Michael Domingo

But today, we are going to stop talking about the stain and start cooking the pasta.

Because with the release of the Workday Flowise Agent Builder, you, yes, you, the person who still gets nervous when editing a calc field—can now build a digital employee.

Is that exciting? Yes.

Is it absolutely terrifying? Also yes.

It’s like handing a chainsaw to a golden retriever. It might cut down the tree, or it might accidentally prune your arm off.

So, severed limbs aside, let’s put on our safety goggles and look at how to build, govern, and unleash your first agent without accidentally violating the Geneva Convention.

First, let’s talk about Flowise. Workday acquired this low-code platform because, frankly, asking functional consultants to write Python is a human rights violation.

Flowise is a "visual" builder. That means you drag little boxes around a screen and connect them with lines, like a conspiracy theorist mapping out Bigfoot’s migration patterns.

Here is your step-by-step guide to building your first agent. Let’s call him "Policy Paul."

Step 1: The "Hello World" Template

Don't be a hero. Do not start from scratch. Open the Agent Builder and select a template. For your first go, pick something low-stakes like "Policy Q&A" or "Simple Report Runner". Starting from blank canvas is hubris. It’s like trying to bake a soufflé before you’ve mastered toast. Just grab the toast template.

Step 2: Give Paul a Brain (The Knowledge Base)

You need to tell Paul what he knows. In Flowise, this is called "RAG" (Retrieval-Augmented Generation), which sounds like a cleaning rag but is actually just a fancy way of saying "Read the Manual".

Action: Upload your "Employee Handbook.pdf" or point Paul to a specific Workday Knowledge Base article.

Step 3: Give Paul Some Hands (The Tools)

An agent that can only talk is just a chatbot. An agent that can do things is a minion. You need to equip Paul with Tools.

Action: Drag in the "Workday Custom Report" tool. Connect it to a specific report, like "My Time Off Balance."

The Logic: Now, if someone asks, "Can I take Friday off?", Paul doesn't just recite the policy; he checks their balance and says, "Technically yes, but you only have 5 hours left, so maybe just work a half-day, Stephanie."

Step 4: The "Don't Be Racist" Check (Testing)

Before you let Paul talk to humans, you test him in the Playground. Ask him weird questions. Ask him illegal questions. Make sure his temperature (creativity) isn't set so high that he starts hallucinating new company holidays like "National Honda Odyssey Day". (Hey! Minivan drivers deserve a day too!)

So you’ve built Paul. He’s smart, he’s helpful, and he’s ready to work. Now you have to face the final boss:

Governance.

Workday has built something called the Agent System of Record (ASOR). Think of ASOR as the DMV for robots. You cannot drive on the corporate highway until you have registered your vehicle and proven it’s not going to explode.

The Governance Checklist:

Register the Agent: You literally have to "register" Paul in ASOR. You define his Role (e.g., "HR Assistant") and his Skills.

Define the Blast Radius: You must explicitly state what data Paul can touch. This is the Security Group for agents. If Paul is a "Policy Agent," he should not have access to the "Compensation" domain. If he does, that’s not an agent; that’s a lawsuit.

Human in the Lead (HITL): For high-stakes actions (like "Approve Bonus"), you must configure a checkpoint where a human (you) has to click "Yes" before Paul executes the task. It’s the "Are You Sure?" popup for reality.

The "Approval" Process:

It’s not just a button in ASOR. It involves your AI Ethics Council (or just Janis in Legal) reviewing the "Risk Evaluation". They will ask: "Does Paul hallucinate?" "Is Paul biased against people named Steve?" You need to have answers. This part will probably take much longer than expected. Heck, it's going to be longer than the actual development cycle. Make sure you are prepared.

If building Paul sounds like too much work, good news! Workday has announced a fleet of Standard Agents (aka "Illuminate Agents") that are coming to take our j...er, I mean, make our lives easier.

Here are a few you should know about:

The Recruiting Agent: It sources candidates, schedules interviews, and writes rejection emails that are almost as cold as the human ones.

The Expense Agent: It looks at receipts, notices you ordered the "premium" guacamole, and flagging it for audit. It is the narc of the AI world.

The Job Architecture Agent: It looks at your messy, chaotic job titles and suggests a clean taxonomy. Finally, someone to tell the tech bros that "Senior Ninja Rockstar II" is not a real job title.

The Payroll Agent: It automates payroll tasks and finds errors. This is the high-stakes one. If this agent messes up, we riot.

Building your first agent in Flowise is going to feel weird. It’s going to break. Paul is going to say something stupid.

But you have to do it. Because the alternative is letting the "Shadow AI" epidemic continue, where your employees are pasting sensitive data into ChatGPT because you didn't give them a better tool. And trust me, they're already doing it (cough, Performance Reviews).

So go build Paul. Register him in ASOR. Give him a limited security policy. And for the love of god, keep him away from the "Termination" business process.

Here's a bonus video of Flowise I found so you can see it in action - https://www.youtube.com/embed/tX2LtzwFa9o