Generative & Agentic AI on AWS

AnyCompany Finance — Workshop

3 days · From AI fundamentals to automated workflows · No coding required

Workshop Journey

📚

Day 1: GenAI Essentials

What can AI do? Foundation models, use cases for finance, responsible AI, and your first GenAI playbook.

💬

Day 2: Prompt Engineering

How do I talk to AI? Master prompt techniques, build reusable templates, and create 6 finance tools with Kiro.

🤖

Day 3: Agentic AI

How do I make AI work on its own? Workflow patterns, skills, hooks, and design an agent for your team.

Who Is This For?

💼

Finance Operations

Finance analysts, accounts payable/receivable teams exploring AI-powered productivity tools

📊

Risk & Compliance

Compliance officers and risk managers evaluating AI for regulatory reporting and fraud detection

🏢

Business Management

Technical managers and team leads looking to prototype internal tools rapidly

🤝

Anyone Curious

Anyone interested in using AI to build software without coding

Prerequisites

  • No coding experience required — Kiro writes all the code
  • An internet connection
  • Access to the AWS workshop environment
⚠️ 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. When the event ends, all sandbox resources are automatically destroyed.

Day 1: Generative AI Essentials

Understanding what AI can do — and where it fits in finance.

Topics Covered

• How foundation models work
• How LLMs predict the next token
• Amazon Bedrock services & models
• Temperature, Top-k, Top-p parameters
• Tokenization: how AI processes text
• Model pricing & cost per assessment
• Why models differ in speed, quality & cost
• When GenAI is (and isn't) the right fit
• Finance use cases: PTP, RTR, Risk, Payments
• Prompt design basics & best practices
• Responsible AI: bias, PII, prompt injection
• Model selection criteria
• Bedrock Guardrails configuration

🧪 Labs

⚡ Interactive Explainers — How LLMs Work Under the Hood

Visual, interactive walkthroughs of the three core stages of how AI processes text. Use these as reference during the day's discussions.

📚 Reference Materials

💡 Day 1 BuilderLabs

BuilderLab 1: Prompt Engineering in Amazon Bedrock Playground (customized prompts)
BuilderLab 2: Responsible AI with Bedrock Guardrails
BuilderLab 3 (Capstone): Create a GenAI Implementation Playbook (5 documents)

Day 2: Prompt Engineering Workshop

Master prompt techniques and build 6 practical finance tools with Kiro.

Topics Covered

• 4 pillars: Clarity, Context, Role, Output
• 4 types of context: Domain, Data, Situational, Constraints
• Chain-of-Thought: Zero-Shot, Few-Shot, Step-Back
• Self-Consistency for high-stakes decisions
• Persona prompting & multi-agent framing
• Structured outputs (JSON) & RAG grounding
• Model-specific tuning & parameters
• Evaluating prompts: rubrics & LLM-as-Judge
• Bedrock Prompt Management & Optimization
• Long conversation management & circuit breakers
• Common prompt mistakes & how to avoid them
• From prompt templates to automated skills

📽️ Presentation

🛠️ Kiro Hands-On Labs

✏️ Prompt Engineering Labs

⚡ Interactive Explainers

Day 3: Agentic AI & Workflow Automation

Make AI work on its own — design automated workflows for your team.

Topics Covered

• Chatbot vs autonomous agents
• The agentic loop: Observe → Plan → Act → Reflect
• 4 levels of AI autonomy (L1–L4)
• Workflow patterns: Chaining, Parallelization, Routing, Orchestration
• Kiro stack: Steering + Skills + Hooks
• Skills vs MCP: instructions vs connections
• Converting templates to SKILL.md
• SKILL.md best practices
• Agent Design Canvas methodology
• Guardrails at every layer

📽️ Presentation

🤖 Agent Design Labs

⚡ Interactive Explainers

💡 What you'll take home from Day 3: A SKILL.md file, a steering file, a hook configuration, and an Agent Design Canvas — all ready to share with your tech team.

Resources & Reference

📦 Downloads

📋 Workshop Deliverables

By the end of the 3 days, each participant will have created these artifacts:

DeliverableDayLabDescription
GenAI Playbook1BL35-document implementation playbook: use case selection, model rubric, performance plan, evaluation framework, deployment plan
6 Finance Tools2L1-5Invoice processor, transaction dashboard, finance presentation, project planner, compliance report — all built with Kiro
Prompt Template2L6/L7Production-ready template with {{variables}} for merchant risk (GREEN/AMBER/RED) or credit narrative (APPROVE/CONDITIONS/DECLINE)
SKILL.md3L8Portable, auto-activating skill file — converts your Day 2 template into a reusable Kiro skill
Steering File3L8Always-on rules for your workspace: currency defaults, PII handling, brand voice, domain context
Hook Config3L8Auto-trigger that runs your skill when new data files arrive — no manual prompting needed
Fraud Detection Skill3L9MCP-connected skill that queries a merchant database and detects fraud patterns
Agent Design Canvas3L10One-page agent design: role, workflow steps, data sources, guardrails, success metrics, business impact

🔧 Tools & Platforms

ToolDayWhat You'll Do With It
Amazon Bedrock Playground1Test prompts across models (Nova Pro, Claude), configure guardrails, compare outputs
Kiro IDE2-3Build finance tools, create skills, configure hooks, connect to databases via MCP
Bedrock Prompt Management2 (demo)Store, version, and share prompt templates across your team (instructor demo)
MCP (Model Context Protocol)3Connect Kiro to a SQLite database — query merchant data in plain English

📖 Financial Terms Glossary

Key financial terms, acronyms, and metrics used throughout the workshop.

💼 Finance Operations

PTPProcure-to-Pay — purchasing through to supplier payment
RTRRecord-to-Report — journal entries through to financial reports
GMVGross Merchandise Value — total transaction value before deductions
GSTGoods & Services Tax — Singapore 9% (since Jan 2024)

📊 Credit & Lending

DSCRDebt Service Coverage Ratio — cash flow ÷ debt payments (min 1.25x)
DTIDebt-to-Income Ratio — monthly debt ÷ income (max 30-40%)
BNPLBuy Now, Pay Later — AnyCompany's version is "PayLater"

🛡️ Risk & Compliance

KYCKnow Your Customer — identity verification before onboarding
AMLAnti-Money Laundering — preventing illegal fund disguising
MCCMerchant Category Code — 4-digit code classifying merchant type

🌏 SEA Regulators

MASMonetary Authority of Singapore
BNMBank Negara Malaysia
OJKOtoritas Jasa Keuangan (Indonesia)
BOTBank of Thailand

💳 Payments

ChargebackCustomer disputes a transaction — payment reversed to customer
VelocityTransaction rate over time — spikes signal potential fraud
Card TestingFraud technique: many small transactions to test stolen cards

🤖 AI Terms

RAGRetrieval-Augmented Generation — AI answers from your documents
CoTChain-of-Thought — AI shows reasoning step by step
MCPModel Context Protocol — connects AI to databases and APIs
SKILL.mdReusable AI skill file — portable between Kiro and other tools
📖 Full Glossary: A comprehensive glossary with 40+ terms, detailed explanations, and currency reference is available in the instructor materials (reference/financial-terms-glossary.md).

🔑 Key Concepts

💡 Persona Prompting

Throughout this workshop, prompts start with a role assignment like "You are a Senior Accounts Payable Analyst...". This is persona prompting — one of the most effective techniques for getting higher-quality results from AI. By telling the AI what role to play, it adjusts its vocabulary, reasoning, and level of detail to match that expertise.

⚡ All Interactive Explainers

Visual, interactive walkthroughs — use as reference anytime.

📚 Reference Pages

⚠️ Reminder: All sample data in this workshop is synthetic. Do not include any confidential information, PII, or sensitive data in your prompts. When the event ends, all sandbox resources are automatically destroyed.