Mahesh Murthy

We just launched a new product: Karma Compass 🎉

It's the first piece of something bigger we are building: an agentic operating system for nonprofits.

We're starting with board management because every nonprofit has a board, it's time consuming and they don't get enough out of their boards, we hope to change that.

If you're on nonprofit staff or a board, I'd love for you to give it a try!

Also follow @compassbykarma for future updates.

compass.karmahq.org

As of this week, Karma Compass works as a connector in Claude, ChatGPT or any agent you use. Ask "What is still open?" and get your board's answer without opening another portal. Find more details and more features we shipped in the latest waypoints edition.

We launched Karma Compass a couple of weeks ago. Since then we’ve been onboarding nonprofits, talking to them every week, and fixing what gets in their way.

Waypoints is our weekly update on what we built from those conversations. Here’s the first one.

We just launched Karma Compass on Product Hunt 🚀

Our AI board assistant drafts agendas, generates meeting minutes, and follows up on tasks giving nonprofit leaders more time for their mission.

Would love your support! Take a look and upvote if you like what we’re building

producthunt.com/products/karma…

The new AI model Jev by @typesafeai is all over my X timeline with some really interesting usecases.

It got me thinking about nonprofit use cases of Jev. Say your CRM has 1,000, 10,000, or even a million donor records. Jev could score each record for something specific like “likely to respond to personal outreach.” Then it's easy to just rank those scores and give your fundraising team 10 people worth looking at today. At TypeSafe’s published price of $0.042 per million input tokens, this could be cheap enough to run every day instead of doing one big segmentation exercise every few months.

Another example: Someone emails asking about a program. Jev could score whether the message probably needs:
- a phone call
- a personal email
- an automated reply
- spam review

And it would use relevant context from your CRM when producing those probabilities. For ranking and routing work with clear rules and a human override, I think there are a lot of possibilities here. Still thinking this through.