In Episode 100 of the Sidecar Sync, we made several predictions about what's coming. One stood out: by Episode 200, hundreds of associations will deploy thousands of AI agents. Not someday. About two years from now.
Building agents is becoming ridiculously easy for non-technical folks, and the capability will be everywhere within months. But here's the thing—many people are still using AI like a fancy search engine. One question, one answer, move on. The shift from "AI as assistant" to "AI as workforce" is about to accelerate, and if you're not prepared, you'll be drowning in options without a strategy.
Right now, you probably use AI something like this: You ask ChatGPT a question. It answers. You copy that answer somewhere. You do something with it. Repeat. Agentic AI works differently. You describe a complete workflow, and it handles the entire process. Think of it like the difference between a calculator and a bookkeeper. One helps you add numbers. The other manages your books.
Here's an example. Instead of asking AI "What should I include in a member retention email?" you'd deploy an agent that monitors member engagement, identifies when someone's activity drops, crafts a personalized outreach message based on their history, sends it, tracks whether they respond, and adjusts its approach based on what works. That's one agent doing the job of multiple staff hours. Every day. Without coffee breaks!
The mental shift here is critical. Stop thinking "What can AI help me with?" Start thinking "What role am I filling?" Just like hiring an employee, you need to think about job descriptions, boundaries, and reporting structures. This reframing changes everything about how you approach AI implementation.
Agent builders designed for people who've never written code are currently launching. We're familiar with several, including the free open-source AI Data Platform for associations called MemberJunction (our sister company), but the specific platform doesn't matter as much as the trend. Soon, every major association platform will have agent capabilities built in. Your AMS vendor will call you about their new agent features. Your email platform will announce agent automation. Everyone will have an agent story.
The crucial part: these aren't platforms for developers. They're built for you—the person who uses Excel, not Python. The person who can describe a workflow but can't write code. Just like ChatGPT made AI accessible to everyone, these agent builders make agent creation accessible to anyone who can explain what they want done. If you can train a new employee, you can build an agent.
The same pattern happened with ChatGPT. November 2022: "What is this?" By spring 2023: Everyone had an account. By 2024: It was weird if you weren't using it. We're about to see that same compressed timeline with agents, except this time associations are primed to move fast. You need to prepare your strategy now, before vendors start pitching you their "revolutionary agent solutions" and you have no framework for evaluating them.
You don't need to wait for agent platforms to begin preparing. Start documenting your processes as if you're training a new employee. That member onboarding workflow you do every week? Write down every step. The monthly report you pull together from three different systems? Document the process. The way you handle certification renewals? Get it on paper.
This documentation becomes your agent blueprint. As you write, you'll notice patterns emerging. Some steps are pure process—check this database, send that email, update this field. Perfect for agents. Other steps require judgment calls—is this member's situation unusual enough to warrant an exception? That needs human oversight. You'll see where information lives, where decisions get made, and where the bottlenecks actually are.
Next time you're training someone new, save that documentation. You're not just onboarding an employee. You're creating the specification for your future agent. The associations that have this documentation ready when agent platforms launch will deploy their first agents in days, not months.
Here's where most organizations will get agents wrong. They'll think about individual tasks rather than complete roles. But agents work best when they have clear job descriptions, just like employees. When you're planning your agents, write actual job descriptions that include:
Let's look at what this means for a Member Service Agent. The responsibility would be handling routine member inquiries and requests. The authority would include updating member records, sending standard communications, and processing routine renewals. It would escalate complaints, special circumstances, and financial adjustments over $100. For system access, it needs CRM (read/write), Knowledge base (read), and Email system (send).
This isn't theoretical. Associations are building these agents right now. The ones with clear job descriptions work brilliantly. The ones without boundaries create chaos, sending inappropriate messages or making promises the association can't keep.
Not all workflows are equal. Some are perfect for agents today, while others should stay human-controlled for now. The smart approach is to map your workflows on two axes: complexity and risk.
LOW RISK HIGH RISK
| |
LOW FAQ responses Payment processing
COMPLEXITY Data entry Certification approvals
Appointment scheduling Member status changes
Basic report pulling
| |
─────────────────────┼──────────────────────┼─────────────
| |
HIGH Data analysis Strategic decisions
COMPLEXITY Report generation Sensitive member issues
Content creation Crisis response
Research tasks
| |
Start with the low complexity, low risk quadrant (upper left). These include FAQ responses, data entry, appointment scheduling, and basic report pulling. If an intern could do it with a checklist, an agent can do it better. These are your training ground for understanding how agents work in your organization.
Next, tackle high complexity, low risk tasks (lower left). Data analysis, report generation, content creation, and research tasks fall into this category. These need more sophisticated agents but mistakes won't tank your reputation. You can experiment here, push boundaries, and learn what's possible.
For low complexity, high risk processes (upper right)—payment processing, certification approvals, member status changes—proceed carefully. The processes are simple but errors have consequences. Build in extra safeguards and human checkpoints.
Keep high complexity, high risk work (lower right) human-controlled for now. Strategic decisions, sensitive member situations, and crisis response need human judgment, at least until agent technology matures further. This framework answers the question you'll face in three months: "We can build agents now. Which one first?"
By Episode 150, here's what agent deployment could look like for associations:
Every employee needs boundaries. So does every agent. The challenge is finding the sweet spot between autonomy and control.
The principle to remember: Automate the routine to free humans for the exceptional. Let agents handle the 80% of standard cases, and let the humans handle the 20% that need judgment, empathy, or creativity. When a longtime member calls furious about a billing error and threatens to quit after 20 years, that needs human empathy and problem-solving. When someone wants to know their certification expiration date, that's perfect for an agent.
Your first agent will be clunky. It'll need constant adjustment. You'll wonder if the effort is worth it. But your fifth agent will be sophisticated because you'll have learned what works, what breaks, and what members actually want. Your twentieth agent will do things you can't imagine today, working together with other agents, passing tasks between them like a well-trained team.
Associations starting in 2025 get two years of learning before we make it to Episode 200. That's two years of refining, improving, and discovering what's possible. The association that waits until 2026 won't be one year behind—they'll be generations behind in practical experience. Every month you wait, the gap grows. Not linearly. Exponentially. The associations deploying agents now are learning lessons that can't be taught, only experienced.
Two years sounds like forever. It's not. By Episode 200, associations will casually mention their "agent ecosystem" like they talk about their email system today. Job postings will list "agent orchestration experience preferred." Board meetings will include agent governance discussions.
Some associations will have 50+ agents handling everything from member service to event planning to content creation. Others will still be doing everything manually, wondering how competitors offer 24/7 service with half the staff. The divide won't be between technical and non-technical associations. It'll be between those who started thinking about agents as employees versus those who kept thinking about AI as a tool.
Manual workflows will seem as antiquated as carbon paper. The question won't be whether you have agents. It'll be how sophisticated yours are compared to everyone else's. The associations that started documenting processes today, that wrote job descriptions for their agents, that thoughtfully designed their agent workforce—they'll be the ones setting the pace.
The tools exist. The cost is negligible. The knowledge is a conversation away. Start documenting your workflows this week. Write that first agent job description. Pick your initial candidate from the low-risk quadrant.
By Episode 110, you could have one agent running. By Episode 150, you could have five. By Episode 200, you'll wonder how you ever worked without them. The world really is your oyster. Every workflow can become an agent. Every repetitive task can be automated. Every manual process can be transformed.
Welcome to the agent era. Turns out Episode 100 was just the beginning.