What Actually Makes WorkBuddy Tick: A Closer Look at Care Workflows

Care Work Is Messy. Here's an Idea.

Let's face it: care work is a tangle. You're juggling patient follow-ups, coordinating home health aides, tracking wellness plans. Spreadsheets, emails, calendars—a dozen different tools. And you're trying to remember what's next. That's where something like WorkBuddy—an AI agent—starts to make sense. Not because it's flashy. Because it can stitch together the parts of a care workflow that usually slip through the cracks.

I spent a while digging into how WorkBuddy is built. What I found surprised me: it's not really about the AI. It's about how the work is structured. The product splits an agent's abilities into five features: Connectors, Skills, Experts, Expert Teams, and Inspiration. Each has a specific job. Once you see that, it all clicks.

Connectors: The Hands and Feet

A large language model on its own is useless. It can't open your email, check your calendar, or pull up a resident's care plan from a cloud drive. It's like a brilliant assistant with no hands. Connectors give it hands.

In care settings, you might have a Connector for your electronic health record, a scheduling app, or Slack. When you connect one, the agent can read and write to that system—with your permission. For example, to automatically create a follow-up appointment, you'd connect your calendar. The first time, you authorize access. After that, just say, "Schedule a follow-up for Mrs. Chen on Thursday at 2 PM," and it happens.

But here's a warning from the source: don't connect everything at once. Each Connector adds a description to the agent's system prompt. Too many options confuse the model. It might pick the wrong tool. That's a risk in care. If you have both a scheduling app and a messaging app, and you ask for a reminder, which one? The agent might guess wrong. Keep it lean.

Skills: The How-To

Connectors give access. Skills give instructions. A Skill is a packaged set of steps—sometimes with scripts—that the agent follows to finish a multi-step task. For care, that could be a "Patient Onboarding" Skill, a "Medication Refill Reminder" Skill, or a "Weekly Care Summary" Skill.

Say you create a "Weekly Care Coordination" Skill. It might include: pull all appointments from the calendar, check each patient's care plan for updates, draft a summary email for the care team, and log it in a shared doc. The Skill defines the method. But it still needs Connectors to access the calendar and doc. Skills and Connectors work together.

A common mistake? Thinking a Skill alone does everything. If the Skill depends on external systems, you need the right connectors and authorization. A Skill is a recipe, not a kitchen.

Experts: The Right Mindset

Now we go up a level. If Skills are about the how, Experts are about the who. An Expert is a persona with a specific methodology and perspective. For care, you might have a "Care Coordinator Expert" that knows how to triage patient needs, or a "Family Liaison Expert" that talks to relatives with empathy.

The source uses a framework called NLP logical levels. Connectors are environment. Skills cover capabilities and behaviors. Experts sit higher—beliefs, identity, values. So a Skill might say "send a reminder." An Expert decides if a reminder is even appropriate, and if so, how to phrase it based on the patient's history and tone preferences.

Care is nuanced. A generic reminder works for some patients. Others need a softer touch. An Expert can encode that judgment.

Expert Teams: When One Mind Isn't Enough

Some tasks are too complex for one Expert. That's where Expert Teams come in. Instead of forcing one agent to play every role, you assemble a team—say, a nurse, a social worker, and an administrator. The team lead breaks down the task, assigns pieces to each member, then integrates their outputs.

Imagine coordinating discharge for a patient. The nurse Expert reviews clinical notes. The social worker Expert checks home support resources. The administrator Expert schedules follow-ups. The team lead combines everything into one discharge plan. That's way more efficient than one agent trying to do all three roles in sequence.

Inspiration: Don't Start from Scratch

Finally, Inspiration. This one's easy to misunderstand. It's not a tool you use—it's a library of complete solutions. When you see a care workflow someone else built—like a "Fall Risk Assessment" template—you click "do same" and WorkBuddy loads the associated Prompt, Skills, and Experts. Then you customize with your own data.

That's a lifesaver for care teams without time to build from scratch. Instead of learning the ins and outs of Skills and Experts, you find a template and adapt it. Like borrowing a proven recipe rather than inventing one.

Putting It All Together

Let's walk through a realistic scenario. You've just had a care coordination meeting with a family. You need to turn that conversation into a plan. Here's how the pieces fit.

First, data. You use a Connector to pull the patient's history from your EHR and meeting notes. Next, apply a Skill to clean up the transcript—extract concerns, current status, decisions. Then bring in an Expert, say a "Care Plan Specialist," to analyze and recommend next steps. If the case is complex, deploy an Expert Team to handle clinical, logistical, and emotional aspects. Finally, save the whole thing as an Inspiration so next time you just hit "do same" and swap in new details.

What This Means for Care Teams

The lesson? WorkBuddy isn't about turning you into an AI expert. It's about wrapping up the technical complexity so a care professional can use it without a degree in prompt engineering. You don't need to know what a system prompt is. You just need to know that an Expert acts like a specialist, and a Skill tells it what to do.

For care practices, that's a big deal. Care work is collaborative, data-heavy, and requires judgment that's hard to codify. But by breaking the workflow into five pieces, you can automate the mundane parts while keeping human judgment where it belongs.

Start Small

If you're new to this, don't try to master everything at once. Start with Inspiration—find a template for a recurring task, customize it. As you hit gaps, tweak Skills. When you need access to a new system, set up a Connector. When you need deeper analysis, create or find an Expert. Only when a task truly needs multiple perspectives should you pull in an Expert Team.

The point is to get work done, not to become an AI architect. That's the real takeaway from WorkBuddy's design: it lets you focus on the care, not the technology.