01
Suppliers rarely describe things the same way
Procurement teams often receive product and service descriptions from many vendors. One supplier might write: “Annual endpoint protection subscription” Another: “Cybersecurity software renewal” Another: “EDR license — 500 endpoints” Your organization may want all three classified under the same procurement category.
02
Historical purchasing data provides the examples
If previous purchases have already been categorized, they can train a StayCharted AMT model.
| Description | Procurement Category |
|---|---|
| CrowdStrike endpoint renewal | Security Software |
| Dell Precision workstation | Computer Hardware |
| Building HVAC maintenance | Facilities |
The model learns from how your organization has classified previous purchases.
03
Bring supplier product photos into your taxonomy
With Vision, StayCharted AMT can learn categories from supplier or product photos as a separate image workflow. Start with photos your team has already labeled, uploaded in category folders or supplied as picture links.
For example, a photo of a backpack can be categorized as Backpacks even when a supplier’s text description is inconsistent. Use an AI Image Classifier Model, or evaluate a Dedicated AI Image Model on Business with Vision.
Text and image models are separate approaches; this does not combine description and photo inputs into one model. Photo categorization does not establish an exact SKU match, read part numbers, or replace your field-mapping rules. Review predictions before applying your catalog labels.
04
Useful across procurement workflows
A custom model can predict fields such as:
- Spend category
- Procurement group
- Supplier type
- Product family
- Cost center
- Commodity class
- Review team
The output doesn't have to follow a public taxonomy unless you want it to.
05
Handle incoming data before it becomes another cleanup project
The model can be used when supplier data is imported rather than waiting until reporting time.
- Supplier File / Feed
- Procurement System
- StayCharted AMT API
- Internal Category
- Standardized Record
This can reduce the amount of manual normalization required before spend analysis, reporting or workflow routing.
06
Start with one messy field
Procurement data can be complicated. You don't need to solve all of it at once. Train on purchases your team has categorized and check the predictions against a separate set of reviewed purchases.