Proprietary AI where it helps. Deterministic math where it counts.
Every number traces to your data.
SpendCraft uses proprietary AI to read, classify and explain your spend, and deterministic logic for every figure you act on. People confirm every decision.
What Proprietary AI Does
Reads, classifies and narrates.
Normalizes vendors and items, classifies every line into your taxonomy, reads documents you upload, and writes labelled explanations that cite the metric behind them.
What Deterministic Logic Does
Produces every number.
Savings ranges, recoveries, counts, Scout severity and Kraljic placement are computed from your data. A language model never generates a figure you act on.
What Your Team Does
Confirms every decision.
Caddies recommend and explain. Nothing changes, sends or commits without a person’s confirmation, and original values stay visible as provenance.
The AI Boundary
A language model never sets a number you act on.
AI-written text is always labelled and cites the metric it rests on. Figures come from deterministic logic, traced to the transactions and files behind them.
Principles
How we keep the AI honest.
Ranges, Not Promises
Savings are shown as Min to Max ranges on addressable spend. Directional, not guaranteed.
Missing Means Missing
If a value is not on file, SpendCraft says so. It never estimates the gap.
Evidence First
Every signal traces to specific data, and every narrative cites the metric that produced it.
People in the Loop
Classification Review and Caddie confirmations keep a person in control at every stage.
Auditable Memory
What SpendCraft remembers traces to recorded events or snapshots, never to labels about people.
Managed to a Framework
AI risk is managed in line with the NIST AI Risk Management Framework.
Questions
What security and finance teams ask.
See the AI boundary on your own data.
A Proof of Value shows every number traced to its source.