Ask AI

Custom Instructions

Make Sapience work your way.

Customizing Sapience for You

Custom Instructions

Telling Sapience how you want all Agents in the system to behave is a quick way to tune the behaviour of every Agent in every Conversation that you have with the system. Have a look at this screenshot, and read more below:

Notion image

Things to know:

Agent Behavior:

  1. Whatever you put into the Custom Instructions field will be silently injected into every Conversation you have with Agents in the Sapience system.
  1. That is powerful, in that it helps orient every conversation to the type of result that you are most interested in. But it comes with risks. If you put in the wrong thing here, it will impact every conversation.
  1. In the screenshot above, you can see that we’ve mainly told the system how we like outputs to be formatted. That’s a great approach. We’ve told it that we like details and numbered lists. All Agents interactions will keep this in mind.

About Me:

  1. Use this to tell Sapience facts about you. This will also be injected into every Conversation, much like the behavioral cues above.
  1. You can happily put 200-300 words of content in here, and you will see immediate improvements in the outputs you get.
  1. You can change this often. For example, if you’re working on a couple of big projects right now, put details in about those and what you’re focussed on. Just remember to review this info and change it as appropriate over time.

Memory

This is part of the overall Memory system in Sapience, and you can think of these as human-coded Memories that you have explicit control over. If you want to larn more about Sapience’s Memory architecture generally, read this article:

Project Agent Customization

Note that if you want to store information to customize Agent behavior that is particular to one stream of work, what you might want to do for that is use the configuration and customization options on a Project Agent, if you have a Project created for that stream of work. This lets you set Agent behavior and knowledge settings that only apply to a single Project, rather than globally. Learn more about here:

Did this answer your question?
😞
😐
🤩