Course Details
A 2-day course, 6 hours per day (12 hours in total), delivered as a hands-on workshop with a capstone project. Intermediate to Advanced level. Learners take home their own Custom GPT, automation workflow templates, and a security, cost and PDPA checklist.
Day 1: Building Custom GPTs, Specialized AI Assistants
Section 1: Overview of Custom GPTs and AI Assistants in the Organization
- What a Custom GPT is, how it differs from ordinary prompting and Assistants, and when to build one instead of repeating prompts
- Components of a Custom GPT: Instructions, Knowledge, Conversation Starters and Actions, with department use cases
- Account requirements (ChatGPT Plus / Team / Enterprise) and organizational usage scope
Section 2: Building Your First Custom GPT with GPT Builder
- Using GPT Builder in both conversational (Create) and manual (Configure) modes
- Setting name, description, profile image and conversation starters, and enabling capabilities (Web Search, Image Generation, Code Interpreter)
- Testing the GPT in Preview and iterating
- Workshop: build a Custom GPT for one of the learner's tasks, such as an email-reply or document-summary assistant
Section 3: Instructions and Knowledge for Accurate, Consistent Answers
- A good instruction structure: role, scope, steps, output format and prohibitions
- Setting persona and tone of voice to match brand and organizational standards
- Uploading knowledge (documents, manuals, policies) and document retrieval
- The scope and limits of knowledge and managing data so the GPT answers from real information
Section 4: GPT Actions Connecting the Custom GPT to External Systems
- What GPT Actions are and how they let the GPT call APIs and real systems
- Basic OpenAPI schema structure defining what the GPT can call
- Setting authentication (API key, OAuth) and connection security, with action examples
- Workshop: connect one GPT Action so the Custom GPT pulls data from an external service to answer users
Section 5: Testing, Publishing and Standardizing GPT Usage
- Testing the quality and safety of a Custom GPT before publishing
- Publishing levels: private, team/organization (Workspace) and public (GPT Store)
- Sharing and versioning the GPT for team use, and cautions on organizational data and PDPA
- Workshop: publish the Custom GPT for the team to trial and collect feedback for improvement
Day 2: AI Automation, Making AI Work End to End
Section 6: AI Workflow Automation Concepts and Choosing Worthwhile Tasks
- What automation is and how AI changes traditional automation
- Assessing routine work: what to automate and what always needs a human (human-in-the-loop)
- Calculating worth (time, cost, errors) and the architecture overview: trigger, action, node
Section 7: Automation Platforms - n8n, Make and Zapier
- Differences between n8n, Make and Zapier and choosing what fits the organization
- Trigger vs action, nodes and passing data between steps
- Setting credentials and connecting services securely, and the benefits of self-hosted n8n
Section 8: Building Your First Workflow with AI in the Steps
- A good workflow structure: take input, process with AI, and send output
- Calling an AI model in the workflow (OpenAI node) to summarize, classify or generate content
- Sending prompts and getting structured output to use in the next step
- Workshop: build a workflow that takes a message, has AI summarize and classify it, then sends the result out
Section 9: Connecting Real Data with Connectors and MCP
- Connecting common services: Google Sheets, Gmail, Google Drive, LINE and Webhook
- The MCP (Model Context Protocol) concept and its role in connecting AI to real systems
- Reading and writing to databases (MySQL / PostgreSQL) and receiving events via webhook
- Workshop: connect a workflow to pull data from a Google Sheet, process with AI, and return results via Gmail or LINE
Section 10: Multi-step Automation and Orchestration
- Chaining multiple steps and branching (conditional logic)
- Error handling and retry for a stable system
- Scheduling daily, weekly and monthly, and setting up a morning brief or automated report
- Integrating the Day 1 Custom GPT into the workflow to work together
Section 11: Security, Cost, PDPA and Capstone Project
- Managing API keys and credentials securely on least privilege, and controlling AI usage cost
- What must not be sent into AI systems and protecting personal data under PDPA
- Capstone project: assemble a Custom GPT into an automation workflow as your own end-to-end automated system, then present and get feedback
- A first-30-days adoption plan and extending toward enterprise AI agents