AI Data Stewardship

The clearMDM Data Steward Agent roadmap introduces a progressive capability stack; Data Quality, Data Governance and AI Data Stewardship, helping Enterprise organisations continuously improve and manage trusted Salesforce CRM data.

The Spring ’26 release extends the AI Data Stewardship Agent capabilities with AI-guided Duplicate Resolution and Automated Data Stewardship Decisions.

AI-guided Duplicate Resolution enables AI match scoring and reasoning to be integrated into all matching operations irrespective of processing type (API, Batch, Real-time, Flow, Event etc.). This provides data stewards with a second opinion and additional context to guide the next action taken (accept, reject, further investigation etc.).

The Automated Data Stewardship Decisions feature takes the AI-guided duplicate resolution process a step further with the AI match scoring and reasoning being used to automate the determination of the next action taken either at the point-in-time of record matching or subsequent bulk processing of existing matches. This latter case can offer significant manual time savings where a backlog of data stewardship tasks exists.

For example, the first time a set of Salesforce CRM Contact is matched, the first stage of duplicate resolution, a high volume of candidate level matches may be generated (relative to the day-to-day operational volume). This exposes the tension between the Sensitivity and Selectivity of the matching algorithm.

• Sensitivity. Set the match score threshold low to ensure that True Positive matches are not missed and accept a higher tolerance for False Positives.
• Selectivity. Set the match score threshold high and accept that True Positive matches may be missed but volume reduces due to a lower tolerance for False Positives.

Generally, confidence in the matching algorithm evolves over time through empirical evidence and as such high sensitivity is preferred initially and a high volume of candidate matches are generated for human data stewards to review and perform the selectivity aspect (i.e. to identify and reject the False Positives). Consider a dataset of 2M Salesforce Contacts being matched for the first time with a 5-10% level of duplication to get a sense of the potential scale of manual review effort. This is where the AI Data Steward Agent can step in and perform an initial review at scale using custom instructions and thresholds to reduce the backlog. The time and cost savings here can be significant to both ad-hoc bulk processing and also day-to-day operational data stewardship.