up channels to invite others to chat. This is useful for automating repetitive conversations in customer service, keeping teams aligned on projects and engaging with your social media followers at scale.

Training your Personal AI

The process of training your AI is simple. Round up your text documents, audio/video, URLs, PDFs and even Excel files and upload them using the Training Studio. After a short period, your AI is ready to answer questions on the data your just uploaded.

To save time training, you can use the Memory Integration feature to sync your AI with Google Drive, Gmail, Microsoft OneDrive and Outlook. This way, as you upload files to the cloud or engage in new email threads, all relevant data is stored and is ready to access.

Your Memory Stack

Whenever you upload a document into Personal AI, it’s broken down to memory blocks and saved to the Memory Stack. These blocks are then used to train the SLM. In general, you should always aim to upload as much high-quality data into your AI as possible. As you add more memory blocks to your stack, your AI’s capabilities increase, providing better accuracy and more nuance when answering to complex questions.

Organize knowledge with AI Personas

Just as your organize the files in your computer into folders, you can train multiple AI Personas to specialize in different topics. Each Persona will focus on the data you train it on, never talking about other topics outside this scope. This is useful for separating executive-level sensitive data from customer information, for example. That way, when you share your Personas, no sensitive data will be leaked.

These Personas can become virtual employees as you train them on more and more information, offering anything from quick reports to decision-making assistance. You can customize their writing style or message structure using simple prompts, helping you automate and augment your work.

Chatting with your AI

You can ask questions to your AI within the Personal AI app, available on the web, desktop and mobile app. It’ll always reply based on the data you trained it on. You can use any prompts you use with other AI chatbots here—the main advantage is the results will be grounded in your data.

When chatting with other (human) users within the Personal AI app, your AI is ready to assist you with 2 modes:

• AI Copilot understands the messages you receive from others and provides a reply draft you can send with a single tap or click.
• AI Autopilot automatically replies with the best possible answer based on your memories.

There’s more than 1-to-1 chatting: when you create a channel, you can distribute your AI to more people at the same time. Simply select the Persona you’d like to use for the channel, activate your preferred AI mode and start adding other users. You can use the premade replies of Copilot to give quick answers to your team or Autopilot to keep in touch with your followers while you’re working on other projects, for example.

Personal score

When Personal AI generates an answer to a prompt, you’ll see a personal score indicator at the top. This shows how much personal data the model used to generate the answer. Higher scores indicate there’s a lot of data to base the response. Lower scores show that you may need to train your AI further.

Deploy your AI

With such a valuable model on your hands, it makes sense to integrate it with other apps. Personal AI integrates with Zapier, letting you connect and automate workflows across 6,000 apps. You can upload new memories into your AI and push out generated messages to any platform, such as a Slack channel, Salesforce record or an Intercom ticket.

‍