Practical AI use cases in procurement that are actually saving time — not just demos

There's a lot of hype around AI in supply chain. Here's what's working in practice versus what's still theoretical:

Actually working:

Supplier email drafting. Using Claude or ChatGPT to draft initial supplier inquiry emails, negotiation follow-ups, and claim letters. Saves 20-30 minutes per communication. The output needs editing but the structure is there.

Document extraction. Extracting line items from supplier invoices and packing lists into a structured format using AI. Reduces manual data entry significantly for teams processing high volumes.

Spend analysis from messy data. Cleaning and categorizing unstructured purchase data using AI before importing to Excel or Power BI. What used to take a day takes an hour.

Still mostly demo:

Automated supplier selection. The AI recommendations are generic without deep integration into your own historical data.

Predictive demand forecasting. Works well in theory but requires clean historical data that most mid-size operations don't have.

Contract analysis. Promising but requires careful validation — the model misses context-specific clauses.

What tools are you using and what's genuinely saving time versus just being interesting?