---
title: "Automation, Squared"
description: "The article presents “second-order automation,” in which one agent designs reusable workflows for another agent to execute. Its spreadsheet system pairs an “architect” that watches a screen recording, generates detailed prompts, identifies required files and context, with a “doer” that searches workbooks and performs the instructions. The design addresses first-order automation’s dependence on users to supply precise prompts and fresh context for every task, which limits scalability. On an approximately 50-task SpreadsheetBench subset, the spreadsheet agent reached 49.5% accuracy on soft-restriction tasks versus OpenAI’s 45.5%, and 32.5% on hard-restriction tasks versus GPT-4o’s 13.4%; some one-to-two-hour tasks finished in under 10 minutes. The article concludes that recorded workflows can let non-technical users create reusable automations with little additional input, although architect performance has not yet been formally benchmarked."
---

# Automation, Squared

[Ramp](https://yomu.fyi/company/ramp) · Hunter Pruett · Aug 12, 2025

**Type:** Explainer

## Summary

The article presents “second-order automation,” in which one agent designs reusable workflows for another agent to execute. Its spreadsheet system pairs an “architect” that watches a screen recording, generates detailed prompts, identifies required files and context, with a “doer” that searches workbooks and performs the instructions. The design addresses first-order automation’s dependence on users to supply precise prompts and fresh context for every task, which limits scalability. On an approximately 50-task SpreadsheetBench subset, the spreadsheet agent reached 49.5% accuracy on soft-restriction tasks versus OpenAI’s 45.5%, and 32.5% on hard-restriction tasks versus GPT-4o’s 13.4%; some one-to-two-hour tasks finished in under 10 minutes. The article concludes that recorded workflows can let non-technical users create reusable automations with little additional input, although architect performance has not yet been formally benchmarked.

## Context

First-order spreadsheet automation requires users to spend substantial time supplying precise prompts and the right context for every run. Missing context or poor instructions can cause failures, while requiring fresh human input for each task limits scalability and reduces the time-saving impact.

## Approach / What changed

The system combines an architect agent and a doer agent. The architect watches a screen recording of a spreadsheet workflow, generates instructions, identifies required files and additional context, and compiles them into a reusable Process. The doer then receives that Process and executes the spreadsheet task using workbook-search and other tools.

## Takeaways

- The architect agent converts a recorded spreadsheet workflow into a reusable Process containing steps, required files, and additional information.
- On an approximately 50-task SpreadsheetBench subset, the spreadsheet agent scored 49.5% on soft-restriction tasks and 32.5% on hard-restriction tasks.
- Tasks that previously took one to two hours were completed in under 10 minutes in some cases, though partially completed tasks still required third-party intervention.

**Tags:** [AI Agents](https://yomu.fyi/topic/ai-agents), [Architecture](https://yomu.fyi/topic/architecture), [Performance](https://yomu.fyi/topic/performance)

- Source: [Ramp](https://builders.ramp.com/post/automating-automation)
- Source URL: https://builders.ramp.com/post/automating-automation
- Ingested by Yomu: 2026-09-01T01:34:42.566Z

[Read original post](https://builders.ramp.com/post/automating-automation)