How to use AI-generated data quality rules in Nakisa Decision Intelligence

In this post from our new series, our CTO, Faraz, explores how AI is transforming data quality management. By prompting decision intelligence platforms with precise instructions, leaders, analysts, and HRBPs can instantly generate comprehensive, reusable data quality rules, covering integrity, validity, and completeness, and adapt them across HR, Finance, Real Estate, and beyond.
Faraz Ahmed blog image
Research and Development at Nakisa
AI generated data quality rules NDI blog

HR data quality has always been a key component of efficient organizational strategy. Lately, I’ve been exploring how AI can accelerate and elevate the way we define and operationalize data quality. Today, I’m sharing a set of comprehensive, AI-generated instructions designed to assess data integrity, validity, and completeness within HR datasets.

What’s exciting is that these rules weren’t manually compiled. They were created with the help of AI, and can now be adapted, extended, and reused across multiple domains such as Real Estate, Finance, and Accounts Payable.

Whether you're improving HR data governance, validating reporting structures, or ensuring stronger cross-functional decision intelligence, these AI-generated rules provide a solid foundation for any data-driven organization, as explained in a previous article.

To get started, I asked NDI, with some detailed instructions as system prompt.

Generate a comprehensive set of data quality rules for the current dataset, with a strong focus on data integrity, validity, and completeness.

Let’s explore the different rules it created below.

Table of contents

HR Data Quality Rules for NDI Demo Dataset

1. Position reporting structure rules

Rule 1 — Exactly one solid-line manager

Requirement: Every position must have exactly one solid reporting line.
Dataset Field: PositionDirectory."Manager Position ID"
Checks:

  • Flag if Manager Position ID is NULL
  • Flag if Manager Position ID appears more than once for same Position ID (shouldn't happen but we check duplicates)

Rule 2 — Valid dotted-line reporting logic

Dataset Fields:

  • PositionDirectory."Dotted Manager Position ID"
  • PositionDirectory."Manager Position ID" 

Checks:

  • If dotted-line manager exists (non-null), solid-line manager must exist
  • A position cannot report dotted-line to itself
  • A position cannot report dotted-line to its direct solid-line manager (redundancy)

Rule 3 — Manager position must exist in directory

Requirement: Solid-line manager must be a valid Position ID. 

Check:

  • Manager Position ID NOT IN PositionDirectory.Position ID → flag
    Same for dotted-line.

Rule 4 — No circular reporting loops

Check for cases like:

A → B → C → A cycles. Detect with recursive CTE.

A reports to B, and B reports to A

2. Position master data completeness

Rule 5 — Exactly one division

Dataset Field: PositionDirectory.Division
Check:

  • NULL, blank, or multiple distinct divisions per Position ID → flag

Rule 6 — Exactly one department

Dataset Fields: Department, Department ID
Check:

  • Missing department name
  • Missing department ID
  • Mismatch between Department and Department ID (if pattern exists)

Rule 7 — Exactly one Cost Center

Dataset Field: PositionDirectory."Cost Center"
Check:

  • NULL, blank, multi-valued

Rule 8 — Valid Business Area & Business Unit

Dataset Fields:

  • Business Area
  • Business Unit 

Check:

  • Non-null required
  • Business Unit must be consistent with Division (optional rule if hierarchy is known)

Rule 9 — Location completeness

Dataset Fields: City, Region, Country
Check: Missing or inconsistent location attributes.

Rule 10 — Invalid position status

Dataset Field: PositionDirectory.Status
Check: Status must be in allowed set:

  • Active, Inactive, Vacant, On Hold, etc. (You can specify actual valid values.)

3. Position-to-Job alignment rules

Rule 11 — Position must link to an existing job

Dataset Fields & Relationship:

  • PositionDirectory: Job Code
  • JobDirectory: Job Code 

Check:

  • Job Code is NULL
  • Job Code not in JobDirectory → orphan positions

Rule 12 — Job has exactly one Job Family & Functional Area

Dataset Fields:

  • Job Family Code
  • Functional Area Code 

Check:

  • Missing job family
  • Missing functional area
  • Same job appears with multiple job families (should not happen in your dataset)

Rule 13 — Job Family functional alignment (industry norms)

Examples:

  • “Software Engineering” should not belong to “Finance”
  • “Manufacturing Tech” should not belong to “HR Operations”

You can customize norms per your organization.

Rule 14 — Hay Evaluation completeness

Fields must be numeric & non-null:

  • Hay Know-How
  • Hay Problem Solving
  • Hay Accountability
  • Hay Total Points
  • Hay Grade If missing → incomplete job evaluation.

4. Employee-to-Position rules

Rule 15 — Employee must have zero or one position

Dataset Field: EmployeeDirectory."Position ID"

Checks:

  • Missing Position ID → employee unassigned
  • Position ID contains delimiters like , or ; → concatenated values
  • Employee appears in multiple rows with different Position IDs → multi-position assignment

Rule 16 — Employee-to-Position must align to PositionDirectory

Check:

  • Position ID in EmployeeDirectory must exist in PositionDirectory. If not → orphan employee positions.

Rule 17 — No employees assigned to vacant positions

Fields:

  • PositionDirectory.Vacant
  • EmployeeDirectory.Position ID 

Check:

  • If Vacant = “Yes”, Position ID should not have employees assigned.

5. Span of control & managerial rules

Rule 18 — Span of control must match actual direct reports

Fields:

  • PositionDirectory."Span of Control"
  • EmployeeDirectory.Position ID
  • Manager Position ID

Compute actual number of direct reports by counting positions where Manager Position ID = this Position ID. Compare to declared Span of Control. Mismatch > threshold → flag.

Rule 19 — Has Subordinates flag must match reality

Field: Has Subordinates If Has Subordinates = Yes, but actual direct reports = 0 → flag. If Has Subordinates = No, but actual direct reports > 0 → flag.

6. Skill data quality rules

Rule 20 — EmployeeSkill must reference valid Employee

Field: EmployeeSkill.EmployeeId 

Check if it refers to an existing EmployeeDirectory.Employee ID.

Rule 21 — PositionSkill must reference valid osition

Field: PositionSkill.PositionId 

Check if it refers to an existing PositionDirectory.Position ID.

Rule 22 — Skill validity dates must not overlap

For both EmployeeSkill and PositionSkill:

  • For same SkillId + EmployeeId (or PositionId)
  • Overlapping Begda/Endda periods → flag

Rule 23 — Employee Skill vs Position Skill gap

For each employee:

  • Compare EmployeeSkill.SkillId vs PositionSkill.SkillId for the Position Flags:
  1. Missing required skills
  2. Employee proficiency below required level
  3. Employee skill expired (Endda < today)

7. Compensation & FTE rules

Rule 24 — Salary must fall within position salary scale

Fields:

  • EmployeeDirectory.Salary
  • PositionDirectory."Salary Scale (Min)"
  • PositionDirectory."Salary Scale (Max)" 

Checks:

  • Salary < Min
  • Salary > Max

Rule 25 — FTE Consistency

Fields:

  • EmployeeDirectory."Employee FTE"
  • PositionDirectory."Position FTE" 

Check:

  • Employee FTE exceeds Position FTE
  • Employee is part-time but Position Hours Per Week = full-time threshold

8. Diversity, demographic & regulatory rules

Rule 26 — Age & birthdate validity

Fields: Birthdate, Age 

Checks:

  • Age computed from Birthdate must match Age field
  • Implausible ages (<16, >90)

Rule 27 — Mandatory diversity fields completeness

Fields:

  • Gender
  • Nationality
  • Ethnicity
  • Disability 

Checks: Missing or non-standardized values.

9. General data integrity rules

Rule 28 — Duplicate Keys

Check for duplicates in primary identifiers:

  • Employee ID
  • Position ID
  • Job Code
  • SkillId (paired with Employee/Position)

Rule 29 — Text fields containing placeholder values

Look for:

  • “N/A”, “Unknown”, “None”, “--”, “TBD”  and flag where structure requires real values.

Rule 30 — Referential integrity across all relationships

Using dataset relationship map :

  • Employee ↔ EmployeeSkill (Employee ID)
  • Position ↔ Employee (Position ID)
  • Position ↔ Job (Job Code)
  • Position ↔ PositionSkill (Position ID)
  • EmployeeSkill ↔ PositionSkill (SkillId)

Check missing references everywhere.

Final thoughts

Data quality is no longer a back-office hygiene exercise, it’s a strategic enabler. The 30 rules outlined here highlight just how interconnected HR data truly is across positions, jobs, employees, skills, compensation, and organizational structure. When even one link breaks, it cascades into flawed reporting, misaligned decisions, and unreliable insights.

What’s clear is this:

  • High-quality HR data strengthens workforce planning and organizational design.
  • Valid, complete, and reliable data fuels AI-driven decision intelligence.
  • Clean datasets create trust. Trust that leaders depend on to act with confidence.

With modern platforms like Nakisa Decision Intelligence, organizations can move from reactive data cleanup to proactive data governance, with AI continuously monitoring integrity, surfacing anomalies, and providing contextual recommendations.

The goal isn't perfection, it’s continuous clarity. And with well-defined rules and AI-powered validation, clarity becomes scalable, repeatable, and transformative.

These 30 data quality rules form a blueprint for stronger HR foundations today and a smarter, more connected enterprise tomorrow.

If you’re a leader, analyst, HRBP, or data-driven strategist, Nakisa Decision Intelligence unlocks organizational insights with a level of clarity and immediacy that simply didn’t exist before. You can request a demo of NDI here, or reach out to your dedicated Client Success Manager to access the preview environment.

I often write about AI and Nakisa innovations. Connect with me on LinkedIn for the latest updates!

Subscribe to our monthly newsletter

You’d also like

Subscribe to our monthly newsletter

Latest resources

Discover how large enterprises can save on time and costs using Nakisa Decision Intelligence's instant validations of workforce data against org design principles and guardrails.

Be the first to know

Subscribe to the Nakisa monthly newsletter.

Try Nakisa Decision Intelligence free . Turn fragmented data into clear, prioritized actions with agentic AI.

HR-Suite-Main

Nakisa HR suite

Nakisa Lease Administration

HR-Suite-Main

Nakisa Workforce Planning

Nakisa Finance
& IWMS