Tokens & Tactics #17: Making AI Work for Nonprofits
Perry Hewitt on her work at data.org helping organizations use AI for social impact, treating AI as a bureaucracy expert, and managing her cross-country move with ChatGPT as project manager.
Welcome back to Tokens & Tactics, our Tuesday series about how people are actually using AI at work.
Each week, we feature one person and their real-world workflow—what tools they use, what they’re building, and what’s working right now. No hype. No vague predictions. Just practical details from the front lines. This week: Perry Hewitt.
Tell us about yourself.
I’ve had a kind of funky career in different sectors with one common theme: using technology and language to enable the world’s best ideas to reach the right, engaged audiences. Currently Chief Strategy Officer at data.org, a nonprofit focused on data and AI for social impact. In this role, I get to hear a lot of great ideas and implementations of AI applied to real, practical problems, to scale them through funding, and to share them with the world.
Also: I love to run road races, with my enthusiasm undeterred by my lifelong mediocrity.
ChatGPT, Gemini, or Claude?
ChatGPT ends up being the go-to, from web search to deep research modes. I do double check with Gemini when I’m unsure of the results, and/or the stakes are extra high. To me, ChatGPT seems more aware of the real-time content out there.
What was your last SFW AI conversation?
Our team was trying to generate solutions for organizations in APAC seeking to make best use of data to improve the financial health of their communities. We had isolated solution strategies through desk research and lots of conversations with actual humans doing this work. But as Paul Ford likes to say, no one does bureaucracy like AI, so let them plot out the step-by-step instructions for getting there. Then, back to actual humans in the loop review and refine -- but AI does this process stuff really, really well.
Problem Statement
Across APAC, many MFIs, FSPs, and nonprofits have thin teams and tight budgets, so data sits in spreadsheets or siloed systems with little time or skill to analyze it. This creates “dark data” and slow decisions—especially harmful when serving low-income clients who need timely, data-informed support.Solution Strategy
Build capacity in layers over 12–24 months: (1) make analysis easier for non-technical staff with low/no-code tools; (2) bring in outside expertise while ensuring knowledge transfer; (3) train a core group of staff with practical programs. Anchor efforts to proven programs like data.org’s APAC Data Capacity Accelerator and global CAN hubs to access curricula, mentors, and peer cohorts.Approach 1: Start with simple analytics tools
Put one live dashboard in front of program leads within 90 days. Use Looker Studio (free, drag-and-drop) or Power BI for Nonprofits (discounted/grant options) for charts and automated refresh; add an open-source BI layer like Metabase when you need SQL-friendly exploration on top of your database. Begin with a single KPI set (e.g., active clients, arrears, on-time follow-ups), publish a monthly snapshot, and iterate.
First "aha!" moment with AI?
Frankly, when I saw the amazing work some of our data.org global innovation challenge awardees were doing with AI across the sector. When you operate programs and interact with partners in low and middle income contexts, you are reminded you don’t need enormous investment to reap the benefits.
Your AI subscriptions and rough monthly spend?
No obscure APIs -- am pretty basic, frankly
$20/mo ChatGPT
$20/mo for Google Gemini
$30/mo for Every
Who do you read/listen to to stay current on AI?
It’s a combo of the filtered hype machine served to me daily by TechMeme, of responsible data and AI resources curated by smart people like Stefaan Verhulst, and of seeing what our partners are actually using that’s getting results.
Your most-used GPT/Project/Gem?
Real talk? Real estate. I am plotting a move from one state to another in 2026, and tracking all the related complexity like sellers and buyers brokers, market analyses, and environmental considerations. See above re: bureaucratic excellence -- ChatGPT has me on timelines with dependencies like a personal project manager.
The AI task that would've seemed like magic two years ago but now feels routine?
It blows my mind that I can ask open-ended questions in incomplete sentences and expect to receive well-articulated answers back-- and often do.
Magic wand feature request?
A “don’t blow smoke” feature. I feel like the models get a sense of where you are leaning and tilt vehemently in that direction. For now, I rely a lot on “give me an argument against X.” But a “don’t blow smoke” button would be nice -- in AI and in the world, for that matter.
If you could only invest in one company to ride the AI wave, who would it be?
Anthropic, because they seem most able to give that leg up to software development.
But please know that my investment choices have included Peloton in 2021.
Have you tried full self-driving yet?
Waymo was way cool. As the worst driver in the world, I am ready for a driverless future. Also: I still thank ATMs so there is no way in hell I’ll be able to stop thanking the nonexistent drivers.
Latest AI rabbit hole?
Does baffled Sora scrolling count? If not, an exploration of Dinky toys and their creator, Frank Hornby. I have my late husband’s collection and am a little obsessed.
One piece of advice for folks wanting to get deeper into AI?
It might seem like AI is your friend and colleague performing magic for you each and every day, but AI can make mistakes. Dive in, but always double check the answers -- it’s fallible like those actual humans you interact with, but seems way more convincing.
Who do you want to read a Tokens & Tactics interview from?
Natalia Quintero from Every. Her husband may be way into Magic the Gathering, but she’s the biggest nerd I know.
If you have any questions, please be in touch.
Thanks for reading,
Noah and Claire