
| Cities monitored | 10 major French cities (shown on screen) |
| PowerPoint length | 8 slides generated automatically (shown on screen) |
| Execution time | 23 minutes for the agent to complete the full quarterly reporting cycle (shown on screen) |
| Data scope | 3 months of financial data (January to March) (shown on screen) |
Who will this analysis save an hour?We’ll send the link with the “actually” takeaway — they won’t have to watch the whole video.
You can use platforms like Codex or Claude Code which allow you to define an agent's behavior using natural language instructions. You provide the agent with 'context' files and 'skill' descriptions in markdown format, and the platform handles the underlying execution logic.
MCP stands for Model Context Protocol. it acts as a universal bridge that allows an AI agent to securely access and interact with your local files, databases, and web applications like Google Drive or Airtable.
Yes, by connecting the agent to tools like Gmail and PowerPoint via plugins or MCP servers, you can instruct it to draft emails based on data analysis and generate formatted slide decks following your brand guidelines.
By implementing a 'Memory' file and a self-evaluation loop, the agent can review its own outputs, identify errors, and store user feedback. It consults this memory before every new task to ensure it doesn't repeat past mistakes.
The best way is to start with a single, well-defined task. Identify the data inputs needed, the required output format, and the specific steps you currently take manually, then translate those into a 'Skill' for the agent.