Generative & Agentic AI on AWS

AnyCompany Leader — Workshop

2 days · From prompt mastery to your first AI agent · No coding required

Workshop Journey

🎯

Day 1: Prompt Mastery & Smart Model Selection

Choose the right model, write better prompts, verify AI outputs. Ends with a Prompt Competition judged by AI.

🤖

Day 2: Agentic AI — From Concept to Your First Agent

Understand autonomous agents, then build one for your team's process in Claude Cowork. Ends with Agent Design Showcase.

Who Is This For?

👔

Finance Leadership

Heads of CoE, Senior Managers, and CFO Office decision-makers evaluating where AI fits in their team's roadmap

💼

Procurement & Sourcing

Cluster & Category Heads, Procurement Managers — daily contracts, vendor scorecards, and sourcing decisions

📊

Finance Operations

Controllership, FP&A, Reporting, and Treasury teams using Claude or AnyCompany GPT for narrative and analysis work

🛡️

Audit, Tax & Specialists

Internal Audit & Integrity, Tax, and Finance Data Solutions — defensibility, research, and bridge-to-IT roles

What You'll Take Home

📝

Reusable Prompt Template

Persona + Chain-of-Thought + structured output, applied to variance commentary, vendor scorecards, audit memos, or tax positions — your choice.

🤖

Working Invoice Agent

Live Claude Cowork project with Project Instructions, a saved Skill, scheduled run, and exception-handling on a deliberately broken invoice.

📊

Model Selection Guide

Haiku / Sonnet / Opus + Effort levels, with a finance cost ladder ($15/mo to $1,200/mo for the same task — and how to control which side you're on).

✅

Governance & Verification Checklist

Numbers, regulations, names, dates — what to verify before the output reaches a reviewer or a regulator. Maps to your existing SOX / ITGC controls.

📋

Agentic Cheat Sheet

One-page reference: Project Instructions · Skills · Scheduled Tasks · Plugins / MCP — when to use what, with finance scenarios.

🗺️

Agent Design Canvas

Strategic one-pager for your next automation — the brief you hand to your tech team next week.

Prerequisites

  • No coding experience required
  • Basic familiarity with AI chat tools (you already use Claude or AnyCompany GPT in your daily work)
  • Bring a laptop with access to Claude Cowork — the workshop's primary tool
  • Bring one repetitive process from your team that you'd like to automate (e.g., variance commentary, vendor scorecard, audit memo, tax research, period-close narrative)
⚠️ Security Notice: Do not include any confidential information, personally identifiable information (PII), or sensitive data in your prompts. All sample data in this workshop is synthetic.

Day 1: Prompt Mastery & Smart Model Selection

Use AI better today — choose the right model, write better prompts, verify the output.

Topics Covered (M1 — M9 + Exercise)

• M1 · AI hierarchy, foundation models, AWS GenAI toolkit
• M2 · 28 finance use cases by function — when AI wins, when it doesn't
• M3–M5 · How LLMs read, understand, and generate text
• M6 · The 1-cent-or-1-dollar model choice
• M7 · Governance & Trust — verification, Bedrock Guardrails, L1 / L2 / L3
• M8 · 4 pillars of prompts + Chain-of-Thought + persona + structured output
• M9 · RAG — grounding AI in your documents (the Tax persona's flagship)
• Exercise · Executive Decision Brief + AI Judge competition
• PII redaction, hallucination detection, audit trail patterns
• When NOT to use AI (Python vs LLM)

⚡ Day 1 Modules — Numbered for the Class Flow

🏆 Day 1 Exercise

Day 2: Agentic AI — From Concept to Your First Agent

Understand autonomous AI, then build one for your team's process.

Topics Covered

• Chatbots vs autonomous agents
• The agentic loop: Observe → Plan → Act → Reflect
• 4 levels of AI autonomy (L1–L4)
• 4 workflow patterns: Chaining, Parallelization, Routing, Orchestration
• The Cowork stack: Project Instructions · Skills · Scheduled Tasks · Plugins / MCP
• Skills deep-dive: anatomy, lifecycle, Skill vs ad-hoc prompt
• Memory & conversations — context budget & recurring token cost
• Plugins & Connectors — the 3-layer auth model (install · activate · authenticate)
• Build your first invoice agent (with deliberate-error iteration)
• Agent Design Canvas — strategic plan for your team's next automation
• Python vs LLM: when to script, when to use AI

⚡ Interactive Explainers (M1–M7)

🤖 Day 2 Exercises

💡 What you'll take home from Day 2: A working agent project in Claude Cowork, an agentic workflow cheat sheet, and an Agent Design Canvas ready to share with your tech team.

Resources & Reference

📋 Cheat Sheets & Frameworks

⚡ All Interactive Explainers