Before AI Uses Your Customer Data, Who Owns It?
SAP and Oracle are both placing AI closer to enterprise data. For UK businesses, the valuable work is resolving conflicting definitions, permissions and freshness before an agent acts.
Data owners and operations directors
5 min read
30 October 2026
2 sources
Decide
Business question
→Locate
Authoritative source
→Check
Access + freshness
→Resolve
Conflicts + owner
→Act
Evidence + trace
Connected records need agreed definitions and accountable owners.
A connected agent can retrieve three customer records quickly; it cannot decide which one is authoritative without a business rule.
- 01Assign owners to consequential data and definitions.
- 02Expose source, timestamp and confidence or uncertainty to users.
- 03Apply the user's access rights across every connected system.
The new interface still depends on old records
Oracle's 27 October announcement describes an AI workspace connecting applications and enterprise data. SAP's TechEd keynote describes a business data layer and tools for adding domain context. Both are vendor visions of connected AI; neither removes the need to determine whether a particular record is correct and accessible to a particular user.
Imagine a UK distributor whose CRM says a customer is active, finance has placed the account on credit hold, and a sales spreadsheet lists an old delivery address. An agent can surface all three. It should not treat the most recently written sentence as an automatic authority to release an order.
Give facts a provenance and an owner
For each high-impact decision, identify the system of record, accountable owner, definition and expected update interval. A credit hold may be controlled by finance; a delivery address may require customer confirmation; order status may depend on a warehouse event. Show the source and last confirmed time, and flag missing or conflicting values.
Keep access controls attached to the data. A conversational interface must not turn an employee's access to one customer account into access to all accounts. Test both retrieval and downstream actions with real role boundaries.
Start with a contested decision
Select one recurring handoff, such as confirming whether an order can be released. Trace the decision across systems, identify conflicts and determine the escalation owner. Add deterministic rules for clear constraints; use AI to explain the evidence or highlight ambiguity where interpretation is genuinely needed.
Measure incorrect releases, time spent reconciling records and unresolved conflicts. If the team cannot agree on the definition of an active hold, an agent rollout will expose that disagreement rather than resolve it.
Use this before approving the work.
- 01Select a decision with a clear owner.
- 02Document authoritative sources and freshness expectations.
- 03Test access rights and conflicting records.
- 04Track corrections and unresolved uncertainty.
Research and references supporting this article. Interpretation and recommendations are Thalenor’s.
The distributor scenario is illustrative; the article's data-ownership recommendations are independent analysis.
Published and reviewed on 30 October 2026. This article provides general technology and operational information, not legal, tax, medical or regulatory advice.