The creation of a language to describe feature patterns in an image / data context pattern in a transaction, allowed us to build skill-based Bot Workers, that use a repository of Pattern Definitions. We are currently calling this language as Pattern’s Pattern Definition Language (PPDL).
When a Bot Worker processes a case, using the Standard Operating Procedure (SOP) configured for the case type (class), there are two possibilities: correct result (case processed correctly – STP case) or an erroneous result (Manual Case). In erroneous cases, Bot Worker would have completed more than 80 to 85% of the work & a human will complete the remaining incremental work in the case. While doing so, the human will perform repair / corrections to the incorrect recognitions / decisions taken by the Bot Worker. These manual repairs will become the guiding factors for the Bot Worker system to learn on the job.
A Teacher Bot Worker uses the manual repairs done, as inputs for continuous on-the-job learning. The Teacher will pick all the repairs, bunch them for similarity & select best candidates to auto generate PDs (Pattern Descriptions) & save the PDs in the Pattern Definition Repository. Using this enhanced Pattern Definition Repository, Bot Workers will be able to correctly process cases, which failed earlier.
This Perform-Repair-Learn cycle provides a continuous learning Bot Worker implementation. Combine this with SOP definitions, multiple SOP execution capabilities of a single Bot Worker, Check Listing, Domain Access Definitions, Referral Capabilities & Architecture for Digital Back Office Room & we have a next generation of Digital Workers to provide round the clock processing in complex service areas.
The continuous, self-learning paradigm reduces the Time to Value (TTV) for an organization dramatically.
Patterns has started its exciting journey of offering Action Bot Workers strengthened by an associated Reviewer Bot Worker, which learn continuously on-the-job, guided by Teacher Bot Workers. And probably, this can be a candidate for Next Gen Hyper Automation.
Aug
13