An AI application survives examination when the claims describe a reproducible technical procedure rather than a business outcome. The work is to separate the data pipeline, the model behaviour and the measurable effect from the product narrative.
What we look at first
- the pipeline that collects, processes and scores data
- features of training, inference and result validation
- a technically reproducible output of the AI system
Patent examples
Granted patents in this area. The numbers are public and can be checked against the Rospatent and EAPO registers.
EA 052928 Method for automated processing of employee data using artificial intelligence models An AI approach to HR data with a technically defined pipeline for analysis and output. EA 052862 Method of conducting a candidate interview using artificial intelligence AI interviewing: collecting, processing and scoring candidate answers in a reproducible procedure. EA 050603 Method of processing court documents with a machine learning model ML processing of legal documents: extracting and structuring meaningful information from case materials. EA 051313 Teaching method using an electronic educational resource An EdTech solution that processes user actions and adapts the learning scenario. EA 048442 Data processing method for determining the completeness of enterprise business processes Analysing enterprise data to assess process completeness and detect deviations. EA 044817 Method and system for assessing the quality of AI-enabled digital solutions using generative technology and AutoML Quality assessment of digital solutions using generative technology and automated machine learning.