Reliable AI starts with reliable data
Bad data leads to wrong decisions, failed integrations and inconsistencies between systems. With AI, “garbage in, garbage out” becomes “garbage in, fantastic garbage out”. The most expensive part is often not the fix but the hours spent finding the problem: a single malformed date among millions of records can delay a critical dashboard by weeks.
At Code10 TechTalks 2026, Antonio Valladares hosted a panel with Damjan Arsovski, Digital Project Manager, and Manuele Borghin, Data Architect, to present the C10 Data Cleaner, Code10’s approach to data quality.
Detect, fix and learn
The solution works in three steps:
- Detect: an observability layer validates incoming records against data contracts stored in a central GitHub repository.
- Fix: when a violation appears, the remediation layer looks for a matching data rule and proposes a fix in a simple cockpit, where a data steward approves changes in bulk.
- Learn: when no rule exists, an AI agent suggests one. Once approved, it joins the repository and handles every future case automatically.
Fixes are sent through a common API to the customer’s middleware, which adapts them to SAP, Salesforce, Workday or legacy systems.
Governed AI, by design
AI is never used to fix the data directly; it only proposes rules. That keeps results deterministic, avoids hallucinations and keeps token costs low. Every decision stays with people, and everything is managed as code.
An open debate
The panel also discusses two questions every data team faces. Should data be fixed at the source, or only flagged? And who should own data contracts, the business or engineering?
Talk to our experts
Struggling with data quality? Contact us to see how the C10 Data Cleaner can help.
