my bots just BROKE Wall St...

Summary

Wes Roth demonstrates how to build an AI “workforce” with Grok Bot, using a simulated investment-research desk as an example. Bots run on a cloud-based computer, so they can continue working when the user’s own devices are offline. The same setup could be adapted for business, work, or study tasks.

How the System Is Organized

• Create specialized bots: Give each bot a name, purpose, and optional personality or avatar.
• Use a chief of staff: Roth recommends making this one of the first bots. It serves as the main point of contact, delegates tasks, creates and manages other bots, and filters notifications.
• Connect apps: The demo uses Google Drive, Sheets, Calendar, and Gmail, with other integrations such as Notion and Canva available. Roth emphasizes that app permissions granted to one bot may be available to the others.
• Give bots their own email: Roth describes setting up AgentMail, a mailbox service for AI agents, so bots can receive newsletters and other information without using the user’s personal email.
• Group bots together: A group chat lets specialists share findings and hand work to one another.

Example Research Team

Roth’s simulated investment desk includes bots for coding, monitoring company filings, tracking supply chains and industry bottlenecks, following market events, analyzing revenue, challenging assumptions, maintaining portfolio records, considering taxes, and paper trading. A skeptical “Bear” bot is intended to identify weaknesses in the team’s conclusions. A paper trader uses the research to test potential trades without real money.

In one group discussion, the bots review filings, supply-chain issues, upcoming events, and competing analyses before the paper trader proposes a hypothetical portfolio change. Roth stresses that this is paper trading, not financial advice.

Automation, Alerts, and Records

• Recurring tasks: Bots can check sources on a schedule—for example, monitoring subreddits before and after market hours, or checking a website or social account every few minutes.
• Targeted notifications: Roth recommends limiting phone alerts to key bots, such as the chief of staff, rather than receiving messages from every agent.
• Long-term journals: Bots can save trades, analyses, opinions, and relevant context to Google Drive. Roth suggests reviewing and backing up this accumulated information over time so it can later be analyzed for patterns.
• Social-media research: Bots can search X for posts and sentiment. Roth notes that accessing X requires paid API requests.
• File handling: Bots can organize files in a designated folder on the user’s computer.

Teaching Reusable Skills

Roth demonstrates the “teach task” feature: the bot observes browser actions, uses them to infer the steps, and writes a reusable skill. He uses Minesweeper as a lighthearted example, then describes a more practical skill for gathering information from multiple online sources and presenting it on a dashboard.

Skills are stored as Markdown text files and can be transferred between compatible models or agents. Roth notes that training a workflow takes initial effort but can save time when repeated.

Practical Guidance

The tutorial’s central recommendation is to start with one recurring task, then add more pieces gradually. Roth says that a complete workflow—rather than a collection of isolated automations—is where the larger time savings come from. He also advises using clear bot responsibilities, a chief of staff to coordinate work, and shared storage so agents can build on one another’s output.

The video is sponsored by SpaceX. The investment examples are simulated and are not financial advice.

Main Speaker and Source

• Wes Roth — presenter and demonstrator of Grok Bot.