Interview with Remy Reya: AI Founders & AI Tech leaders playbook for nonprofits

Table of Contents

• Dek
• TL;DR
• The playbook
• Case snapshots
• Copy/paste checklist
• Open questions
• Resources & links
• Watch / listen (timecoded highlights)
• Footer CTA

Dek

What you’ll learn: how Compass Pro Bono used GPTs to save 10–12 hours per person per week, which AI patterns translate straight to small nonprofits, and a responsible, change‑management framework any AI founder or leader can copy this quarter.

TL;DR

• Nonprofits are a high-leverage audience for AI: big upside because teams are resource‑constrained, but adoption lags due to bandwidth and governance concerns. (Intro / 00:00 — see video: https://www.youtube.com/watch?v=SdBJxAEN48M&t=0)
• Start small: use LLMs for repeatable work (grant drafts, social posts, conference apps) before buying bespoke systems. Compass Pro Bono saved an average of ~4–5 hrs/week after initial rollout and ~10–12 hrs/week after 8 months. (see https://www.youtube.com/watch?v=SdBJxAEN48M&t=300)
• Change management > model choice: designate an AI champion, codify workflows, then augment with RAG/custom GPTs—this reduces lock‑in and accelerates trust. (see https://www.youtube.com/watch?v=SdBJxAEN48M&t=1500)
• Measure time saved and where it goes: guard against the “impact treadmill” by deciding up front whether saved hours buy strategic time, relationship building, or faster throughput. (see https://www.youtube.com/watch?v=SdBJxAEN48M&t=1660)

The playbook

Why it matters

Problem: nonprofits do high-impact work on shoestring budgets; time is the scarcest resource. AI can multiply capacity quickly, but most nonprofits lack the time and infrastructure to experiment safely—so the gap between potential and practice is wide.

Audience: this post is written for AI Founders & AI Tech leaders who are building tools for mission-driven organizations, and for nonprofit leaders deciding whether to adopt AI now or later.

What to do — a four-step framework you can apply this week

Remy Reya’s approach at Compass Pro Bono is pragmatic: don’t buy the shiny stack first—prepare the org, pilot the low-risk wins, iterate, then scale. Below is a checklist that mirrors that path.

1. Groundwork (week 0–4)
   • Audit repetitive tasks: list processes that take >2 hours/week and are templateable (grant asks, conference apps, program reports).
   • Pick an AI champion: give one staffer 30–60 minutes/week to experiment and report back.
   • Standardize basic infra: move files to Google Drive, centralize comms on Slack, and ensure a single source of truth for documents (Remy cited switching away from Dropbox/Teams as preparatory). (see https://www.youtube.com/watch?v=SdBJxAEN48M&t=1800)
2. Pilot (weeks 2–8)
   • Choose two micro pilots: one external (grant draft generator) and one internal (meeting notes/transcription automation).
   • Use off‑the‑shelf LLMs + RAG to train a small custom GPT on your best 10–20 grant applications (Remy’s team did this and reached ~70% of a usable draft). (see https://www.youtube.com/watch?v=SdBJxAEN48M&t=1260)
   • Require human review on all AI outputs; track edits to refine prompts and retrieval documents.
3. Iterate and measure (months 1–6)
   • Survey the pilot users monthly: time saved, where saved time is spent, comfort/confidence with outputs. Remy reports initial 4–5 hrs/week saved, rising to 10–12 hrs/week across the team. (see https://www.youtube.com/watch?v=SdBJxAEN48M&t=300)
   • Use the craft prompt template (context, role, action, format, target) to increase output quality.
   • Keep a human‑in‑the‑loop policy for sensitive clients and mission‑critical decisions.
4. Scale and govern (months 3–ongoing)
   • Lock down vendor discounts, shared licensing, and a policy for what data can be uploaded. Remy negotiated nonprofit discounts to lower the cost barrier. (see https://www.youtube.com/watch?v=SdBJxAEN48M&t=1200)
   • Create transparent norms: when AI is used, how humans sign off, and what falls outside automation.
   • Decide what saved time should buy—strategic planning, relationship work, or right‑sized workload to reduce burnout (avoid the “impact treadmill”). (see https://www.youtube.com/watch?v=SdBJxAEN48M&t=1660)

How others did it (case + quote)

Compass Pro Bono built custom GPTs trained on internal docs for grant writing, social media, and program design. Remy: “We rolled out our suite in Oct 2024…in January people reported saving 4–5 hours a week; by June it was 10–12 hours.” (video timecode: https://www.youtube.com/watch?v=SdBJxAEN48M&t=300)

Metrics that matter

• Hours saved per staff per week — initial target: 2–5 hrs; mature target: 8–12 hrs (Compass benchmarks: 4–5 → 10–12 hrs/week). (see https://www.youtube.com/watch?v=SdBJxAEN48M&t=300)
• Grant throughput — number of proposals completed / month; target +25–50% with templating + LLM drafts.
• Quality delta — percent of AI draft retained after human edit (aim for ≥60–70% to convince skeptics).
• Adoption rate — % of staff using AI tools weekly; aim for 60–80% within 3–6 months.

Pitfalls and how to avoid them

• Pitfall: starting with tech, not process. Fix: document repeatable workflows first, then automate.
• Pitfall: vendor lock‑in and data siloing. Fix: use platform‑neutral tools (Remy called Perplexity “Switzerland of AI” for model switching). (see https://www.youtube.com/watch?v=SdBJxAEN48M&t=1500)
• Pitfall: losing human touch in mission work (e.g., tutoring, counseling). Fix: keep humans in the loop where empathy and relationships matter.
• Pitfall: saved time absorbed back into more work (the impact treadmill). Fix: set an organizational rule on how to allocate saved hours (strategic projects, coaching, reduced hours). (see https://www.youtube.com/watch?v=SdBJxAEN48M&t=1660)

Case snapshots

• Grant writing GPT — trained on 15–20 successful applications; produces a 70%–usable first draft. (see https://www.youtube.com/watch?v=SdBJxAEN48M&t=1260)
• Meeting transcription + notes — Blurow + Whisper Flow to capture calls and summarize action items automatically. (see https://www.youtube.com/watch?v=SdBJxAEN48M&t=1200)
• Thought leadership assistant — a custom GPT that helps staff translate institutional knowledge into LinkedIn drafts and talk outlines, lowering the barrier to public presence. (see https://www.youtube.com/watch?v=SdBJxAEN48M&t=2400)

Copy/paste checklist

• Run a 1‑week audit: list top 10 repetitive tasks and time spent per week.
• Identify 1 AI champion and schedule 30-minute weekly experiments.
• Collect the 10–20 best examples of the output you want (e.g., winning grants) and upload them for RAG training.
• Deploy one custom GPT for a single use case (grant or social posts) and require human review.
• Survey staff month‑over‑month: hrs saved, what staff did with that time, confidence level.
• Create basic policies for data sharing and vendor discounts; negotiate nonprofit pricing.
• Plan a 3‑month roadmap: pilot → iterate → scale; set adoption KPI (e.g., 60% weekly use).
• Publish a public-facing note that your org uses AI responsibly to build trust with stakeholders.

Try this next week

1. Pick one repetitive deliverable (e.g., grant cover letter). Save 10 past examples to a folder.
2. Open ChatGPT or Perplexity and prompt: “Using these examples, draft a grant response for [program summary]. Keep it under 600 words.”
3. Schedule a 30‑minute review with the grant lead to edit the draft and document changes for the next iteration.

Open questions

• If your team saves 10 hours/week, how will you allocate that time—more output or less burnout?
• Which services in your mission‑delivery chain must remain human? Where is automation acceptable?
• What governance structure will you use to audit AI outputs for equity and bias?

Resources & links

• ChatGPT / OpenAI — good baseline LLM for fast prototyping and prompt iteration.
• Perplexity — platform‑neutral model toggling; useful to avoid lock‑in and test multiple backends. (Remy recommends as “Switzerland”.)
• Blurow — AI meeting note taker; useful for turning calls into action items automatically.
• Whisper Flow — live transcription for events and interviews; reduces manual note burden.
• Happenstance — network engagement tools (used by Compass Pro Bono to activate volunteer networks).
• Compass Pro Bono — a practical exemplar of AI rollout in nonprofits; case examples in the episode. (see https://www.youtube.com/watch?v=SdBJxAEN48M)

Watch / listen (timecoded highlights)

1. Intro & what Compass Pro Bono does — 00:00 (https://www.youtube.com/watch?v=SdBJxAEN48M&t=0)
2. Why AI matters for nonprofits — 01:16 (~76s) (https://www.youtube.com/watch?v=SdBJxAEN48M&t=76)
3. How Compass uses AI day‑to‑day (grant GPTs, social, program design) — 05:00 (300s) (https://www.youtube.com/watch?v=SdBJxAEN48M&t=300)
4. Ethics, jobs and the human element — 09:40 (580s) (https://www.youtube.com/watch?v=SdBJxAEN48M&t=580)
5. Grant writing and building internal GPTs — 21:00 (1260s) (https://www.youtube.com/watch?v=SdBJxAEN48M&t=1260)
6. Vendor lock‑in, Perplexity as platform-neutral option — 27:40 (1660s) (https://www.youtube.com/watch?v=SdBJxAEN48M&t=1660)
7. Time savings, burnout, and the “impact treadmill” — 33:00 (1980s) (https://www.youtube.com/watch?v=SdBJxAEN48M&t=1980)
8. Bias and homelessness systems risk — 45:00 (2700s) (https://www.youtube.com/watch?v=SdBJxAEN48M&t=2700)

Footer CTA

Get weekly, 5‑min, founder‑ready AI insights from Homebase → http://www.thehomebase.ai/newsletter

Credits

Guest: Remy Reya, Director of AI & Thought Leadership, Compass Pro Bono. Episode: “Nonprofits Are Saving 12 Hours a Week With AI” — full interview on The AI Chopping Block with David Stepania. Source material: episode transcript and timestamps from the original video: https://www.youtube.com/watch?v=SdBJxAEN48M.

This article was created from our video Nonprofits Are Saving 12 Hours a Week With AI with a little help from AI.