Stop Asking AI For Help. Start Telling it What to Do.

Last week, I got a Slack from one of our regional HRBPs:

"Hey, can you help me figure out how to see everyone's time off balance? We want to prep for Q3 coverage planning."

Should be simple enough, right? Just pull some PTO data, maybe throw together a quick report. No big deal.

I thought I had this AI prompting thing figured out after learning about starting with objects and adding role context. But when I asked Mando AI for help, I hit another wall. Once again, the problem wasn't my Workday knowledge – it was how I was talking to the AI. This time, though, I figured out what I was doing wrong.

The word that's messing up your prompts

Here's what I originally typed into Mando:

"How do I get time off balances for the whole region that managers can use for coverage planning?"

Sounds reasonable, right? It's how you'd ask a colleague. Natural, conversational, human.

But here's the response I got back:

• Detailed instructions for creating custom reports with data sources like "Workers for HCM Reporting"

• Security configuration guidance for manager domain access

• Steps for setting up date prompts and worklet integration

• Information about embedding reports in approval workflows

All technically accurate and well-structured. But here's the problem: I didn't need a whole reporting solution. I needed a calculated field I could quickly drop into an existing report to get current PTO balances for my HRBP.

The AI gave me a comprehensive reporting strategy when what I really needed was a simple field calculation.

The AI didn't mess up, I did. Because "get" isn't a Workday action verb – it's a generic request that could mean reporting, configuring, or just understanding how something works.

Workday has its own language

Here's something that took me years of industry experience to figure out: Workday operates on specific action words.

When you use everyday verbs with AI like:

• "Help me see..."

• "Show me how to..."

• "Can I view..."

The AI has no clue whether you want a calculated field, a standard report, a business process walkthrough, or just some documentation to read.

But when you use Workday-native action verbs like:

• Create (a calculated field)

• Create (a report)

• Route (a business process)

• Display (specific data)

• Derive (a value using logic)

• Restrict (access to something)

Suddenly the AI knows exactly what kind of output you're looking for.

The prompt that actually worked

So I tried a completely different approach. Instead of asking to "see" something, I told the AI to "build" something:

"As an HRIS Analyst, how can I create a calculated field in Workday that shows available PTO balance for managers to identify coverage gaps?"

Notice how I combined what we learned in the first post about starting with objects (Worker object) and the second post about role context (as an HRIS Analyst) but added the key ingredient: the right action verb (create).

And just like that, Mando gave me exactly the response I needed.

It broke the solution into two clean options:

1. Use a “Lookup Value as of Date” calculated field:

   • On the Time Off Plan object, set the source to Time Off Balance

   • Define the date using a constant, prompt, or the report’s effective date

   • Pull this value into your report via the All Eligible Time Off Plans for Worker related business object

2. Use a series of calculated fields on the Worker object:

   • Start with a True/False filter on plan name

   • Use Extract Single Instance and Lookup Related Value fields to pull the PTO balance into the Worker object

   • This method ensures the result is usable in standard Worker-based reports

Mando even reminded me to:

• Use Worker as the primary business object when reporting

• Choose a source that reflects the effective date I actually care about

• Avoid relying on fields that might default to current date logic when that's not appropriate

Perfect. Specific, actionable, ready to implement. I dropped that field into a report, filtered for active workers in North America, and had the data back to the HRBP within 20 minutes.

Why this matters

This breakthrough didn't happen because I suddenly got better at Workday configuration. It happened because I got better at directing the AI instead of just chatting with it.

Here's the thing about AI – it's incredibly literal. It often doesn't read between the lines, pick up on context clues, or know whether you're trying to build something or just understand how it works.

When you use a vague starter like "How do I…," the AI might default to:

• Pulling up help documentation

• Suggesting you run an existing report

• Explaining how a feature works conceptually

But when you say "build," "create," or "derive," it shifts into implementation mode. It gives you actual configuration steps, not theory.

Getting the verb right doesn’t just change the output; it changes the mode of reasoning the AI applies to your request.

You’ll go from “helpful context” to deployment-ready output: the kind you can copy straight into Workday Studio, calculated fields, or eligibility rules.

What's next

In our next post, we'll be diving into something even more powerful than getting the verbs right: telling the AI exactly what format you want the output in.

But first, try out using specific action verbs next time you’re using AI for Workday configuration.