AI Specialization: Gems, Custom GPTs & Claude Skills
Overview
This 4-hour build workshop moves beyond using AI to creating your own specialised AI assistants. You will learn how to design, configure, and deploy custom AI agents across three leading platforms: Gemini Gems (Google), Custom GPTs (OpenAI), and Claude Skills/Projects (Anthropic). By the end of the session, you will have built at least one working AI specialist of your own — trained on your context, instructions, and knowledge — ready to use in your daily workflow.
Skills Acquired:
- System prompt engineering for AI agent design
- Custom GPT building with knowledge upload and capabilities
- AI persona design: tone, expertise, and boundary setting
- Knowledge base preparation for AI retrieval
- Cross-platform AI specialisation strategy
- Gemini Gem creation and Google Workspace integration
- Claude Project (Skills) configuration and knowledge loading
- Iterative testing and refinement of custom AI agents
- Sharing and deploying AI assistants within a team
- Custom GPT Actions introduction and API awareness
This course is included in the AI Annual Learning Pass.
Duration:
1/2 Day
Who Should Take This Course
Prerequisites
- Active accounts on Gemini Advanced, ChatGPT Plus/Team, and Claude Pro (at least one)
- A specific role, task, or domain you want your AI specialist to master (come with an idea)
- Basic comfort writing structured instructions (bullet points, numbered lists)
Why You Should Take This Course
By the end of this course, learners will be able to:
- Design and build a Gemini Gem tailored to a specific role or task
- Create and publish a Custom GPT with a system prompt, persona, and uploaded knowledge
- Configure a Claude Project (Skill) with custom instructions and contextual knowledge
- Write effective system prompts that define AI behaviour, tone, and boundaries
- Evaluate and iterate on custom AI agents for quality and consistency
- Understand the use cases, limitations, and best-fit scenarios for each platform’s customisation approach
Course Outline
AI Specialization: Gems, Custom GPTs & Claude Skills
| Module 1: The Anatomy of a Custom AI Agent
40 minutes |
What Makes an AI Specialist?
- System prompts vs. user prompts: the hidden layer that controls behaviour
- Persona, tone, expertise, and boundaries — the four pillars of a great AI agent
- Knowledge injection: uploading files, documents, and structured data
- Comparison of Gems, Custom GPTs, and Claude Projects side by side
Writing Effective System Prompts
- The system prompt template: role, context, rules, tone, and output format
- Common mistakes: being too vague, too restrictive, or contradictory
- Testing your system prompt before building the agent
- Iterative refinement: the 3-round testing loop
Hands-On: Write a system prompt for your chosen AI specialist and peer-review it with a partner
| Module 2: Building Gemini Gems
50 minutes |
Gems Interface and Configuration
- Accessing Gems in Gemini Advanced (gemini.google.com)
- Setting the name, description, and instructions for your Gem
- Connecting Google Workspace: Docs, Drive, and Gmail integration
- Sharing Gems with team members or keeping them private
Hands-On: Build a Gemini Gem: create a Meeting Summariser Gem that takes meeting notes and outputs structured action items
Advanced Gem Configuration
- Multimodal Gems: accepting image and document inputs
- Using Gems for recurring workflows
- Limitations: what Gems cannot do (no web browsing, no real-time data)
| Module 3: Building Custom GPTs
60 minutes |
GPT Builder Walkthrough
- Accessing the GPT Builder in ChatGPT (chat.openai.com/gpts/editor)
- Configure tab: name, description, instructions, conversation starters
- Knowledge tab: uploading PDFs, Word docs, and CSVs as retrieval sources
- Capabilities: enabling web browsing, image generation (DALL-E), and code interpreter
Hands-On: Build a Custom GPT: create a Brand Voice Assistant loaded with your brand guidelines (or a sample document)
Publishing and Sharing
- Privacy settings: private, link-share, or public GPT store
- GPT Actions: connecting to external APIs (introduction only)
- Testing and debugging your Custom GPT
- Monitoring GPT usage and iterating based on real interactions
| Module 4: Building Claude Projects (Skills)
50 minutes |
Claude Projects Overview
- What Claude Projects are and how they differ from standard conversations
- Setting project instructions: Claude’s version of the system prompt
- Uploading knowledge files to your project
- Persistent memory within a project context
Hands-On: Build a Claude Project: create a Research Assistant project loaded with 3 topic-specific PDFs and a detailed instruction set
Claude’s Strengths for Specialisation
- Long-context advantage: why Claude suits document-heavy specialists
- Nuanced instruction-following for complex, rule-based workflows
- Combining Claude Projects with Claude’s API for advanced use cases (preview)