Skip to content

Latest commit

 

History

History
37 lines (28 loc) · 4.22 KB

File metadata and controls

37 lines (28 loc) · 4.22 KB

GitHub Copilot Chat Assistant

This document organizes repository patterns into a concise GitHub-friendly matrix using the provided seven-pattern AI capability taxonomy.

Canonical seven AI patterns

  1. Recognition (perception, classification, extraction)
  2. Conversational / Human Interaction (dialogue, assistants)
  3. Predictive Analytics & Decision Support (forecasting, recommendations)
  4. Generative Content (text, image, code generation)
  5. Optimization & Automation (planning, orchestration, process automation)
  6. Anomaly Detection & Security (threats, robustness)
  7. Knowledge Retrieval & Reasoning (RAG, grounding, long-context reasoning)

Repository pattern mapping

Pattern Why (evidence from this repo) AI tools that can enhance or automate work Mapped AI pattern(s)
Agent-based advisory / Domain-specific copilots The PM Risk Assessor prototype and agent materials (agents/pm-risk-assessor, ai-agent-q1-2026) show an agent encoding PM rules, stage-aware guidance, and scenario tests. Azure AI Foundry; Copilot Studio; ChatGPT Enterprise; RAG stacks such as Azure Cognitive Search, LlamaIndex, and LangChain; agent QA frameworks Conversational / Human Interaction; Knowledge Retrieval & Reasoning; Generative Content
Stage-aware decision support (adoption lifecycle) README and ai-agent-q1-2026 emphasize stage confusion, evidence ladders, and stage-aware guidance for pilot-to-scale transitions. Workflow automation such as Power Automate and Azure Logic Apps; decision-support agents; dashboards such as Power BI; policy engines Predictive Analytics & Decision Support; Optimization & Automation; Knowledge Retrieval & Reasoning
Governance, compliance & risk mapping cyber-ai-profile, NIST AI RMF references, and responsible-ai-usage.md stress control mappings, GRC integration, and NIST alignment. Policy-as-code such as OPA; compliance scanners; Microsoft Responsible AI toolkits; SBOM tools; GitHub Actions gating Anomaly Detection & Security; Knowledge Retrieval & Reasoning; Recognition (classification of controls and risks)
Human-in-the-loop review & accountability The repository repeatedly stresses human responsibility, review, and acceptance tests; human checkpoints are built into agent validation. Annotation UIs such as Label Studio; review queues in Teams or Jira; explainability tools such as InterpretML; HITL workflow automation Conversational / Human Interaction; Knowledge Retrieval & Reasoning; Optimization & Automation
Knowledge-grounding & cross-ecosystem integration References to SharePoint, OneDrive, prompt libraries, and structured artifacts imply reuse of knowledge-management assets to ground agents (prompts/, frameworks/). Vector databases such as Pinecone; Azure Cognitive Search; connectors to SharePoint or Confluence; RAG pipelines using LangChain or LlamaIndex Knowledge Retrieval & Reasoning; Recognition (entity extraction); Conversational / Human Interaction
Reporting, measurement & evidence translation README suggests Power BI risk heatmaps, CSV exports, Evidence Ladder, and notebooks for benchmarking. ETL tooling such as Azure Data Factory; Power BI dashboards; Jupyter notebooks; AutoML; anomaly detection Predictive Analytics & Decision Support; Optimization & Automation; Recognition (metric classification and labeling)
Prompt engineering & reproducible enablement artifacts The repository includes a prompt library, templates, examples, and suggested artifact workflows for reuse and governance. PromptLayer; prompt versioning in Git; CI for prompts using GitHub Actions; prompt testing frameworks Generative Content; Conversational / Human Interaction; Knowledge Retrieval & Reasoning

Notes

  • This version keeps the provided canonical taxonomy unchanged.
  • The table wording is tightened for readability in GitHub Markdown while preserving the original intent.
  • File and folder references remain in inline code for easier scanning in repository documentation.

Possible next edits

  • Re-map the table to a different canonical seven-pattern taxonomy.
  • Add explicit file and line references for each evidence statement.
  • Split the matrix into "repository evidence" and "recommended tooling" sections for presentation use.