How I Automated 20 Hours of Work With AI Agents

Key Concepts

• AI Automation/Orchestration
• Zapier Agents (Automated Agents)
• Model Context Protocol (MCP)
• Prompt Engineering (Warm-up Questions, Meta-Prompting)
• Internal Tooling & Automation
• Hackathons for AI Adoption
• Candidate Risk Detection

Instant Dossier Creator

• Main Topic: Using AI to quickly gather information about people before meetings or events.
• Key Points:
  • Wade uses Claude (an AI assistant) connected to Zapier, HubSpot, and ZoomInfo to create instant dossiers.
  • Claude uses MCP (Model Context Protocol) to access data from different tools. MCP allows agents to interact with data sources without needing specific API knowledge.
  • Claude projects allow for custom system prompts to guide the AI's behavior.
  • The system looks up contact and company information in HubSpot and ZoomInfo, then searches the web for additional details.
• Example: Wade demonstrates looking up Lars from Social Puppy. The system retrieves Lars's LinkedIn profile, job title at Zapier, location, and information about his side gig, Social Puppy. It also pulls information from HubSpot, noting that Lars isn't a paying customer.
• Technical Terms:
  • MCP (Model Context Protocol): A way for AI agents to interact with data sources.
• Challenges: MCP is still relatively new and can be buggy, leading to occasional errors.

Company Brief Generator

• Main Topic: An internal tool built at Zapier to quickly generate reports on companies before meetings.
• Key Points:
  • The tool uses Zapier Interfaces and pulls data from three sources:
    1. Web Search: Publicly available information about the company.
    2. Glean Search: Internal knowledge base (like an internal Google) that searches Slack, Google Drive, and other connected tools.
    3. Data Bricks: A replica of the customer database to understand usage patterns.
  • The output is a report with a company summary, glean summary, and usage visualizations.
• Example: Wade shows a report generated for Hampton. The report includes a company summary with details about Hampton's business, a glean summary with internal communications and issues, and visualizations of Zapier usage.
• Real-World Application: Preparing for customer dinners or meetings by having quick access to relevant information.
• Benefits: Helps employees be well-informed and prepared, leading to better customer interactions.
• Actionable Insight: Zapier templates under "lead management" and "agents" can be used to build similar tools.

Strategy Memo Analysis with ChatGPT

• Main Topic: Using ChatGPT to analyze and understand strategy memos or business reviews.
• Key Points:
  • Wade uploads a document (in the example, a leaked Mr. Beast memo) to ChatGPT.
  • He starts by asking ChatGPT to succinctly describe the document's contents.
  • He then uses the prompt "be 100x more specific" to get more nitty-gritty details.
  • This helps him understand the document and identify potential blind spots.
• Example: Analyzing the Mr. Beast memo to understand the core philosophy and what makes a YouTube video go viral.
• Warm-up Questions:
  • Describe it back to me.
  • Before we get into it, if there's any questions that you think you need answered so you have full context, please ask them now.
• Meta-Prompting: Asking ChatGPT to write the prompt for a specific task.
• Blind Spot Analysis: Asking ChatGPT to identify potential blind spots in a strategy or plan.
• Actionable Insight: Use warm-up questions and meta-prompting to get better results from ChatGPT.

Zapier Agents for Email Automation

• Main Topic: Using Zapier Agents to automate tasks, specifically email responses.
• Key Points:
  • Zapier Agents are fully automated agents that wake up based on events in your software (e.g., new email, new lead).
  • Wade demonstrates creating an agent that drafts replies to emails about job opportunities.
  • The agent analyzes the email content for keywords like "job," "career," etc.
  • It then drafts a reply directing the sender to the career page.
• Step-by-Step Process:
  1. Create a new agent in Zapier Agents.
  2. Choose a trigger (e.g., new email in Gmail).
  3. Define the agent's behavior (e.g., write a draft reply).
  4. Provide detailed instructions and prompts to guide the agent.
  5. Test and turn on the agent.
• Benefits: Automates repetitive tasks, saves time, and improves efficiency.
• Actionable Insight: Zapier Agents can be used to automate a wide range of tasks beyond email responses.

Fostering AI Adoption Within a Company

• Main Topic: Strategies for encouraging employees to embrace AI tools.
• Key Points:
  • Zapier has a company value: "Don't be a robot, build a robot."
  • They use hackathons to get employees familiar with AI tools.
  • Hackathons promote accountability (show and tell) and knowledge sharing.
  • They hold hackathons every 3-6 months to keep employees updated on new developments.
• Code Red Hackathon: A week-long event where the entire company paused their regular work to build things with AI.
• Show and Tell: At the end of the hackathon, everyone presented what they built.
• Results: Zapier increased AI usage from near zero to almost 90% of employees using it daily.

Candidate Risk Detector

• Main Topic: Using AI to detect potentially fraudulent job applicants.
• Key Points:
  • The system hooks into tools like Ashb, Slack, and Verafhone IP API.
  • It analyzes applicant data, including IP addresses and phone numbers.
  • It compares the data to other applicants to spot mismatches or suspicious patterns.
• Benefits: Saves HR teams time and resources by identifying and filtering out fraudulent candidates.
• AI Orchestration: Combining AI with workflow automation to solve complex problems.

Synthesis/Conclusion

The video showcases practical applications of AI and automation in a business context, emphasizing time-saving strategies and increased efficiency. Wade Foster demonstrates how Zapier leverages AI tools like Claude and ChatGPT, along with its own platform, to streamline various processes, from gathering information on individuals and companies to automating email responses and detecting fraudulent job applicants. The key takeaways include the importance of prompt engineering, the power of internal tooling, and the effectiveness of hackathons in fostering AI adoption within a company. The discussion also highlights the shift towards AI orchestration, where AI is integrated into existing workflows to solve complex problems and automate tasks that were previously impossible or impractical.