Power Platform Tenant State Scenarios
Budget: ₹12,500 – ₹37,500 INR
Phase 2 of my Governance IQ dataset is ready to move from raw research to repeatable insight, and I need your help. The purpose of this phase is two-fold:
1. Craft a set of realistic Power Platform tenant-environment scenarios that accurately mirror what admins see in the Power Platform Admin Center. Each scenario must surface the governance signals I care about—apps and flows, connectors and agents, users and licensing, environments, DLP policies, usage patterns, and any governance risks.
2. From those scenarios, design a robust system prompt for a large-language model so that, when the model is given a short scenario description, it returns a complete tenant-state JSON document ready for analytics or downstream automation.
Scope of work
• Write clear, narrative-style scenarios that cover multiple environment types (e.g., dev, prod, training) and reflect real-world governance challenges.
• Define or refine the JSON schema (if adjustments are needed) so every required field—object counts, policy settings, license allocations, usage and entitlement consumption—has an unambiguous place.
• Build the LLM prompt (few-shot or instruction-tuned style) so the model reliably converts each scenario into valid JSON without hallucination.
• Provide sample interactions: the plain-English scenario you feed the model, the resulting JSON, and a brief commentary on why the output meets governance expectations.
Acceptance criteria
• At least eight distinct tenant-environment scenarios delivered in a single document.
• A system prompt (and any supporting examples) that produces syntactically correct JSON matching the agreed schema in ≥95 % of test runs.
• Example JSON files validated by a linter and by round-tripping through the Power Platform Admin Center export where feasible.
I can supply the Phase 1 baseline data and the preliminary JSON schema as soon as we start. Familiarity with Power Platform governance concepts, PowerShell Admin Center cmdlets, and prompt engineering for OpenAI or Azure OpenAI will make the work faster, but a sharp eye for detail and structured data quality is what matters most.
1. Craft a set of realistic Power Platform tenant-environment scenarios that accurately mirror what admins see in the Power Platform Admin Center. Each scenario must surface the governance signals I care about—apps and flows, connectors and agents, users and licensing, environments, DLP policies, usage patterns, and any governance risks.
2. From those scenarios, design a robust system prompt for a large-language model so that, when the model is given a short scenario description, it returns a complete tenant-state JSON document ready for analytics or downstream automation.
Scope of work
• Write clear, narrative-style scenarios that cover multiple environment types (e.g., dev, prod, training) and reflect real-world governance challenges.
• Define or refine the JSON schema (if adjustments are needed) so every required field—object counts, policy settings, license allocations, usage and entitlement consumption—has an unambiguous place.
• Build the LLM prompt (few-shot or instruction-tuned style) so the model reliably converts each scenario into valid JSON without hallucination.
• Provide sample interactions: the plain-English scenario you feed the model, the resulting JSON, and a brief commentary on why the output meets governance expectations.
Acceptance criteria
• At least eight distinct tenant-environment scenarios delivered in a single document.
• A system prompt (and any supporting examples) that produces syntactically correct JSON matching the agreed schema in ≥95 % of test runs.
• Example JSON files validated by a linter and by round-tripping through the Power Platform Admin Center export where feasible.
I can supply the Phase 1 baseline data and the preliminary JSON schema as soon as we start. Familiarity with Power Platform governance concepts, PowerShell Admin Center cmdlets, and prompt engineering for OpenAI or Azure OpenAI will make the work faster, but a sharp eye for detail and structured data quality is what matters most.