Rework Process Generator

The purpose of this article is to explain the value, usage, and configuration requirements of the Rework Process Generator AI Agent.

AI Agents in Tulip

Using the Rework Process Generator

Overview

The Rework Process Generator is a digital assistant that is able to generate rework instructions based on data contained in your Tulip tables.

Agent capabilities

• Review tables in your instance for previously reported defects and solutions for these defects
• Generate Rework instructions for defective parts

How to use the agent?

1. Create an automation
2. Add the AI agent to the automation
3. Trigger the automation by app usage when the operator wants to find rework instruction
4. Get the result in the form of table data manipulation, i.e. Agent generated solution to this defect/issue.

Use Cases

The Tulip Rework Process Generator can be used in two primary use cases: In-app troubleshooting and automated quality control.

In-App Troubleshooting for Operators

This use case directly helps shop-floor operators. When an operator identifies a defective part, they can use this agent to quickly get repair instructions.

An operator is on the assembly line and notices a specific defect, like "a pneumatic cylinder is leaking." Instead of searching through manuals or waiting for a supervisor, they can enter this issue into the Rework Instructions AI Agent.

Based on documents, work instructions, historical defect data, the agent generates clear, step-by-step rework instructions. This use case provides immediate, accurate, and standardized repair instructions, which reduces downtime, minimizes human error, and improves first-pass yield.

Automated Quality Control and Rework

This use case automates the process of generating rework instructions for quality control personnel or supervisors.

A quality control inspector uses a Tulip app to log a failed quality check. For example, a batch of 50 products is found to have a specific type of scratch. The inspector logs this defect, and then a Tulip automation could automatically trigger the Rework Instructions AI Agent to create instructions and append it to the rework.

The agent is automatically fed the defect type and quantity. It then searches the Tulip tables for the most effective rework procedure for that specific scratch type. It then creates the detailed instructions, which is available for an operator to pull up and start the rework process.

This use case streamlines the rework process by automatically generating instructions. It ensures consistency across all rework tasks, improves traceability, and provides data that can be analyzed to identify common defects and improve manufacturing processes over time.

In full-page Agent chat

In Automations

Agent configuration

The prompt of this agent is structured in the following sections:

• Goal
• Instructions
  • Task
  • Inputs
  • Outputs
  • Constraints
  • Capabilities and Reminders

Goal

This agent is able to analyse records from the defects table and actions table to generate rework instructions.

Depending on which tables this data is stored in, point this agent to look at all the appropriate tables in your workspace.

Goal:
You are an AI assistant with expertise in manufacturing processes and rework procedures. Your main responsibility is to review defect data and rework instructions from Tulip tables. You help shop-floor operators by providing accurate, step-by-step rework instructions in clear language, based strictly on the available data.

Instructions

Task:
1. Look at the following Tulip tables that show defects and how they were fixed -
(Operational Artifact) Defects
(Process Artifacts) Actions

2.  Determine the main steps needed to repair the defect.
Write easy-to-follow, step-by-step instructions to fix the defect.


Input:
The user will prompt with a defect/issue for you to analyse.


Output:
- Write a step-by-step plan telling operators exactly how to rework the item and fix the defect. 
- Also include the data source of where this information was found.
- Answer user requests in a concise manner. If there is a question, answer it directly and keep the answer short. Offer more details if the user asks for it.


Constraints:
- Use clear, jargon-free language understandable by shop-floor personnel.
- Do not guess information, base your recommendation solely on the data in Tulip.
- Remain neutral — report observed data without assuming root causes unless supported by evidence.
- Always clarify or ask follow-up questions if needed.
- If data is missing or ambiguous, note it explicitly in the summary.


Capabilities and Reminders:
- Review defect and action data from the specified Tulip tables.
- Provide step-by-step rework instructions using only information from Tulip—never invent or assume steps.
- Write in clear, simple language suitable for shop-floor operators.
- Always include the data source for each step in your instructions.
- If the necessary data is insufficient or unclear, ask the user for more details or clearly state what’s missing.
- Do not proceed without enough information; never guess at causes or solutions.
- Follow the defect and rework instructions exactly as shown in the available data.

Task

Here we define the exact work that we want the agent to do. In this case it is to generate rework instructions based on the data in specific tulip tables. Point this agent to look at the data in the appropriate tables in your workspace.

Inputs

As an example, we want to find rework instructions for a leak in a pneumatic cylinder - “my pneumatic cylinder is leaking”. The fake manufacturing facility in this scenario has this data stored in the actions and defects tables.

Outputs

Configure this section based on how you want the operators to see the outputs.

Constraints

Here we define standard guidereails for the agent to follow.

Capabilities and Reminders

Reinforce what you want the agent to do.

Tools used

The tools used by this AI Agent are the following:

• App Tools

  • getApp
  • getAppFolder
  • getStep
  • getTrigger
  • getWidget

• Data Tools

  • getTables
  • getTable
  • getRecord
  • getRecords
  • countRecords
  • updateRecord

Other AI Agents to read about

• Solution Design Agent
• Launch Readiness Reviewer
• Shift Summary Reporter
• Test Data Generator