> For the complete documentation index, see [llms.txt](https://docs.karini.ai/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.karini.ai/agentic-ai-workshops/automotive-and-manufacturing.md).

# Automotive & Manufacturing

Four hands-on labs that build working agents for manufacturing on Karini AI and AWS: from drawings and parts to supplier catalogs and machine diagnostics.

Manufacturers don't lack data. They lack the time to turn it into decisions. Part numbers live in thousand-page catalogs, supplier price lists arrive in a different format every time, drawings have to be read symbol by symbol, and when a machine starts to fail, the evidence is scattered across logs, maintenance records, and procedures. Each of these is skilled work that engineers, buyers, and maintenance teams do by hand, every day.

These four hands-on labs show how agentic AI takes on that work on Karini AI, running on AWS. Each lab builds a working agent end to end, from prompt and tools to a deployed copilot or workflow, using real manufacturing data. Each one keeps people in control: agents ground every answer in your catalogs and systems, pause for approval when a decision matters, and say plainly when they can't find an answer instead of guessing.

## The Labs

| Lab                                                                                                                             | What you build                                                                                                                   | Where it fits              | Time         |
| ------------------------------------------------------------------------------------------------------------------------------- | -------------------------------------------------------------------------------------------------------------------------------- | -------------------------- | ------------ |
| [SpecToBOM](/agentic-ai-workshops/automotive-and-manufacturing/assembly-to-bom-intelligence.md)                                 | A deep agent that reads a P\&ID drawing, asks for missing specs, and returns a part-numbered, annotated bill of materials        | Engineering and quoting    | 60 to 75 min |
| [SpecMatch](/agentic-ai-workshops/automotive-and-manufacturing/specmatch-build-a-knowledge-graph-agent-for-industrial-parts.md) | A knowledge graph agent that matches a plain-language fitting spec to the exact ordering number                                  | Sales and customer service | About 60 min |
| [CatalogSync](/agentic-ai-workshops/automotive-and-manufacturing/supplier-catalog-normalization.md)                             | A deep agent that maps any supplier's price catalog, pauses for your approval, and validates every line against the ERP          | Procurement and ERP        | 60 to 75 min |
| [FaultLens](/agentic-ai-workshops/automotive-and-manufacturing/predictive-diagnostic-assistant.md)                              | A webhook-triggered workflow that screens anomalies, gathers evidence, and emails a cited diagnosis when a pump develops a fault | Maintenance and operations | 75 to 90 min |

Together they follow a product through the business: from the customer's drawing, to the right part, to the supplier's price, to keeping the line running once it's built.

## What You Will Learn Across the Labs

* **Grounded answers:** knowledge graphs, vector search, and ERP data so every answer traces to a source
* **Human in the loop:** agents that pause for approval or missing inputs mid-task, then continue
* **Multi-agent design:** routers, subagents, and agents that hand work to each other
* **Event-driven automation:** workflows that start themselves from an outside system, with no one asking
* **Right-sized models:** comparing Claude Sonnet and Haiku side by side, and choosing a primary and fallback model for each job
* **Governance:** guardrails, predefined database calls, and traces that show every step an agent took

## Suggested Order

Start with [SpecMatch](/agentic-ai-workshops/automotive-and-manufacturing/specmatch-build-a-knowledge-graph-agent-for-industrial-parts.md) if you're new to Karini AI. It introduces prompts, tools, recipes, and copilots with a single agent. [CatalogSync](/agentic-ai-workshops/automotive-and-manufacturing/supplier-catalog-normalization.md) and [SpecToBOM](/agentic-ai-workshops/automotive-and-manufacturing/assembly-to-bom-intelligence.md) add deep agents that pause for human input, and SpecToBOM adds a subagent. Finish with [FaultLens](/agentic-ai-workshops/automotive-and-manufacturing/predictive-diagnostic-assistant.md), the most advanced lab, which combines a router, two agents, and a live webhook trigger.

## Before You Begin

* A Karini AI account at [https://app.karini.ai](https://app.karini.ai/) , set up by your Karini AI representative
* A short platform orientation with a Karini AI instructor
* Templates, skills, datasets, and credentials prepared by your instructor for each lab

## Where This Leads

Each lab is the first link in a longer chain. A validated BOM can flow into quoting, material availability, and production planning. A normalized supplier catalog can feed procure-to-pay. A diagnosis can open a work order and check spare parts. The same platform, patterns, and governance carry each agent from a workshop exercise to production.