How to build an AI agent team from scratch

TL;DR

An AI agent team can automate sponsorship deals end to end. Four agents coordinate via Slack and Hyper Agent: Marco handles the pipeline, Vetter conducts due diligence, Ron writes concepts, Sam creates presentations. The system self manages from incoming deal to final review, reducing manual vetting and speeding decisions.

Transcript

So the video you're watching right now is sponsored. And honestly, the interesting part here is that I didn't vet the sponsor, I didn't research the company, and I didn't even come up with the first draft of the ideas for this video. A team of agents did all of that for me automatically. So by the time I sat down to actually look at this deal, ther... Read More

Key Insights

• AI agents can be connected to Slack, Gmail, and Google Drive to automate a workflow
• A dedicated sponsor pipeline agent can trigger due diligence automatically when a deal is received
• A vetting agent compiles risk and feasibility data into a usable outline for decision making
• A writer agent creates multiple video concept ideas based on the vetted data
• A presenter agent turns the vetted ideas into a presentable deck for review
• Agents can be configured to work with specific tools and integrations to fit the workflow
• The system can be extended by adding more specialized agents to handle different tasks
• Creating a new agent is possible by using a prompt to define role, prompts, and tools

Questions & Answers

Q: What is the main benefit of using an AI agent team for sponsorships?

The main benefit is that multiple AI agents can automatically handle the end to end sponsorship workflow, from initial deal intake to due diligence to producing a presentation, with minimal manual prompting. This reduces repetitive work, speeds up decision making, and keeps the human in the loop only for final approval when needed.

Q: Who are the four agents described and what does each do?

The four agents are Marco, Vetter, Ron, and Sam. Marco handles the sponsorship pipeline and uses Slack, Gmail, and Drive to trigger the process. Vetter performs full due diligence and audits the deal, creating a risk outline. Ron acts as writer and generates video concepts based on the vetted data. Sam builds a presentation deck summarizing the findings for quick review.

Q: How do the agents interact to produce a final decision deck?

The agents interact by passing information through a shared workflow initiated from a deal. Marco starts the process, Vetter compiles a due diligence report, Ron generates concepts and notes, and Sam creates a decision deck. The dashboard shows each agent’s activity, instructions, and ability to delegate tasks to others, culminating in a complete presentation.

Q: What platforms and tools are mentioned as part of the workflow?

The workflow uses Slack for messaging, Gmail for communication, Google Drive and Google Docs for storage and collaboration, and Google Slides for the final presentation deck. These integrations enable seamless data sharing and allow agents to perform tasks without manual interference.

Q: Can new AI agents be created from scratch in this setup?

Yes, new AI agents can be created from scratch by providing a prompt that defines the agent’s role, model, tools, and integrations. The setup includes options to select models, set capabilities, and connect to various services, enabling rapid customization for different parts of the workflow.

Q: What role does the presentation agent play in the process?

The presentation agent, Sam, takes the vetted data and concepts and turns them into a clear, shareable deck. It compiles the key findings, recommends concepts, and formats the material for quick review by the human decision maker, helping to speed up the final evaluation stage.

Q: How is the automation scaled for multiple deals?

Automation scales by running parallel workflows for each incoming deal. Each deal triggers the same chain of Marco, Vetter, Ron, and Sam, with dashboards tracking activity and costs. The system can be extended with additional agents or rules to handle increasing volume while maintaining consistency and speed.

Q: What is needed to start building this AI agent team?

To start, you need a platform that supports agent creation and orchestration, a basic prompt to define the roles, and access to tools like Slack, Gmail, and Google Drive. A human quarterback should monitor outcomes, approve final results, and tune prompts, enabling iterative improvement and reliable automation.

Summary & Key Takeaways

• A four agent system is shown that automates the sponsorship pipeline from intake to final presentation.

• The video demonstrates how different agents share tasks and hand off work to produce a finished decision deck.

• The process emphasizes automation, collaboration between agents, and a quick review for approval or rejection.