Make your semantic model AI-ready with systematic checks
Copilot Optimization Checklist: 15 Requirements with Automated Validation
Make your semantic model AI-ready with systematic checks
Microsoft Copilot in Power BI is only as good as the semantic model it works with. Feed it a poorly documented model with cryptic names and implicit measures, and you’ll get hallucinated DAX and frustrated users.
This article provides a comprehensive checklist of 15 requirements for Copilot-optimized semantic models — and shows how to validate them automatically.
The Copilot Readiness Problem
When Copilot generates a DAX query or answers a natural language question, it relies on:
- Table and column names to understand structure
- Descriptions to understand meaning
- Relationships to understand connections
- Measures to understand metrics
- AI Instructions to understand business rules
If any of these are missing or unclear, Copilot guesses. And guessing fails.
The Manual Approach (Doesn’t Scale)
Teams try to optimize models by:
- Reading Microsoft documentation
- Manually checking each table/column
- Hoping developers remember the rules
- Discovering issues after users complain
This doesn’t work at enterprise scale.
The Automated Approach
Define requirements as validation rules. Run them in CI/CD. Block deployment if requirements aren’t met.
The 15 Requirements
Category 1: Naming (4 Requirements)
Requirement 1: Business-Friendly Table Names
Tables must use human-readable names, not technical identifiers.
Validation Rule:
- id: COPILOT_001
name: "Table names must be business-friendly"
check:
type: regex_match
target: table.name
pattern: "^[A-Z][a-z]+([A-Z][a-z]+)*$" # PascalCase, no underscores
severity: error
Requirement 2: Business-Friendly Column Names
Columns must use descriptive names without abbreviations.
Validation Rule:
- id: COPILOT_002
name: "Column names must be business-friendly"
check:
type: no_abbreviations
target: column.name
blocked_patterns: ["_id$", "^dt_", "^amt", "^qty", "^num"]
severity: error
Requirement 3: Measure Names Describe Purpose
Measures must clearly describe what they calculate.
Validation Rule:
- id: COPILOT_003
name: "Measure names must be descriptive"
check:
type: regex_not_match
target: measure.name
pattern: "^(Measure|M_|SUM_|AVG_|COUNT_)\\d*"
severity: error
Requirement 4: No Duplicate Names
Measures with overlapping names confuse Copilot.
Validation Rule:
- id: COPILOT_004
name: "No duplicate measure names or synonyms"
check:
type: unique_names
target: measure.name
include_synonyms: true
severity: error
Category 2: Documentation (4 Requirements)
Requirement 5: All Tables Have Descriptions
Every table must have a description explaining its purpose.
Validation Rule:
- id: COPILOT_005
name: "All tables must have descriptions"
check:
type: not_empty
target: table.description
severity: error
Requirement 6: All Columns Have Descriptions
Every column must have a description explaining its meaning.
Validation Rule:
- id: COPILOT_006
name: "All columns must have descriptions"
check:
type: not_empty
target: column.description
severity: error
Requirement 7: All Measures Have Descriptions
Every measure must have a description explaining what it calculates.
Validation Rule:
- id: COPILOT_007
name: "All measures must have descriptions"
check:
type: not_empty
target: measure.description
severity: error
Requirement 8: Descriptions Under 200 Characters
Descriptions must be concise to fit in Copilot’s context window.
Validation Rule:
- id: COPILOT_008
name: "Descriptions must be under 200 characters"
check:
type: max_length
target: "*.description"
max: 200
severity: warning
Category 3: Structure (3 Requirements)
Requirement 9: Star Schema Topology
Fact tables must have one-to-many relationships to dimension tables. No snowflake chains, no many-to-many without bridges.
Validation Rule:
- id: COPILOT_009
name: "Model must follow star schema"
check:
type: star_schema
fact_tables: ["Sales", "Orders", "Transactions"]
dimension_tables: ["Customer", "Product", "Date", "Geography"]
severity: error
Requirement 10: Explicit Relationships
All relationships must be explicitly defined with clear cardinality.
Validation Rule:
- id: COPILOT_010
name: "All relationships must be explicit"
check:
type: relationships_complete
require_cardinality: true
require_cross_filter: true
severity: error
Requirement 11: Data Categories Set
Geographic, date, and URL columns must have appropriate data categories.
Validation Rule:
- id: COPILOT_011
name: "Data categories must be set for typed columns"
check:
type: data_category_required
rules:
- pattern: "city|town"
category: City
- pattern: "country|nation"
category: Country
- pattern: "zip|postal"
category: PostalCode
- pattern: "url|link"
category: WebUrl
severity: warning
Category 4: Measures (2 Requirements)
Requirement 12: No Implicit Measures
All aggregations must be explicit DAX measures, not implicit sums.
Validation Rule:
- id: COPILOT_012
name: "No implicit measures allowed"
check:
type: no_implicit_measures
target: column
flag_summarizable: true
severity: error
Requirement 13: Measures Have Format Strings
All measures must have appropriate format strings.
Validation Rule:
- id: COPILOT_013
name: "All measures must have format strings"
check:
type: not_empty
target: measure.formatString
severity: warning
Category 5: AI Features (2 Requirements)
Requirement 14: Synonyms Defined
Important tables, columns, and measures should have synonyms for natural language queries.
# Good: Multiple ways to ask for the same thing
synonyms:
"Total Revenue": ["Sales", "Revenue", "Income", "Total Sales"]
"Customer Count": ["Number of Customers", "# Customers", "Customers"]
Validation Rule:
- id: COPILOT_014
name: "Key measures should have synonyms"
check:
type: synonyms_defined
target: measure
min_measures_with_synonyms: 0.5 # At least 50% of measures
severity: warning
Requirement 15: AI Instructions Configured
The model should have AI Instructions defining business rules and guardrails.
# Good: Explicit instructions for Copilot
ai_instructions: |
- Revenue calculations exclude returns
- Use Date[Date] for all time intelligence
- Customer segments: Premium (>$10K annual), Standard (others)
- Fiscal year starts April 1
Validation Rule:
- id: COPILOT_015
name: "AI Instructions should be configured"
check:
type: not_empty
target: model.aiInstructions
severity: warning
Running Validation
CLI Command
mdde powerbi validate --copilot ./model.tmdl
Output
COPILOT READINESS VALIDATION
============================
Model: Sales Analytics
Checked: 15 requirements
PASSED (11):
✓ COPILOT_001: Table names must be business-friendly
✓ COPILOT_002: Column names must be business-friendly
✓ COPILOT_003: Measure names must be descriptive
✓ COPILOT_004: No duplicate measure names
✓ COPILOT_005: All tables have descriptions
✓ COPILOT_009: Model follows star schema
✓ COPILOT_010: All relationships explicit
✓ COPILOT_012: No implicit measures
✓ COPILOT_013: All measures have format strings
✓ COPILOT_014: Synonyms defined
✓ COPILOT_015: AI Instructions configured
FAILED (2):
✗ COPILOT_006: All columns must have descriptions
Missing: Product[SKU], Customer[LegacyID], Date[FiscalWeek]
✗ COPILOT_011: Data categories must be set
Missing: Customer[City] (should be City)
Customer[Country] (should be Country)
WARNINGS (2):
⚠ COPILOT_007: 3 measures missing descriptions
- [Margin %]
- [Units Sold]
- [Discount Amount]
⚠ COPILOT_008: 2 descriptions over 200 characters
- Table "Orders" (245 chars)
- Measure "Complex Calculation" (312 chars)
SCORE: 73/100
STATUS: NOT READY FOR COPILOT
Fix errors before deployment
CI/CD Integration
GitHub Actions
# .github/workflows/copilot-validation.yml
name: Copilot Readiness Check
on:
pull_request:
paths:
- 'models/**/*.tmdl'
- 'models/**/*.yaml'
jobs:
validate:
runs-on: ubuntu-latest
steps:
- uses: actions/checkout@v4
- name: Install MDDE
run: pip install mdde
- name: Validate Copilot Readiness
run: |
mdde powerbi validate --copilot ./models/ \
--fail-on-error \
--output copilot-report.html
- name: Upload Report
uses: actions/upload-artifact@v4
with:
name: copilot-readiness-report
path: copilot-report.html
- name: Comment on PR
if: failure()
uses: actions/github-script@v6
with:
script: |
github.rest.issues.createComment({
issue_number: context.issue.number,
owner: context.repo.owner,
repo: context.repo.repo,
body: '❌ Copilot readiness validation failed. See report for details.'
})
Azure DevOps
# azure-pipelines.yml
trigger:
paths:
include:
- models/*
steps:
- task: UsePythonVersion@0
inputs:
versionSpec: '3.11'
- script: pip install mdde
displayName: 'Install MDDE'
- script: |
mdde powerbi validate --copilot ./models/ \
--fail-on-error \
--output $(Build.ArtifactStagingDirectory)/copilot-report.html
displayName: 'Validate Copilot Readiness'
- task: PublishBuildArtifacts@1
inputs:
pathToPublish: $(Build.ArtifactStagingDirectory)
artifactName: CopilotReport
Auto-Fix Capability
Some requirements can be automatically fixed:
mdde powerbi fix --copilot ./model.tmdl --dry-run
Fixable Issues:
Non-Fixable Issues (require human input):
# Preview fixes
mdde powerbi fix --copilot ./model.tmdl --dry-run
# Apply fixes
mdde powerbi fix --copilot ./model.tmdl --apply
# Fix only specific categories
mdde powerbi fix --copilot ./model.tmdl --only naming,formatting
Scoring System
Each requirement contributes to an overall score:
Score Interpretation
Generating Reports
Markdown Report
mdde powerbi validate --copilot ./model.tmdl --output report.md
HTML Report (Visual)
mdde powerbi validate --copilot ./model.tmdl --output report.html
JSON Report (Programmatic)
mdde powerbi validate --copilot ./model.tmdl --output report.json
Integration with Model-Driven Development
When using MDDE’s metadata-driven approach, many requirements are satisfied by default:
MDDE Configuration:
# mdde.yaml
powerbi:
copilot:
require_descriptions: true
max_description_length: 200
auto_synonyms_from_glossary: true
generate_ai_instructions: true
Key Takeaways
- 15 requirements cover Copilot readiness — Naming, documentation, structure, measures, AI features.
- Automate validation — Run checks in CI/CD, not manually.
- Fail fast — Block deployment if requirements aren’t met.
- Score provides visibility — Track readiness over time.
- Auto-fix where possible — Reduce manual remediation effort.
- Model-driven development helps — Many requirements satisfied by default.
Copilot readiness isn’t a one-time effort. It’s a continuous quality gate that ensures your semantic models remain AI-ready as they evolve.
This is part of the Power BI + MDDE integration series.
Previous: “ From Entity to Measure: Automated DAX Generation ” Next: “Semantic Model DevOps: CI/CD for Power BI” (coming soon)