Here’s how I handed off my FAST Manager role to an AI agent.

When most people hear “AI,” they think of ChatGPT: great for brainstorming, writing, research, and social posts. But what if AI could actually complete tasks for you instead of just helping you think through them? A small group of animal advocates in our movement have been using AI agents that do exactly that.

AI agents connect directly to your workflow and move data between tools. Imagine logging new PayPal donations, syncing them with your donor database, and posting a daily impact report in Slack automatically. These are actual tasks completed, just like having your own assistant!

Some of the newer tools even go a step further: they can literally take over your cursor and do the work for you! They’re still a bit rough, but this is coming fast. The future’s here, and it’s our chance to ride the wave and put these tools to work for animals.

In this post, I’ll walk through how we built our own agent that nearly replaced the entire FAST Manager role. The goal is to show how AI can handle repetitive tasks better left to machines, freeing us up to spend more time on the work AI can’t (at least not yet).

If tech isn’t your thing or you’re short on time, here’s the TL;DR:

TL;DR (Too Long, Didn’t Read)

• Up until now, FAST used to be a Google Group – a big mailing list that allowed members to send an email to thousands of subscribers in one. It required a surprising amount of time to run properly for the 600 organizations and 1,800 community members using it since 2012.

• Not only did I hand off 95% of my FAST Manager role to an AI Agent, but because of that, FAST is more secure, faster, richer in features (like the new ability to select what type of emails you’d like to receive), and just overall better across the board.

• FAST used to frequently disrupt my day with context-switching, and I was losing about 3–7 hours of productivity every week. If my role wasn’t volunteer-run, I’d value it at around $1,500/month. Now, it runs itself for $60/month, and I reclaim 25+ hours at less than $3/hour.

• You might be surprised how much AI Agents can automate. For FAST, I had 15+ weekly tasks: managing accounts, submissions, digests, issues, and handling approximately 25 different types of support requests daily related to our list, resources, academy, and more.

• A few tools now manage everything 24/7: n8n (our AI agent of choice), Mailgun (our email provider), a single Google Sheet and form, plus some GPT written code. Together they handle rule-checking, member validation, preferences, and sending without delays.

• Are you interested in automating your own work? If so, fill out this form to let us know what kind of work you’d like to have automated, even if you’re not sure it can, and we’ll review it and reach out.

So, what is FAST anyways?

In short, FAST used to be a Google Group – a classic “listserv,” which was basically just a big mailing list that let someone send an email out to thousands of people in the group all at once.

But before it even became a Google Group, FAST started in 2012 as a simple email list – literally an Excel sheet with a few dozen reps from a handful of animal orgs. The original team ran it first, then it passed to Chris Holbein, then Amanda Cramer, and finally to me. Over the years, it just kept growing. What began as a small, volunteer-run effort now connects nearly 1,800 community members from over 600 organizations, all united in the fight to end factory farming.

And as the movement grew, so did the group’s needs. I took over in 2024 and have been slowly improving things bit by bit, but like every manager before me, I quickly found that FAST somehow still managed to take up significantly more time than you’d expect.

What does a FAST Manager actually do?

…and why would it take so much time that it would need automating in the first place?

Here's what I was dealing with on the daily:

• Adding new members, but only after researching their role in animal advocacy first

• Removing subscribers when requested from the list, from the forum, and from the portal

• Responding to the many requests of updating member emails based on new jobs/positions

• Reviewing every single daily email through FAST to make sure it follows our (many) guidelines

• Explaining rejections to emails submitted, whether because of their quality or contents.

• Tracking and making sure members don’t submit emails too frequently or making sure their links work

• Tweaking member’s Google Groups settings when requested (daily vs. weekly digests)

• Walking people through posting job openings or volunteer gigs in the right spot

• Explaining to new members how to properly send emails so they hit the FAST list properly

• Tracking down “missing” emails or digests and pushing them out if they stalled

• Fixing bounced or blocked emails and nudging people to check their spam filters when issues occur

• Pointing users to quick how-tos for sharing files, links, PDFs, or forum posts (where to host files)

• Preparing and publishing the weekly job and victory digests from posts published on the FAST Forum

• Updating forum accounts tied to old emails and handling portal-access hiccups

• Letting members post anonymously to FAST while still screening for abuse

• Reviewing resource submissions for the Resource Hub and performing quality control on links

• Reviewing course submissions and queueing them for publishing on FAST Academy

• Fixing submissions where users added “FAST” in the email title directly or hit “reply-all”

• Responding daily to around 25 different support style tickets from 1,800 members

I’ve done my best to manage these items by using templated responses, updating FAST so that things are more intuitive, and adding processes to solve recurring problems. But even with all that – it was a lot.

The challenges and goals

My goal wasn’t just to automate everything. I wanted to show that FAST could actually run better without me, aka a human involved. My idea was to, for the first time ever, give members control over what kinds of emails they get, without changing how they use FAST.

Now, when someone sends an email to FAST, the AI automatically reads it, then sorts it into the right category, and only sends it to members who’ve asked for that type of content. Just like that, everyone gets a more relevant, personalized FAST experience.

The first big challenge was getting off Google Groups. It’s both old and hasn’t had a real update in over a decade. I imagine Google has kept it around because of the many old, but massive communities that still rely on it, but unfortunately it was never going to support the kind of functionality we need.

Second challenge: I didn't want to create any websites or custom dashboards. The goal was to prove you could build this with tools you already use every day. This is especially feasible when tasks follow predictable patterns that AI can handle more consistently than humans.

Privacy Concerns

It’s reasonable to be cautious about AI, especially with all the noise around data privacy and models “training on your emails.” FAST doesn’t work that way. According to n8n’s policy, your data isn’t used to train public AI models. Access is strictly controlled with privacy-by-design practices like data minimization, RBAC, and encryption.

FAST emails are never used for training. Still, it’s smart to review the privacy policies and documentation of any LLM or agent platform you use.

Here’s what FAST looks like under the hood now

Let me walk you through each component:

Google Sheets:

We’ve moved every member out of Google Groups and into Google Sheets. That’s now where our entire database lives, member preferences and all. Full circle, considering FAST started as an Excel sheet!

Technically, our Google Group still has one member left – the AI agent! We kept it that way so people could still use the old Google Groups email to submit messages during our transition. This means that if you send an email to the old FAST submission address, it simply forwards it to our more modern one at [email protected].

With the new system in place, when you send an email, you’re no longer sending it to 1,800 recipients – you’re sending it to one: the AI agent, who grabs it, processes it, and takes it from there.

Google Forms:

This is how we collect and update member preferences. The form, linked in the footer of every FAST email, lets people subscribe, unsubscribe or pick which kinds of emails they want to receive. That Google Form feeds directly into our master Google Sheet, where the AI Agent makes sure every member’s preferences are safely stored.

Mailgun:

Instead of relying on Google Groups to send emails, we now use Mailgun to send all 1,800+ emails in one go. It’s fast, easy, cheap, and reliable. Mailgun’s API gets its instructions directly from the AI: who to send the emails to and what to send.

JavaScript:

We used a bit of code to stitch member emails into a single BCC list based on their preferences, and a small Cloudflare Worker for confirmation links. We also added a step to ensure images display properly in emails (cid mapping, for the nerds reading).

ChatGPT:

I used ChatGPT to write the above JavaScript! No coding experience required. I simply told it my goal, copied what it gave me, and followed the steps. If you can follow directions, you can do this too! ChatGPT also came in clutch as a thought partner for some of the more frustrating edge cases.

n8n:

Finally, we’ve got n8n running the whole show. It’s our new “Fast Manager” AI Agent – reading the incoming emails, applying the right rules, checking who the sender is, matching it to member preferences, and sending the emails out. All automatically, 24/7.

Our n8n AI Agent also scrapes the FAST Forum’s Careers & Volunteering and Victories & Achievements sections to build our Weekly Digests, connects to Hive Slack to send relevant emails to the right channels, and even handles most of the FAST support and inquiry emails too!

The workflow is straightforward:

Form ➜ Sheet ➜ n8n ➜ Mailgun ➜ Members’ inboxes. No humans in the loop!

So, what does my role at FAST look like?

Pretty minimal: unless something crashes, all I do is keep an eye on things. But even that’s automated! The agent simply emails me if a message fails to send or if there’s any kind of error. Which means that my interactions with FAST are about the same as anyone else’s – just an occasional check-in.

But building something this automated meant dealing with edge cases – all those unexpected scenarios that could break the system.

Edge Cases

This is exactly where ChatGPT shines. We’ll dip into some technical detail below, but keep in mind that most agents don’t need this much complexity, and when they do, LLMs make excellent agent-building partners.

Here’s one example: What’s stopping someone from entering another person’s email on the preferences form and changing their settings? To prevent that, you can send a confirmation email to verify it’s really them. But then you hit more issues. For example, what if the link never expires? Or what if someone edits the link parameters to spoof another update?

The solution: Whenever someone updates their FAST preferences, n8n sends them an email with a confirmation link. That link includes the row number, the new settings, a timestamp, and a secure signature. A Cloudflare Worker (a tiny program that runs on the cloud) double-checks the signature and rejects anything older than 48 hours. If even a single character is changed, or the link is stale, it fails. Only a valid link updates the Sheet, and the user lands on a clean “Preferences updated” page.

The best part is that I didn't write a single line of code for this – I simply explained the problem to ChatGPT and it gave me the exact steps: how to generate signed links, set up the Worker, and hook it all up in n8n.

Several other edge cases showed up, but like the one above, those were quickly taken care of by GPT.

You're probably wondering about the practical side: how long does this actually take to set up?

Time Investment

So how much time did this actually take, and how does that compare to the time I usually spend on FAST? It’s hard to pin down. Some days FAST barely took any time; other days it pulled me in for hours. I’d estimate I was losing around 3–7 hours a week in lost productivity.

The automation came together in two weeks, but I spent two months strategizing, debugging, and testing before I was confident enough that this was the right solution. While two months seem long, FAST has been running for over 10 years and will for many more, so that time investment will pay off. Not just through direct impact, but by teaching me how to think in agents and build them.

Now I can automate other things much faster. So, even if you worry the setup time might outweigh the time saved, it’s worth remembering: you’re also building a skill, which compounds later.

Build Your Own AI Agent

If you want to try building an AI Agent yourself, tools like n8n, Make, Lindy, or Zapier make it possible, with n8n being the most powerful and popular right now. You can dive in, experiment, and figure things out by asking GPT (or your favorite LLM) for updated steps when you get stuck.

Or, if you’d rather skip the trial-and-error, Vegan Hacktivists is now offering a service to design custom workflows for orgs like yours, complete with 1-on-1 guidance so you actually learn how it works. If that sounds easier, just reach out to us and tell us what you'd like to automate.

See you on the other side. :)

Recommended Resource

The point of using AI isn’t to add more complexity, but rather to actually save time and free you up to do more for animals. A lot of people in nonprofits feel stuck here: they know AI could help, but they’re not sure where to start or how far along they are.

That’s why I highly recommend you check out AI Impact Hub’s AI Adoption Maturity Quiz. It’s a simple way to see where you’re at – from “just getting curious” to “building custom AI tools.” There’s a quick quiz if you want guidance on next steps, or you can dig into the full PDF if you prefer the details.

AI Impact Hub has also been a big help in my own AI journey, offering practical skills, ongoing learning, and a peer community to learn alongside. If you haven’t checked it out yet, I’d definitely recommend it.