Clase 03 - De la Automatización a la Entrega Final: Tu Agente de IA a Nivel Profesional

Summary

The class demonstrates how to move from basic automation to a more complete, professional AI solution. The instructor builds a recruitment workflow that receives résumés, extracts and evaluates their contents, records results in a spreadsheet, and presents the data in an interactive dashboard.

Main Ideas and Lessons

Map the process before choosing tools

Automation should begin by mapping the existing workflow, not by opening an automation platform. The recruitment process covered in the class includes:

1. Receive a candidate’s résumé.
2. Download and read it.
3. Compare the candidate’s profile with the vacancy requirements.
4. Assign a score and evaluation.
5. Record the information in a spreadsheet.
6. Decide whom to invite for an interview.

Automate operational work, not human accountability

An AI agent can handle repetitive tasks such as extracting résumé details, applying consistent criteria, and organizing results. The final hiring decision should remain with a person, especially when judgment and consequential decisions are involved.

Build and test the workflow incrementally in n8n

The instructor creates a new n8n workflow and uses a form to collect a candidate’s full name, email address, and résumé in PDF format. The form can be adapted or replaced with another service, provided it can be integrated.

Extract and prepare résumé data

• Use an Extract from File node to retrieve text from the uploaded PDF.
• Use an Edit Fields node to select and organize information needed later, such as the candidate’s name, email, and résumé text.
• Give nodes descriptive names to make the workflow easier to understand and maintain.
• Test each step with sample data. Pinning test data can avoid repeatedly submitting the form during development.

Give the AI clear instructions and criteria

The workflow sends the candidate’s details and résumé text to an AI model; the demonstration uses an Anthropic model.

The prompt follows the IDEAL approach discussed in earlier classes: define the agent’s identity or role, intention, relevant details, examples, action, and limits. In the example, the agent acts as a recruiter assessing candidates for a Junior Data Analyst vacancy, including requirements such as Excel and SQL experience.

The prompt defines a scoring scale and explains what different score ranges mean. The instructor emphasizes that criteria should be tailored to the specific job and business context rather than treated as universal. Structured formatting, including Markdown, can make instructions and priorities clearer. A well-calibrated prompt may not require the most advanced or expensive model; the model should suit the task.

Store results in a spreadsheet

Google Sheets is added to the workflow and configured to append each evaluation to a sheet. The example records the date, candidate’s name and contact details, résumé summary, score, and a suggested contact message. Keeping records in the organization’s spreadsheet also makes the data available for later review.

Test the full workflow with different résumés

The instructor submits sample candidates to see how the agent applies the same criteria to each. Candidates should be assessed against the actual vacancy, and applicants should keep their résumés accurate while making relevant experience clear.

Extend the automation into a management interface

• The class introduces Claude’s connectors, which can connect the assistant to tools and data sources such as Google Drive.
• It demonstrates Artifacts as a way to create an interactive view of data from a prompt. The example dashboard summarizes candidate totals and score distributions, including candidates who fully, partially, or do not meet the criteria.
• A further example shows how Claude Code could be used to create a custom careers page and HR dashboard. These are possibilities for extending the workflow, rather than part of the core n8n build.

Broader Takeaways

• AI automation can reduce repetitive administration and free staff to spend more time on interviews, analysis, and other work requiring human judgment.
• A workflow can be adapted to different tools and business processes, provided the necessary integrations are available.
• Building a professional solution involves more than making an AI call. It requires process mapping, clear criteria, data handling, testing, and a useful interface for the people acting on the results.
• The class ends with a Daxus training-program promotion and a question-and-answer segment. These are separate from the workflow lesson.

Speakers and Sources Featured

• Ingrid — The main instructor, who explains and demonstrates the recruitment automation.
• Unnamed co-host or moderator — Supports the live class, reads or responds to chat, and helps manage the presentation.
• Socrates Cabral — Named in the opening testimonial montage; he describes professional and salary improvements he attributes to training.
• Other unnamed Daxus students or alumni — Appear in the testimonial montage and report career, promotion, salary, or productivity gains.
• Live audience members — Represented through chat comments and questions, but not individually identifiable as speakers from the subtitles.