Stop Casting a Wide Net. Start Stacking Constraints.

There’s nothing like getting called into the spotlight. Recently, our VP of HR asked for a simple headcount trend analysis for the next HR leadership team meeting.

"Just give me a clean view of active employees by quarter. And make it easy for me to pull in the future."

I decided I should consult Mando AI because I had a lot of other things on my plate. I'd already mastered the object-first approach, role context, Workday action verbs, and output format. This was going to be another quick win. 

I fired off what seemed like a solid prompt: 

"Help me analyze headcount trends by providing guidance on reporting options." 

Here’s the response Mando gave me. It walked me through Workday's entire trending infrastructure: 

• How to enable "Worker Trending" in tenant setup 

• Running the "Create Worker Trending Data" task 

• Security requirements for the "Trended Worker Data" domain 

• Options for "All Active and Terminated Workers" data source 

• Management Reporting Dashboard configuration 

• Matrix reports vs. composite reports for headcount analysis 

All technically accurate. But here's the thing – and not to brag – but I already knew all of this. Our tenant has had trending enabled for months, security is configured, and the data task runs nightly. What's missing from this breakdown? Step-by-step instructions to actually build the specific report I needed for Thursday's presentation. 

The response gave me Workday system administration guidance when what I needed was executable configuration steps.

The fifth element: population precision

Here's what clicked for me: the four-step framework we've built so far gets you solid AI responses, but it doesn't guarantee relevant AI responses.

Our prompt structure: 

1. Start with the object (Worker, Position, etc.) 

2. Set your role ("As a Compensation Analyst...") 

3. Use a Workday-native verb (create, build, derive...) 

4. Specify the output format (matrix report, calculated field, etc.) 

But there's the fifth element that increases the usability of the AI response: 

5. Stack your constraints (population filters, time boundaries, exclusion rules) 

Without the fifth element, you get correct but broad responses. With constraints, you get deployment-ready configurations that address your specific business scenarios.

The problem: one constraint isn't enough

"Headcount trends" sounds specific, but it's actually wide open:

• Trends over what time period - monthly, quarterly, yearly? 

• All employees or specific populations? 

• Which departments or organizations? 

• Including contractors and consultants? 

• Active employees only, or including terminated workers? 

Without stacking constraints, AI gives you the most general answer possible. Great for learning about Workday capabilities, not so great for leadership presentations.

When constaints go wrong: the over-filtering trap

Before I show you what worked, here's what doesn't: 

"As an HRIS Analyst, create a matrix report showing active full-time employees by department over the last 24 months, quarterly breakouts, excluding contractors, interns, expatriates, people on leave, recent transfers, pending terminations, and anyone hired in the last 30 days."

This prompt stacked eight constraints. The result? A population that changes so dramatically between quarters that trend analysis becomes meaningless. This left Mando going  ¯\_(ツ)_/¯

Over-constraining is real. When you stack too many conditions, you filter your workforce in ways that actually obscure trends rather than reveal them. The sweet spot is three to five constraints - enough precision to eliminate noise, not so much that you lose meaningful patterns.

The solution: strategic constraint stacking

So I rewrote the prompt with four strategic constraints: 

"I need to show headcount growth patterns by department over the last 6 quarters for executive reporting - what's the most effective Workday approach?" 

This constraint combination solved specific business problems: 

• Headcount growth patterns = focus on change over time, not static counts 

• By department = the organizational view leadership needed 

• Last 6 quarters = specific, meaningful trend timeframe 

• For executive reporting = audience context that drives format decisions 

This completely transformed Mando's response.

Now I got actionable, implementation-focused guidance:

• Specific data source recommendation: Use "Trended Workers" data source for historical headcount analysis

• Report type guidance: Trending custom reports with start/end period columns for growth calculations

• Technical implementation: Create calculated fields for growth percentage comparisons between current and prior year 

• Executive formatting: Leverage delivered reports like "Headcount and FTE by Month" as starting points

• Advanced options: Composite reports with repeating column groups for dynamic quarterly display

Same underlying business need, but instead of getting an overview, Mando delivered a deployment roadmap with specific techniques, data sources, and configuration approaches.

Why strategic constraints matter

The difference goes beyond cleaner data. Constraint stacking prevents the technical “gotchas” that make reports unusable for executive audiences:

• Data source precision: Stacked constraints forced the right data source selection ("Trended Workers"), giving me access to quarterly snapshots and department groupings. Without constraints, I got generic guidance about multiple data source options without clear direction.

• Time period logic: The 4-quarter constraint eliminated confusion about whether to use monthly snapshots, annual averages, or point-in-time counts. Executive trend discussions need consistent time boundaries.

• Population clarity: Excluding contractors and interns removed hiring noise that would confuse headcount growth patterns. These populations have different approval processes and timing that don't reflect core business decisions.

• Downstream impact: These same constraint patterns feed workforce planning, budget forecasting, and organizational design discussions. Get the logic right once, reuse it everywhere.

Your constraint categories

Every headcount analysis needs constraints from these categories:

Population scope

• "Active employees in North America"

• "Individual contributors in Technology"

• "Managers and above excluding VPs"

• "With assigned cost centers"

• "Eligible for benefits"

Time boundaries

• "Last 6 quarters"

• "Fiscal year 2024 through current"

• "With assigned cost centers"

• "Eligible for benefits"

• "Monthly snapshots for Q4"

Exclusion rules

• "Excluding contractors and interns" 

• "Not on unpaid leave" 

• "Excluding pending transfers" 

Mix and match three to five constraints from these categories to get trend-specific analysis instead of generic guidance.

Common constraint combinations that work

• Leadership dashboards: Active employees + department grouping + quarterly periods + excluding contractors

• Budget planning: All employees + cost center grouping + fiscal year boundaries + including approved headcount

• Organizational analysis: Active + specific regions + management levels + excluding recent acquisitions

• Capacity planning: Active + project teams + monthly trends + excluding people on leave

Each combination solves different business problems by filtering to the population that matters for the question being asked.

Your turn to try this

Take any headcount prompt you've used recently - something like: 

"Show me workforce data by organization." 

Now add strategic constraints: 

"As an HRIS Analyst, create a matrix report showing active headcount by Supervisory Organization for the last 6 quarters, excluding contractors and employees on unpaid leave." 

You're not just asking for workforce data anymore. You're asking for trend-ready population analysis.

The real test: forwardable output

When I sent the revised report to our VP of HR, she used it directly in the leadership presentation without any modifications. That's the real measure of constraint stacking success - not just clean data, but analysis that executives trust and use immediately. 

The difference was telling Mando AI exactly what population and timeframe I needed analyzed, not just what concept I wanted explained.

What's next

In our next post, we're diving into security context - how to ask Mando why you can't access certain data and uncover the domain policies behind permission restrictions. 

Until then, practice constraint stacking. Take your next prompt and add three filters you normally apply after pulling the report. Constraints don't add friction, they add credibility. That's what turns a Workday report into an executive-ready deliverable.