The webinar covered Tess AI's robust multi-agent and automation feature, highlighting its potential to help companies with digital transformation using generative AI.
An agent in Tess AI is similar to a GPT, but more robust, letting you combine different AI models. It can be simple or complex, using lots of tools and steps if needed.
An example of a simple agent takes an audio file as input, transcribes it using Deepgram (which is considered better than OpenAI's Whisper for Portuguese), and formats the output as JSON. This agent shows off Tess AI's ability to go beyond a traditional GPT by including specific tools for different jobs.
Tess AI offers an environment with multiple AIs, letting you create more complex agents. You can add steps like browsing the internet, processing documents, connecting with apps (Google Sheets, Google Drive), and using different AI models for transcription (Assembly AI, REV AI, Whisper).
The success of generative AI depends on the quality of the training material. Tess AI lets you use spreadsheets, documents (PDFs), web links, and other data as context for agents, which really boosts their results. You should focus on building a solid knowledge base!
A WhatsApp chat agent can be trained to work like a salesperson, qualifying leads. The key to success is giving good sample conversations and materials about sales methods as context.
To build a strong knowledge base, it's crucial to document WhatsApp conversations. A simple automation with Zapier/Make can be set up to transcribe WhatsApp audios and feed a Google Docs with the chats, creating valuable training material.
This agent transcribes audio files sent to a specific folder in Google Drive and feeds a Google Doc with the transcriptions, making it easy to document audio and classes.
An example WhatsApp chat agent shows how to create automation in Make to reply to received messages. The agent is set up in Tess AI and runs via API through Make. The conversation gets stored in a database and a spreadsheet for analysis and improvements.
A more complex automation handles .zip files exported from WhatsApp, pulls out audio, transcribes them, and puts all the conversations together in a .txt file for agent training.
Tess AI has an API, so you can connect it to a bunch of platforms like Zapier, Make, n8n, and your own systems too. The platform focuses on training and improving the AI, while the integrations take care of the side tasks.
Using the API uses up credits, so it's important to think about how efficient your automations are to keep costs under control. For big data volumes, Tess AI Orchestrator (still in development) brings more reliability and better value.
* You can transcribe long audios through the API (more than one hour).
* Tess AI works with n8n hosted on a VPS
* You can combine multiple agents using the "AI Assistant" feature
* To integrate with Typebot or Menchat, use the API and webhooks.
* Connecting with Apple apps can be done indirectly via Zapier/Make.
* To create a copywriter agent, use PDFs and copywriting history as context.
