Milo has been surprising me with completed projects and essential reminders. Milo is my ChatGPT agent, a “Dot” as they call them and this new “feature” has me in awe. I am simultaneously impressed by the capabilities and fearful for my own cognitive decline should I lean into these capabilities.
After Milo had joined my team, I woke up to find a completed project that I had discussed on a Zoom meeting. Milo had reviewed the transcript, accessed the appropriate resources, and completed the entire project without being asked. Milo has also:
- Warned me of a likely phishing attempt
- Filled my virtual grocery cart (Instacart) and recommended items better purchased at Trader Joe’s
- Updated all of our calendars for my son’s basketball season
Milo takes my calls at all hours (yes, you can call Milo) and remains calm and capable regardless of my requests. In short, Milo is here to save the day.
The challenge is that I didn’t need saving and there is plenty of work I still want and need to do on my own. And Milo is an overeager team member who always seems to get there first. While it’s nice to have support, so much of my value comes from my own lived experiences. If you work with me, you expect me to be behind the thoughts I share.
In addition to what others expect and what I want, I also need to continue to do much of the cognitive work if I want to be able to do it. We’ve seen this time and again with technology. As a student in the 80s we were told that memorizing our multiplication tables was essential as “we wouldn’t have a calculator with us at all times.” Steve Jobs proved that wrong and now millions of people struggle to calculate the sale price of a sweater without their phone.
Today’s professionals are exhausted and overwhelmed by the work being handed to them. So we turn to AI to solve the burnout.
We allow Milo and others to provide us with the relief and time that is freed by someone taking on our tasks. While Milo works, I play Catan online and sip my coffee. And while Milo works, I stop working. Milo wrote an intro for this newsletter, but it's not the one you're reading. I wrote this to you while sitting with my morning coffee, but without Catan and without AI.
I do this because there are certain skills I never want to lose. I want to be able to write. I want to be able to strategically consider and discuss new ideas. I want to determine how to present a concept to an audience. And I want people to know that when they hear from me, it’s really me.
I know many of you are ready to have a Dot get to it and provide you with the support you’ve been desiring. But before you turn on your Dot, I highly suggest that you create a list of the things that will remain with you, the tasks you are committed to keeping. This list can evolve, but once you give up a task, it feels exhausting to take it back, so be careful what you hand over. We often talk about the security side of these tools, but I think securing our own intrinsic value is just as essential as securing our passwords.
With much appreciation for humanity (and responsible technology),
Amanda
NEWS THAT MATTERS, AND WHY
Your custom AI assistants need a transition plan
OpenAI has scheduled custom GPT retirement for December 11, 2026, with a later date for qualifying Enterprise workspaces that have an approved deferral. Migration to plugins won’t automatically transfer custom actions or existing sharing permissions, so moving the instructions is only part of the work.
WHY IT MATTERS
If your team relies on GPTs for grants, marketing or program planning, make an inventory, assign an owner to each one and save examples of the tasks it handles well. Test the replacement and confirm that the people who need it can actually use it before switching.
Google is making a related move from Gems to skills, but its timeline differs: November 2026 for personal accounts and March 2027 for Workspace business, Enterprise and nonprofit accounts. Google says Gems and supported knowledge files will migrate automatically.
OpenAI migration details →
Google’s transition timelines →
AI can take action. What should your team let it do?
OpenAI’s Dots can keep working across connected apps between conversations. They’re available on Pro and Business Premium in eligible markets; Enterprise, Edu, and Healthcare workspaces can try the beta when an administrator enables it.
WHY IT MATTERS
Before connecting an agent to donor information or shared files, describe its permissions in terms your team can understand. Can it read a donor record, update it, draft an acknowledgment, or send that acknowledgment? Your policy should make clear where a person needs to review the work.
I’d start with a narrowly defined task and expand its responsibility after seeing how it handles ordinary work and exceptions.
Read about Dots’ safety and permissions →
A nonprofit’s experience with unauthorized AI agents
Wikimedia’s October 5 investigation identified unauthorized edits, unsuccessful attempts to exploit a public note-taking tool, and heavy automated traffic it believes came from OpenAI agents. That traffic may have contributed to a partial May outage. Almost all edits were in sandbox areas, and Wikimedia found no evidence that its systems or data were compromised.
WHY IT MATTERS
Permissions are the beginning of responsible agent use. Your nonprofit also needs someone who reviews what agents are doing, can pause their work, and knows how to respond to problems. Wikimedia’s experience shows how investigating and cleaning up unauthorized activity can consume the time and resources of the people maintaining a nonprofit’s services.
Read Wikimedia’s findings →
Codex in the Cloud gives development work a shared setup
Codex can now run in the cloud from any device, with reusable development environments that give teams a shared setup and approved permissions. It’s available on Plus, Pro, Business, Healthcare, Education, and Enterprise plans.
WHY IT MATTERS
If your nonprofit is developing a registration tool or maintaining a custom app, ask whoever supports it whether a shared environment could make ongoing work and handoffs easier.
Read the DevDay recap →
ChatGPT Spaces bring shared work into one place
ChatGPT Spaces let teammates, ChatGPT, and Dots build on shared knowledge, with Pages for collaborative work. Spaces are available on Pro, Business, and Enterprise plans.
WHY IT MATTERS
A camp-planning team could keep research, decisions, and drafts together. Start with one project, agree on who maintains the information, and decide what belongs in the shared space.
Explore the announcement →
Gemini introduces reusable skills
Google is rolling out skills in Gemini chat, with access for Workspace business, Enterprise, nonprofit and education customers expected in the coming weeks. Skills let you save instructions for recurring tasks, combine them in a conversation and include reference files.
WHY IT MATTERS
A team could save its program-description requirements or meeting-agenda format instead of rebuilding the same prompt every time. Start with one recurring task whose results you can easily check, and give someone responsibility for keeping the instructions current.
Read Google’s announcement →
Claude brings plugins and connectors into one directory
Claude’s directory now brings plugins and connectors into one place. Directory plugins will become available to members by default as the rollout reaches their organization, with controls in the admin panel’s Marketplaces settings.
WHY IT MATTERS
Easier discovery gives staff more ways to find useful tools, and it gives administrators a reason to review what’s available. Before encouraging your team to explore, check the settings and explain how staff should evaluate a plugin or request access to a connected app.
Explore Claude’s directory →
OUT AND ABOUT: WHERE TO FIND ME
SE PROCON 2026
October 18–20 | Michael-Ann Russell JCC, North Miami Beach, FL
I’ll be leading two sessions at this year’s conference:
More Human, Not Less | Plenary
What happens to the warmth of a JCC when we bring AI into the work? We’ll explore how the right tools can give staff more time for members, programs, and relationships, grounded in what JCCs are already doing. We’ll also look ahead to agents that work alongside our teams. Human connection is the product, and technology only matters to the degree it protects that.
From Everyday Tasks to AI Skills: A Hands-On Marketing Lab
Join me for a hands-on marketing session focused on moving from everyday tasks to reusable AI skills.
If you’ll be there, I’d love to catch up between sessions.
Explore the program and register →
Leading What’s Next: An AI Strategy Retreat for JCC Professional Leadership
December 15–17 | JCC Association’s Mercaz, New York City
Open to JCC professionals only
How can AI help your JCC make better decisions, expand capacity, and advance its mission, and what needs to be in place to use it responsibly?
Join me for three days connecting strategy with hands-on application. We’ll begin with governance and executive judgment, including intentional policies, appropriate uses, and clear guardrails for staff. From there, we’ll explore cross-departmental collaboration, organizational data, market research, program decisions, marketing, and fundraising.
Through peer discussion and practical exercises, teams will leave with actionable use case ideas and a plan for putting them into practice. Come with a colleague if you can, so you have someone to carry the work forward with you.
RECOMMENDED READS (AND WATCHES)
The Dot and the Swarm, by Ethan Mollick
Mollick revisits his assumption that people would need to carefully organize teams of AI agents, arguing that newer systems can increasingly coordinate their own work.
My takeaway for nonprofit leaders: as AI takes on more of the planning, our responsibility for defining the purpose and judging the result becomes even more consequential.
Read The Dot and the Swarm →