I run a two-person research nonprofit with an AI agent as my operations team

I'm Amartya Kumar. Three weeks ago I started the Oryx Centre for Strategic Futures — a non-profit research centre working on AI governance and strategic foresight, run by exactly two people, bootstrapped at $0, from Noida, India.

A two-person institute has an obvious problem: the work of an institution — applications, deadlines, correspondence, research, a website, records — does not shrink to fit two people. So I gave the operations layer to my Muse agent, and I supervise it the way you'd supervise a very fast, very literal chief of staff.

What it actually does, every day:

• Drafts and sends our institutional outreach — funding applications, partnership proposals, speaker pitches — each one going out only after my approval
• Watches my inbox hourly and reports only what's substantive: decisions, offers, document requests
• Runs deep research scans — most recently, a country-by-country map of where a nonprofit like ours can register for free (60+ jurisdictions checked; France and South Africa are now in motion)
• Maintains the unglamorous stack: our website, business email, and local AI model routers for evaluation work
• Remembers every thread, deadline and decision, so nothing depends on my memory at 2 a.m.

What that's produced in the first weeks: our founding working paper published open-access with a DOI (10.5281/zenodo.23601643), funding applications in with ARIA (UK) and the Survival and Flourishing Fund, speaker applications at GITEX and other summits, and more than 180 institutions contacted — with every reply tracked.

The honest limits: I approve every external send, it gets things wrong that I have to catch, and supervising it is real work. It's leverage, not magic. But for a team of two trying to do the work of twenty, leverage is the whole game.

If you want to try Muse yourself, here's my invite code:

6E1TU7

New users can redeem it in the Muse app under Settings > Redeem token (within your first 48 hours after joining), or on muse.ai under Settings > General > Usage > Redeem invite code. We both get benefits when you redeem it.

If you're running something small and ambitious — a lab, a nonprofit, a one-person company — I'd genuinely like to hear how you're using agents. Where do they help, and where do they fall down?