Methodology for Creating a Personal Assistant
Creating an effective AI assistant is not just about clicking through settings – it requires a thoughtful approach and a clear methodology. In this article, we will introduce a proven process for creating a personal assistant in AYETO that will be truly useful and tailored to your needs.
Why do you need a methodology?
Many users jump straight into creating an assistant without prior planning. The result is often an assistant that: - Responds too generally - Does not understand the context of your work - Does not use available tools effectively - Gradually loses its usefulness
With a clear methodology, you can avoid these problems and create an assistant that will be your true helper.
Phase 1: Define the purpose and target audience
Ask yourself key questions
Before you open AYETO and start building, answer the following:
-
What will the assistant be used for? - Answering customer questions? - Helping with writing documents? - Data analysis? - Teaching and education?
-
Who will use the assistant? - Only you personally? - Your team? - External users?
-
What are the expected outputs? - Short answers to questions - Extensive documents - Structured data - A combination of all of the above
Practical example
Poorly defined purpose: "I need an assistant for work."
Well-defined purpose: "I need an assistant for technical support of product XYZ that will answer customer questions in Czech, use our product documentation, and escalate complex issues when necessary."
Phase 2: Choose the right AI model
Criteria for choosing a model
In AYETO, you have various AI models available. When choosing, consider:
| Criterion | Description |
|---|---|
| Speed | How quickly do you need responses? |
| Quality | How complex are the tasks the assistant will solve? |
| Price | What is your credit budget? |
| Capabilities | Do you need tools, vision, reasoning? |
Recommendations by task type
- Fast responses, FAQ: Smaller, faster models
- Complex analyses, creative generation: Advanced models with reasoning
- Working with images: Models with Vision support
- Using tools (Knowledge base, web search): Models with "tools" support
💡 Tip: If the selected model does not support some Tools, the AI tool settings will be ignored. Always verify the capabilities of the selected model.
Phase 3: Create instructions (System Prompt)
Instructions are the heart of your assistant. High-quality instructions determine: - Personality and tone of communication - Domain expertise - Limits and rules of behavior
Structure of effective instructions
# Role
A short description of who the assistant is.
# Tasks
What the assistant should do (list).
# Rules
How it should behave, what it must not do.
# Response format
How it should structure outputs.
# Context
Important information about the environment, company, product.
Example instructions
# Role
You are a technical support specialist for the AYETO application. You are friendly, patient, and always try to help.
# Tasks
- Answer user questions about AYETO features
- Provide step-by-step instructions
- Help solve technical problems
- Recommend suitable features based on user needs
# Rules
- Always respond in Czech unless the user communicates otherwise
- Never invent features that do not exist
- If you are not sure, admit it and recommend contacting support
- Be concise, but thorough
# Response format
- Use bullet points for clarity
- Number steps in guides
- Highlight important information in **bold**
Phase 4: Configure AI tools
AYETO offers a wide range of tools that expand the assistant’s capabilities:
Basic tools (recommended)
| Tool | Use case |
|---|---|
| Web search | Up-to-date information from the internet |
| YouTube transcription | Analysis of video content |
| DOCX generator | Creating Word documents |
| XLSX generator | Creating Excel spreadsheets |
Advanced tools
| Tool | Use case |
|---|---|
| REST API | Integration with external services |
| SOAP API | Legacy systems |
| Python interpreter | Advanced calculations and analyses |
| Chart generator | Data visualization |
When to enable which tools?
- FAQ assistant: Usually does not need any tools
- Research assistant: Web search + YouTube transcription
- Documentation assistant: DOCX/XLSX generators
- Analytical assistant: Python interpreter + Chart generator
- Integration assistant: REST/SOAP API
⚠️ Important: Enable only the tools the assistant truly needs. More tools = higher credit consumption and potentially slower responses.
Phase 5: Set up the knowledge base ('Knowledge Base Search')
The knowledge base allows the assistant to use your own documents as a source of information.
When to use 'Knowledge Base Search'?
✅ Yes, use 'Knowledge Base Search' when: - You have company documentation - You need answers based on specific data - You want an assistant with "memory" of your materials
❌ No, do not use 'Knowledge Base Search' when: - General model knowledge is sufficient - You do not have quality source documents - You only need a fast chat
Preparing documents for 'Knowledge Base Search'
- Formats: PDF, DOCX, TXT, and other text formats
- Quality: Structured, readable documents
- Scope: Relevant information, not "everything we have"
Advanced settings for 'Knowledge Base Search'
| Parameter | Description | Recommendation |
|---|---|---|
| Chunk size | Size of text blocks | 0 = system default |
| Vector store K | Number of documents for retrieval | 0 = system default |
Phase 6: Testing and iteration
Test scenarios
Before putting the assistant "into production," test it:
- Common questions – Does it answer typical questions correctly?
- Edge cases – How does it respond to unusual questions?
- Out of domain – Does it correctly refuse topics outside its area?
- Tools – Does it use tools effectively?
- RAG – Does it return relevant information from documents?
Iteration process
1. Test the assistant
2. Identify weaknesses
3. Adjust instructions or settings
4. Test again
5. Repeat until you are satisfied
Phase 7: Sharing and management
Sharing options
In AYETO, you can: - Share with colleagues – Add users using their email address - Set permissions – Read-only or write access - Export – JSON export for backup or transfer - Create a template – For repeated use in the gallery
Automated tasks
For advanced use, you can set up automated tasks: - Regular reports - Automatic data processing - Scheduled notifications
Summary of the methodology
| Phase | Key activity |
|---|---|
| 1 | Define the purpose and target audience |
| 2 | Choose an appropriate AI model |
| 3 | Write high-quality instructions |
| 4 | Enable the required tools |
| 5 | Set up the knowledge base (optional) |
| 6 | Test and iterate |
| 7 | Share and manage |
Checklist before launch
- [ ] Clearly defined assistant purpose
- [ ] Appropriate AI model selected
- [ ] Structured and complete instructions
- [ ] Only the necessary tools enabled
- [ ] Relevant documents uploaded (if using RAG)
- [ ] Basic test scenarios completed
- [ ] Sharing permissions configured (if needed)
Next steps
Now you have a complete methodology for creating a personal assistant. We recommend:
- Start simple – Create your first assistant with minimal configuration
- Expand gradually – Add tools and documents as needed
- Monitor usage – Track credit consumption
- Share experience – Collaborate with your team on improvements
Do you have questions about creating assistants? Use our AI support assistant or contact us.