40 Use Cases, One Platform
Kale Holding is one of Turkey's largest industrial groups. Kale Seramik makes ceramics at scale; Kalekim makes building chemicals. Both run on SAP, both carry decades of product, contract and operational history.
The question was never whether AI could help. It was where to start. Instead of one pilot, Kale mapped the whole landscape: over 40 candidate use cases, grouped into six domains, spanning operations, finance, commercial and IT.
The Use Case Landscape
Over 40 use cases grouped into six domains across Kale Seramik and Kalekim. Each one reads from the same governed layer, so the answers agree with each other.
Supply chain & inventory
Dead stock moved before it ages, orders proposed before they are asked for.
Finance & controlling
Controllers ask in plain language instead of waiting on the next report cycle.
Pricing & packaging
Commercial teams price from live stock and sales data, not last quarter's sheet.
Product information access
One agent answers from SAP, non-SAP systems and the document library at once.
Contract & document review
Every draft checked against the template before it reaches a signature.
IT operations & audit
The first line of support answers itself, and audit evidence collects itself.
What Makes It Work
One Foundation Under Every Use Case
40 use cases do not mean 40 projects. SAP transactional data, non-SAP product attributes and the technical document library land in the same governed lakehouse. Each new use case is a question asked of data that is already there.
Stock, price and lead time carry one definition across both companies. A pricing agent and a finance report reading the same number is not a coincidence — it is the same column.
Agents Where People Already Work
The same product agent answers on the public site, inside the B2B portal and in Teams. Dealers, sales engineers and internal staff each get the slice they are allowed to see — one agent, one permission model, three front doors.
Documents Treated as Data
Contracts, datasheets and audit evidence are indexed alongside structured records. A deviation from a template, a missing renewal date or a configuration proof for an audit is a query, not a week of reading.