Enhanced AI Search
Enhanced AI Search is the capability the screening modules are built on. It is what lets a match survive a spelling, a transliteration or a reordered name, and what keeps the resulting alert count workable.
The DetectX® approach
Several matching techniques run together rather than in sequence: fuzzy string matching for partial and approximate matches, probabilistic entity matching and deduplication across disparate datasets, and vector-based matching with natural language processing for results a keyword search cannot reach.
Scoring is applied on top, so a result arrives with a risk score and a priority rather than as an undifferentiated hit list. Data is processed as it streams, so a result is available when the decision is being made.
- Fuzzy string matching
- Partial and approximate matches, which is where inconsistent source data actually lives.
- Entity matching
- Names and transactions resolved across datasets that do not share an identifier.
- Semantic and contextual search
- Vector matching and NLP, for relevance a keyword cannot express.
- Pattern recognition
- Patterns across large datasets that conventional methods pass over.
- Hierarchical clustering and LSA
- Latent Semantic Analysis and machine-learning-enhanced scoring, used to surface relationships that are not stated in the data.
What it changes
The cost of screening is not the matching, it is the review. Matching that scores and prioritises is what decides how much of an analyst's day is spent on results that were never going to be alerts.
- Scale without a second system
- The same layer serves a single transaction screen and monitoring across global operations.
- Relationships, not just records
- Hidden connections between entities are surfaced as part of the search rather than found in a later investigation.
Why DetectX®
Every module that screens, scores or profiles reads through this one layer. That is why a name matched by Name Screening and the same name matched by Transaction Filtering agree: there is one matching implementation, not one per module.
Regulatory requirements
DetectX® supports the FATF and GAFI recommendations, and is used against Swiss regulatory requirements including GwG, FINMA-GwV, VsB 16, KAG, FinfraG, FidleG, BEHG and KKG.
It also supports MiFID and EMIR reporting obligations, and BASEL II and III requirements.
See Enhanced AI Search on your own data
A demo runs against a scenario you choose, so the alert volume and the match quality are yours rather than ours.
Request a demo