Folks are still super careful with AI agents helping with sensitive tasks!

A lot of people I come across who are considering agents worry about the stochastic nature of LLMs. They don't want to trust them with complex tasks that require several steps because they might miss one or get it wrong. They worry that the agent will go off topic, access apps or generally do things they shouldn't. Even worse for regulated industries, you need to produce evidence of how you comply with different laws. Humans leave a natural audit trail of SOPs they follow, tasks they complete etc. so you can easily work with an auditor but what do you do when it's AI doing the work?

At Portia AI, we built an open source SDK from the ground up to address this challenge and we'd love to know if you think it meets the need and if there's something we can learn from others in this space. Agents built with Portia plan transparently, execute statefully, can be interrupted for human input and auth deterministically.

Let me know what you think if you have a moment? Your feedback would be super helpful 🙏🏼