> ## Documentation Index
> Fetch the complete documentation index at: https://docs.chemcloud.com/llms.txt
> Use this file to discover all available pages before exploring further.

# Exercises: Quality & Technical

> Hands-on walkthrough of ChemCloud for Quality and Technical teams — product setup, approvals, document management, and AI-powered QC workflows.

# ChemCloud Exercises: Quality & Technical

**Purpose:** Hands-on walkthrough of core ChemCloud workflows for Quality and Technical teams, covering product setup, approvals, AI document analysis, and QC processes.

***

## What We Want to Get Out of This

1. **Show how an end-to-end workflow works** — from product setup through approvals, documents, and QC checks
2. **Demonstrate the capability of AI** — see how ChemCloud's AI agents process documents, answer questions, and surface insights
3. **Feel how an AI agent can be like an additional member of the team** — delegate tasks, ask questions, and let it handle the manual work

***

## Before You Start

* Log in to the ChemCloud demo environment
* Have some internal specification and supplier product documents to hand
* Keyboard shortcut to remember: **Cmd+K** (Mac) or **Ctrl+K** (Windows) opens the AI search/chat anywhere in the app

***

## Exercise 1: Create a Product with an Internal Specification (\~10 mins)

**Objective:** Pick a raw material you know well. Create it as a product in ChemCloud, fill in the details, set the document requirements, and upload an internal specification.

<Accordion title="Step-by-step Guide">
  <Steps>
    <Step title="Create a Product">
      Go to **Products** in the top navigation (or use the **+** button). Click **Add Product**.
    </Step>

    <Step title="Enter Product Details">
      Fill in the raw material code, name, and any other details you'd normally track (annual usage, UOM, etc.).
    </Step>

    <Step title="Set Document Requirements">
      Configure which documents are required for this product — e.g., internal spec, SDS, COA.
    </Step>

    <Step title="Upload an Internal Specification">
      Upload your internal specification document against the product.
    </Step>

    <Step title="Try the AI">
      Open the AI Chat (**Cmd+K**) from the search bar in top navigation and try asking it questions about the product you just created. Also try to trip it up — ask something ambiguous, or something the system shouldn't know yet.
    </Step>
  </Steps>
</Accordion>

**Expected Outcome:** Product created with an internal specification uploaded and document requirements set.

***

## Exercise 2: Set Up Approved Materials with Approvals (\~15 mins)

**Objective:** For the product you created, add approved materials from different suppliers at different stages of the approval process. Upload supplier documents against each one.

<Accordion title="Step-by-step Guide">
  <Steps>
    <Step title="Add Approved Materials">
      Go to **Add Approved Material** (from the **+** button in top navigation). Choose your product, select a supplier (or add a new one), and select the supplier's product.
    </Step>

    <Step title="Set Approval Statuses">
      Set up a realistic mix — some fully approved, some in progress, some not started.
    </Step>

    <Step title="Upload Supplier Documents">
      Upload supplier documents (specs, SDS, etc.) against each approved material. See if the AI correctly detects document types and extracts dates.
    </Step>

    <Step title="Use Conversations">
      Try using the **Conversations** feature to leave notes or tag a colleague on one of the in-progress approvals.
    </Step>

    <Step title="Try the AI">
      Open the AI Chat and ask it questions about your approvals and documents — e.g., what's pending, what's missing, compare specs between two suppliers. Try to induce errors — ask about a supplier that doesn't exist, or ask it to compare documents you haven't uploaded yet.
    </Step>
  </Steps>
</Accordion>

**Expected Outcome:** Multiple approved materials set up with different approval statuses and supplier documents uploaded.

***

## Exercise 3: Send a Document to ChemCloud via Email (\~5 mins)

**Objective:** Forward a supplier document to ChemCloud's AI email agent and see how it processes it automatically.

<Accordion title="Step-by-step Guide">
  <Steps>
    <Step title="Find Your Email Address">
      Navigate to **ChemCloud AI → Email Inbox** to find your ChemCloud email integration address.
    </Step>

    <Step title="Forward a Document">
      Forward a supplier document (spec, SDS, or COA) to that email address.
    </Step>

    <Step title="Check the Results">
      Check back in ChemCloud to see how it was processed. Did it get the right product? The right document type? Were the extracted details accurate?
    </Step>

    <Step title="Try the AI">
      Ask the AI about the document you just emailed in. Try asking it to summarise the document or compare it against something already in the system.
    </Step>
  </Steps>
</Accordion>

**Expected Outcome:** Document automatically filed against the correct product and supplier with key data extracted.

***

## Exercise 4: Use the AI to Compare Supplier Documents (\~10 mins)

**Objective:** Now that you have documents uploaded against multiple suppliers, use the AI to compare them and flag key differences.

<Accordion title="Step-by-step Guide">
  <Steps>
    <Step title="Ask for a Comparison">
      Open the AI Chat and ask it to compare two supplier specifications for the same product.
    </Step>

    <Step title="Test with Vague Prompts">
      Try a vague prompt — e.g., "compare these suppliers" without specifying what to compare on. See how it handles it.
    </Step>

    <Step title="Ask About Missing Data">
      Try asking about something that isn't in the documents, or ask it to compare documents you haven't uploaded yet.
    </Step>

    <Step title="Follow Up">
      Ask a follow-up question based on the AI's first answer. Does it maintain context?
    </Step>
  </Steps>
</Accordion>

**Expected Outcome:** AI identifies and presents key differences between supplier specifications.

***

## Exercise 5: Place an Order, Upload a COA, Compare to Internal Spec (\~20 mins)

**Objective:** This is the core QC workflow. Place an order, simulate receiving a delivery, upload a COA, and use the AI to check it against your internal specification.

<Accordion title="Step-by-step Guide">
  <Steps>
    <Step title="Place an Order">
      Place an order for one of your approved materials. Complete the purchase steps.
    </Step>

    <Step title="Simulate a Delivery">
      Update the delivery status to simulate receiving the goods.
    </Step>

    <Step title="Upload a COA">
      Upload a Certificate of Analysis against the delivery.
    </Step>

    <Step title="Ask the AI to Check the COA">
      Ask the AI to check the COA against your internal spec and flag anything out of spec. Does it identify pass/fail on specific parameters? Does it flag borderline results?
    </Step>

    <Step title="Try to Catch It Out">
      Ask about a parameter that's borderline. Ask it to make a release recommendation and see if it's confident when it shouldn't be.
    </Step>
  </Steps>
</Accordion>

**Expected Outcome:** End-to-end QC workflow completed with AI-powered COA analysis against internal specification.

***

## Exercise 6: Open-Ended AI Exploration (\~10 mins)

**Objective:** Just play around. Ask the AI whatever questions come to mind about the data you've entered.

<Accordion title="Ideas to Try">
  <Steps>
    <Step title="Document Compliance">
      Ask about document expiry or compliance gaps across your products.
    </Step>

    <Step title="Supplier Coverage">
      Ask about supplier coverage — which products have single-source suppliers?
    </Step>

    <Step title="COA Comparisons">
      Try comparing COAs across different batches or deliveries.
    </Step>

    <Step title="Try to Break It">
      Ask something the system definitely shouldn't know. See if it admits uncertainty or confidently gives a wrong answer. Screenshot it if so!
    </Step>
  </Steps>
</Accordion>

### Things to Note

* What questions did you want to ask that it couldn't answer?
* Were any responses wrong or misleading?
* Did you manage to get it to confidently give a wrong answer? (Screenshot it if so!)