Overview

Just getting started?

If you're new to building AI agents, we recommend taking a crawl, walk, run approach to building your agent.

1. Start in the Crawl phase. Build out one clear use case, and launch your agent. Use the transcripts feature to see how users are interacting with your agent and make improvements to it.
   1. Example: Start with customer support. Answer the most common questions with the knowledge base, and use the API step or a function to create tickets that can't be answered. Spend your time optimizing this and answering more and more complex questions to increase your automation rate.
2. Expand to the Walk phase by layering on use cases. Once you have maxed out your single use case across the understand, decide, and respond categories. Add on new workflows and repeat.
   1. Example: Once you've got support tickets locked down, start adding on a sales function. This can be within the same agent or in another separate agent. You can also start to expand to multiple channels. Go from just the chat widget to email, slack, discord, or embedded more natively in your app/website.
3. Experiment in the Run phase. Start to push the boundaries of your AI agent by connecting to new models and new backend tools to see if you can optimize performance. Follow along with our developer experiments on YouTube and our GitHub repo to see what experiments we're learning from with our own AI agent — Tico!