Bloom-aligned objectives

• Understand: what “agentic workflow” means in a business context (without coding)
• Apply: workflow decomposition into steps, decisions, validations, and handoffs
• Analyze: which steps should be AI-assisted vs fully automated vs kept human-controlled

Topics

• Aligning to AB-730 workflow outcomes (business content, data analysis, meetings/collaboration, defining Copilot’s role in the workflow)
• Workflow decomposition patterns
  • Trigger → intake → enrich → decide → act → validate → log → notify
  • Exception-first thinking: what can go wrong and how the workflow should respond
• Human-in-the-loop placement
  • Approvals for sensitive decisions
  • “Draft vs send” separation for external communications
• Defining “AI role” per workflow
  • Drafter, analyst, meeting scribe, policy assistant, checker, coordinator (and when each is risky)

Activity (30 min): “My Top 3 automations”

Participants shortlist 3 recurring workflows and score them by frequency, effort, risk, and automability.

Micro-lab 1 (45 min): “Workflow blueprint”

Pick 1 workflow and produce:

• Before map (current steps, pain points)
• After map (agentic steps, automations, approvals, validations)
• Definition of Done + success metrics (time saved, error reduction)

Bloom-aligned objectives

• Apply: structured prompting to produce SOPs, checklists, templates, and decision memos
• Analyze: gaps, assumptions, missing inputs, and verification needs
• Create: a reusable “prompt pack” for a chosen workflow

Topics

• Prompting for orchestration (not just content)
  • Clarifying-interview prompts (ask questions before generating)
  • Plan-first prompts (steps + dependencies + pitfalls)
  • Output contracts (exact sections, fields, tables, owners)
• Using Copilot Prompt Gallery patterns to standardize prompting across the org
• Grounding prompts in enterprise context (reference files/threads where permitted; avoid sensitive leakage)

Lab 2A (60 min): “SOP generator”

Input: a messy description of a process (e.g., weekly status update, onboarding checklist, invoice follow-up).
Output: SOP with:

• steps, owner, inputs/outputs, decision rules, escalation rules and validation checklist

Lab 2B (45 min): “Prompt pack builder”

Create 8 reusable prompts for the chosen workflow:

• intake prompt, summarizer prompt, decision prompt, draft prompt, validator prompt, exception prompt, escalation prompt, final output prompt

Deliverable: SOP + prompt pack v1.

Bloom-aligned objectives

• Apply: Copilot in Power Automate to create and modify cloud flows using natural language
• Analyze: triggers, conditions, approvals, error handling, and auditability
• Create: 2 production-style personal flows with validation and exception paths

Topics

• Copilot in Power Automate: where it helps (flow creation, refinement, troubleshooting)
• Core building blocks (business-user depth)
  • Triggers (email received, form submitted, file created)
  • Actions (create task, post message, update row, create document)
  • Conditions, parallel branches, approvals, retries/timeouts
• Operational readiness for personal automations
  • Naming, documentation, run history review
  • Notifications and failure handling
  • Minimal logging (what happened, when, and why)

Lab 3A (75 min): “Inbox-to-task automation”

Build a flow that:

• detects specific inbound emails
• extracts key fields (subject, sender, date, request type)
• creates a Planner/To Do task
• posts a Teams notification
• routes exceptions to a manual review step

Lab 3B (60 min): “Document approval workflow”

Build a flow that:

• triggers on a file added to a SharePoint/OneDrive folder
• requests approval
• updates status metadata
• notifies requester and logs outcome

Deliverable: 2 flows with a short runbook (how to test, how to handle failures).

Bloom-aligned objectives

• Apply: AI assistance to convert collaboration artifacts into executable plans
• Analyze: ownership gaps, ambiguous decisions, and missing follow-ups
• Create: a repeatable “meeting → action tracker → stakeholder update” workflow

Topics

• AB-730 alignment: managing meetings and collaboration using Microsoft 365 Copilot patterns
• Meeting-to-execution pipeline
  • agenda creation from threads
  • minutes, decisions, action items
  • follow-up emails and task creation
• Automation chaining
  • Flow triggered from a meeting output artifact (notes/transcript summary)
  • Auto-create tasks + reminders + escalation

Lab 4A (75 min): “Meeting recap to action tracker”

Input: sanitized meeting transcript excerpt + chat thread summary.
Outputs:

• decisions + action items (owners, due dates)
• risk/dependency list
• stakeholder follow-up message
• optional flow step to create tasks automatically

Lab 4B (30 min): “Weekly status pack automation”

Create a structured update template and a flow-assisted method to compile:

• workstream updates
• RAG statuses
• top risks + asks

Deliverable: meeting-to-execution template + an automation-assisted workflow.

Bloom-aligned objectives

• Understand: agent types (declarative vs more advanced) and when to use which
• Apply: build an agent that uses knowledge sources and executes an agent flow
• Create: one personal “work agent” and one agent flow that automates a process

Topics

• Agents for Microsoft 365 Copilot: why and when to build an agent
• Build with Microsoft 365 Copilot Agent Builder (declarative agents)
  • describe agent, configure behavior, add knowledge sources
• Copilot Studio fundamentals and agent flows
  • Copilot Studio as low-code agent/flow platform
  • Agent flows overview and benefits (consistent execution, process visibility)
  • Connectors for integrating business apps (Power Platform connectors concept)
• Decision guide: when Agent Builder is enough vs when Copilot Studio is required

Lab 5A (75 min): “Build a personal work agent (Agent Builder)”

Create an agent for one scenario:

• Weekly project status pack generator, or
• Customer email triage assistant, or
• Onboarding checklist coordinator

Deliverable: agent definition + knowledge sources + test transcript (3 test cases).

Lab 5B (60 min): “Add an agent flow (Copilot Studio)”

Build an agent flow that:

• collects required inputs (structured fields)
• runs steps in order (create/update artifacts, notify, log)
• handles an exception route

Deliverable: working agent flow + run history evidence.

Bloom-aligned objectives

• Understand: common failure modes and attack patterns affecting AI-assisted automation
• Evaluate: workflows for oversharing, consent risk, and prompt-injection exposure
• Create: a personal guardrail checklist and a “safe automation” standard for your team

Topics

• Governance and security best practices for Copilot Studio / agents (admin and maker perspectives)
• Identity and consent risk in automations
  • how app consent works and why it matters for automation-connected tools
  • real-world OAuth-consent phishing patterns observed in Copilot Studio contexts (what to watch for, how to reduce risk)
• Indirect prompt injection in enterprise workflows (documents, emails, web content)
  • why it happens and practical mitigations
• Practical guardrails (business-user friendly)
  • least privilege + minimal connector permissions
  • separate “draft” from “send/execute”
  • mandatory validation steps for external outputs and KPI reporting
  • approval gates for high-risk actions

Workshop (75 min): “My safe automation standard”

Create:

• “Never automate without” checklist (validation, exception handling, owner, rollback)
• “Never paste / never process” sensitive data list
• A team-ready escalation path (what to do when an automation misbehaves)

Final simulation (45 min): “Agentic workday”

Teams complete a timed sequence:

1. Turn a request email into a structured plan
2. Generate an approved response draft
3. Create tasks and reminders automatically
4. Produce a weekly status summary from inputs
5. Apply guardrails (consent risk check + injection-safe prompting + validation steps)