AI agent workflow automation

Overview

A recurring office workflow required gathering information, keeping context across steps, validating intermediate results, and producing a structured final output. The work was repetitive, but still needed judgment and review.

FerrisDev built an AI-assisted workflow that automated the repetitive parts while keeping the output structured and suitable for human use.

Why a Fixed Script Wasn't Enough

The existing process took a few hours because information had to be collected, interpreted, checked, and reformatted manually. A simple script was not enough because the task required context handling and flexible reasoning over changing inputs.

The Agent Workflow

The workflow used an agent harness built on the Rig (rig.rs) framework. It could gather context, call tools, validate intermediate outputs, and produce structured results. The system was designed around repeatability: the same workflow could be run again with new inputs while keeping the output format consistent.

Progress in Slack

The agent posted periodic status reports to Slack so the team could follow the workflow as it progressed.

Outcome

The process was reduced from a few hours to roughly 10-20 minutes while preserving a reviewable final output and giving the team a clearer view of the work in progress.

Where This Approach Fits

This kind of system is useful when a team has a recurring knowledge-work process that is too variable for a fixed script, but structured enough for an AI-assisted workflow with clear inputs, checks, and outputs.