Christopher Nolan has a new movie about a wooden horse packed with soldiers (aka The Odyssey), and somehow that turned into the best AI metaphor I've heard all year. Asked whether AI might be a gift that quietly turns into something darker, the same way the horse did, Nolan laughed and said, “I think AI is a Trojan horse that everybody knows the Greeks are inside.” Nobody's confused about what's hiding inside this one. We already know exactly what it does: it takes real jobs, real judgment, and real relationships, and offers to do them faster, cheaper, and without a person in the room.
That's not an argument against opening the gates. It's an argument for deciding, on purpose, what you actually want walking through them. Nolan calls himself a techno-skeptic rather than a technophobe, someone who welcomes new tools without assuming they should replace whatever came before just because they're newer. That's close to the whole philosophy behind this newsletter: AI earns a seat at your organization's table when it protects the human relationships that are the actual reason people give, volunteer, and show up, not when it quietly starts standing in for them.
Every story in this issue is really the same question wearing a different outfit. A CRM that tells your fundraisers who to call. A model that can run a project for hours without a person checking in. None of it is dangerous by itself. It only gets dangerous if you let it in without deciding what it's allowed to touch and what stays yours. AI's trojan horse isn't hiding anything. The only real question left is whether we're paying attention to what walks out of it.
With much heart,
Amanda
AI News that Matters
Bloomerang brings predictive donor intelligence into its platform via Dataro
Bloomerang, the donor CRM used by more than 24,000 nonprofits, has rolled out an integration with Dataro that flags which donors are at risk of lapsing, which are ready for a bigger ask, and what to do about it, directly inside the platform your team already logs into. It went live this month and doesn't require exporting data or juggling a separate login.
Why it matters
If you're on Bloomerang, this is worth a look. If you're not, it's a signal of where donor CRMs are headed generally: from a place you store data to a tool that tells you what to do with it. Worth asking your own CRM vendor whether something similar is on their roadmap, and deciding now how much judgment you want a tool making on your behalf versus how much you want it just surfacing information for a person to decide on.
Read more
New research: most nonprofits use AI, few use it well
A new Blackbaud Institute report found that 85% of nonprofit staff use AI at work, but only 33% say their organization uses it effectively, and just one in ten nonprofits qualify as "AI-adaptive" under the report's maturity framework. The gap comes down to four things: unclear return on investment, inconsistent staff use, weak data health, and little to no governance.
Why it matters
Individual staff experimenting with AI doesn't automatically translate into organizational benefit. Without a policy, reasonably clean data, and shared expectations across the team, most of that time savings stays with the one person using it and never reaches the mission. If your organization doesn't have a written AI policy yet, this is as good a nudge as any to start one.
Read more
Claude Opus 5 launches, built for long, unsupervised work
Anthropic released Opus 5 this month at roughly half the price of its top-tier Fable 5 model while matching much of its capability. It's built to stay on a task for hours at a stretch, correcting its own errors and working around blockers instead of stalling, and it checks its own output along the way rather than needing someone to catch mistakes after the fact. It's also more likely to ask a clarifying question or flag a problem with your instructions before it just guesses.
Why it matters
This is the tier built for handing off real work, not just quick drafts. If your team has mostly used AI for single tasks like an email or a first pass at copy, Opus 5 is worth testing on something bigger: a multi-step research project, cleaning up years of messy program data, or pulling together a full grant packet from materials you already have.
Claude Cowork is no longer tied to one laptop
Starting August 3, Claude's Cowork feature, which works through a task across your files, calendar, email, and other connected tools until it's done, can run in Anthropic's cloud instead of needing to stay open on a single device. Your team can hand off a task, close the laptop, and pick the same session back up later from a phone, and scheduled tasks can keep running with no device turned on.
Why it matters
For small teams without a dedicated "AI person" glued to a desk, this closes a real gap. Something like a weekly donor newsletter draft or a board report pull can run in the background and get checked on whenever, from wherever.
Claude can now take action inside Microsoft 365, not just read it
An update to the Microsoft 365 connector lets Claude send email and manage drafts, labels, and auto-replies in Outlook, create or edit calendar events, and create or update files in SharePoint, on top of the read-and-search it could already do.
Why it matters
If your organization runs on Microsoft 365, this is a real jump from "AI that can look things up" to "AI that can act on your behalf." It's a good moment to revisit who on your team should have that permission turned on, and whether your AI policy already covers what Claude should and shouldn't be allowed to send under someone's name.
ChatGPT Work launches with a rebuilt desktop app
OpenAI's new ChatGPT Work takes on a project, gathers information across your connected apps, and hands back a finished spreadsheet, slide deck, document, or simple web app instead of a chat reply. It launched alongside a rebuilt desktop app for Mac and Windows that folds chat, Work, and OpenAI's coding tool Codex into one app, available on every plan including free.
Why it matters
This is OpenAI's answer to Claude Cowork and Microsoft's Copilot Cowork, so "can it take on a whole task and finish it" is now a fair question to ask of any of the major AI vendors. Still worth a small, real pilot with one defined task and a specific person reviewing the output before trusting it with anything involving donor or member data.
NotebookLM has a new name: Gemini Notebook
Google renamed NotebookLM to Gemini Notebook in mid-July, folding one of its most popular AI tools further into the Gemini brand. Nothing changes for existing users: same site, same notebooks, same source-grounded answers, just a new name and logo. The one real addition is a secure sandbox that lets it write and run code against your own uploaded documents for deeper analysis, rolling out first to paid tiers.
Why it matters
If your team uses this to make sense of grant reports, board materials, or survey data, there's nothing you need to do. If you've referenced "NotebookLM" by name in any staff training materials, update it to Gemini Notebook when it's convenient. Old links will keep working.
Free Resources
Free Resource · K–12 Educators
Claude for Teachers, free for verified K-12 educators
Anthropic launched a free version of Claude built for teachers, including lesson-planning tools tied to academic standards in all 50 states and a privacy partnership with the American Federation of Teachers. It's free for at least a year for verified US K-12 teachers.
Why it matters
If your organization runs an early childhood center or afterschool program, this is worth putting directly in front of your education staff. It's free, built with actual teacher input and AFT-backed privacy standards, and could save real prep time without your team having to invent an AI workflow from scratch.
Learn More
Where to Find Me
A couple of places I'd love to see you.
Upcoming Webinar · Thursday, July 30, 9:00–10:00 AM CST · Free
Human-Centered AI Implementation
Jamf Nation Webinar. A practical session for IT and operations leaders on rolling out AI in a way that keeps people in control, accounts for real costs, and delivers genuine value.
Save My Seat
JCC Conference · Oct 18–20, starts 12:00 PM · North Miami Beach, FL
JCC ProCon 2026 — Kehillah 2.0
Two and a half days of insight, connection, and strategy for JCC professionals with a focus on community, hosted by MARJCC.
Learn More