Should you build or buy AI agents?

Edition 116 | September 4, 2025

If I were them, I’d budget at least $10 billion of that towards finding Dario Amodei a hairstylist.

Welcome back to Building AI Agents, your biweekly guide to everything new in the field of agentic AI!

We’re trying out a new format today! Please let us know what you think of it in the poll at the bottom, below Quote of the Day. Here’s last Thursday’s issue for comparison.

Thanks for you help making Building AI Agents better for readers like you!

In today’s issue…

• OpenAI’s new model to power voice agents

• The threat of intelligently mutating agent viruses

• Should you build or buy AI tools?

• Salesforce replaces 4,000+ positions with agents

• 15 operating principles for enterprise agents

…and more

🔥 IN CASE YOU MISSED IT

Readers’ favorite items from the past week

1. The state of AI in 2025

2. Build your own stock portfolio agent

3. Agents that reliably follow rules

4. How to build a coding agent: a free workshop

5. Automate your phone using agents

 🤖 AGENT OF THE WEEK

This guide walks you through building an essential Retrieval-Augmented Generation (RAG) agent that turns static documents into a conversational assistant. Using n8n with GPT-4.1, Pinecone, and Google Drive, you’ll create an agent that can instantly answer questions from employee handbooks, customer FAQs, product manuals, sales playbooks, and more in real time.

Step-by-Step:

1. Open up n8n: Create a new workflow (free account lasts 2 weeks!)

2. Add your docs: Connect Google Drive and load documents for the agent

3. Process the docs: Split the documents into chunks* and generate embeddings** with OpenAI

   1. Text chunking → Breaking a big document into smaller, searchable pieces so the AI doesn’t get overwhelmed

   2. Embeddings → Turning each chunk into a “numeric fingerprint” so the AI knows what it’s about and can find the right information fast

4. Store for retrieval: Send embeddings to the Pinecone vector database

5. Chat with the Agent: GPT-4.1 answers questions by pulling the most relevant info from your vectorized documents

6. Test it live: Try “What’s our vacation policy?”, “How do I reset my account’s password?”, or “What’s the return policy?”

Bonus: We’ve put together a detailed video walkthrough and the exact n8n agent template. Access both in our Building AI Agents Community here!