How to Create Custom AI Agents with Document Intelligence

Custom AI agents are one of the most powerful features of lookatmy.ai, allowing you to create specialized assistants that understand your specific domain, have access to your documents, and follow your unique instructions. In this guide, we'll walk through creating a recruiter agent that can search through resumes and match candidates to job requirements.

Why Custom AI Agents?

Custom agents go beyond generic AI chat by:

• Specializing in your specific use case with tailored personalities and behaviors
• Accessing your uploaded documents to provide context-aware responses
• Following specific instructions you define for consistent results
• Remembering context across conversations for more coherent interactions

Step-by-Step: Creating a Recruiter Agent

Let's create a practical example: a recruiter agent that can search through a collection of resumes to find the best candidates for specific job requirements.

Step 1: Create Your New Bot

Navigate to the Customize page and click on "New Bot" to start creating your agent.

Choose the following settings for our recruiter agent:

• Bot Name: "Talent Scout"
• Personality: Select "Professional" and "Expert" to ensure formal, detailed responses
• Description: Add a brief description explaining this bot's purpose

The personality combinations matter - "Professional" ensures business-appropriate communication, while "Expert" provides detailed, technical analysis of candidates.

Step 2: Define Custom Instructions

The real power comes from the custom prompt. Click "Edit Custom Prompt" to add specific instructions.

Here's an effective prompt for our recruiter agent:

markdown
Use our pool of resumes from the @Resume Collection to find
good candidates for the given job descriptions. List their strong
match and unmatched skills.

This prompt:

• References a specific document collection using the @ mention
• Defines the task clearly (matching candidates to jobs)
• Specifies the output format (listing matched and unmatched skills)

Step 3: Upload Your Document Collection

Before using the agent, you need to upload your resume collection. Navigate to the Storage page and upload your PDF resumes.

Pro tips for document uploads:

• Group related documents in collections for easier reference.
  • This will save time and follow seperation of concerns for each work/client in the long run.
• PDFs work best for structured data like resumes
• You can upload up to 500MB of documents on the Pro plan

Step 4: Using Your Recruiter Agent

Now let's see the agent in action. When you ask it to find candidates, it automatically searches through your document collection.

Example query: "Find a sales engineer with 5 years experience and Fortune 500 client background"

The agent will:

1. Search through all uploaded resumes
2. Analyze relevant experience and skills
3. Match candidates to your requirements
4. Present findings in a structured format

Step 5: Review the Results

The agent provides detailed analysis of each candidate, highlighting why they match (or don't match) your requirements.

The response includes:

• Top candidates with match percentages
• Strong matches listing specific qualifications
• Experience details from their resumes
• Potential concerns or missing requirements

More Use Cases and Example Prompts

Legal Document Analyzer

Prompt: "Analyze contracts in @Legal Documents for liability clauses, termination conditions, and payment terms. Highlight any potential risks."

Use case: Quickly review multiple contracts for standard terms and identify areas needing legal review.

Research Assistant

Prompt: "Search @Research Papers for information relevant to the user's question. Cite specific papers and page numbers. Summarize key findings and methodologies."

Use case: Academic research, literature reviews, or technical documentation analysis.

Customer Support Agent

Prompt: "Use @Product Documentation and @FAQ Collection to answer customer questions. If the answer isn't in the documentation, acknowledge this and suggest contacting support."

Use case: Automated first-line support that accurately references your actual documentation.

Financial Analyst

Prompt: "Analyze financial reports in @Quarterly Reports. Calculate key metrics like growth rates, margins, and ROI. Compare against industry standards and previous periods."

Use case: Quick financial analysis and reporting from uploaded statements and reports.

Content Editor

Prompt: "Review documents in @Drafts Collection for grammar, clarity, and tone. Suggest improvements while maintaining the author's voice. Flag any factual inconsistencies."

Use case: Editing and proofreading assistance with consistency checks across documents.

Marketing Campaign Agent

Prompt: "Analyze @Campaign Performance reports and @Competitor Analysis docs. Generate campaign ideas based on successful strategies. Include target audience, key messages, channels, and expected ROI. Reference specific data points from the documents."

Use case: Data-driven marketing strategy development using historical performance data and competitive intelligence to create targeted campaigns.

Advanced Tips

1. Personality Combinations

• Expert + Concise: Technical accuracy with brief responses
• Creative + Friendly: Brainstorming and ideation assistant
• Professional + Cheerful: Customer-facing support agent

2. Document Organization

• Create separate collections for different document types
• Use descriptive file names for better search results
• Update collections regularly to keep information current

3. Prompt Engineering

• Be specific about output format (bullets, tables, summaries)
• Reference collections explicitly with @ mentions
• Include examples of desired responses in your prompt

4. Testing and Iteration

• Start with simple queries to test basic functionality
• Gradually increase complexity as you refine the prompt
• Save successful prompts as templates for similar agents

Common Pitfalls to Avoid

1. Overloading the prompt: Keep instructions clear and focused on one main task
2. Vague references: Always use specific collection names with @ mentions
3. Ignoring personality: The personality affects tone and detail level significantly
4. Not organizing documents: Poor file organization leads to poor search results

Conclusion

Custom AI agents transform lookatmy.ai from a chat interface into a powerful business tool tailored to your specific needs. Whether you're recruiting, analyzing documents, or automating support, the combination of custom instructions and document intelligence creates agents that truly understand your domain.

Start with one focused use case, test thoroughly, and iterate based on results. As you become comfortable with the system, you can create multiple specialized agents for different aspects of your work, each with its own personality, knowledge base, and expertise.

Ready to create your first custom agent? Sign up for lookatmy.ai and start building your AI-powered workforce today.