The second element of the Capability Method — what keeps a scored domain from drifting back to where it started.
A baseline score is a photograph, not a fix. STEER IT is what runs continuously after SEE IT, so the estate doesn’t quietly decay again the month after the assessment ends.
What It Actually Is: AI Housekeeping
Routines that run to schedule over every record — not a one-off cleanup project, and not a person manually checking a sample. They catch what never reached the system of record in the first place, and hand the genuinely hard cases to a person rather than guessing.
The Three Routines
- Semantic field validation — catching records that are technically complete but substantively wrong
- An operator co-pilot — support at the point of entry, not just a correction after the fact
- Duplicate detection & cleanup — checked each cycle, not case by case
The Human Boundary
AI analyses, identifies patterns, and prepares a decision. It never approves a record, accepts a risk, or replaces a governance decision — a person always does. This is a standing position, not something softened per client.
What You Receive
- The housekeeping routines piloted on real data from your own domain, not a demo dataset
- A baseline count of cases still routed to a person for judgement — the number this element is designed to bring down
- A clear line between what the routines catch automatically and what still needs a person, so the boundary is never ambiguous
The Proof — an Honest AI Account
Every capability flagged for AI housekeeping was checked against SAP’s own published feature set, not assumed. As it stands today: 55% map to a real, named SAP AI feature already running — among them SAP Joule, Integration Advisor and SAP AI Core. The other 45% run on deterministic, rule-based logic — fuzzy matching, business rules — that SAP doesn’t market as AI at all. That’s a confirmed negative, not an unchecked gap, and it will shift as SAP ships more. We’d rather tell you which is which than round up.
Separately, every AI-housekeeping verdict was checked against one fixed rule across all eleven domains: same capability, same verdict, every time it recurs. Zero exceptions found.
Where This Starts
STEER IT is what the Business Case package pilots and the Value Realisation package puts in motion: a piloted routine, a measured dependency on manual judgement, and a clear boundary for what AI is and isn’t allowed to decide.
Related: Digital Transformation · Inside the Method · Home
