AA Advanced RPA flashcards

32 cards. Read the term and the definition here. The deck at the top of this page is optional.

RPA vs AI decision rule

Use RPA when decision logic is expressible as explicit rules on structured input; use AI when input is unstructured and labeled examples exist for training.

Precision formula

Precision equals true positives divided by the sum of true positives and false positives; it measures how often a positive prediction is correct.

Recall formula

Recall equals true positives divided by the sum of true positives and false negatives; it measures how many actual positives the model captures.

Confidence threshold routing

AI output above an upper threshold auto-processes, between thresholds goes to human review, and below a lower threshold is rejected or escalated.

Human-in-the-loop design

A pattern where the bot pauses and queues a work item for human review, then resumes with the reviewer's decision; required for regulated or high-value decisions.

Straight-through processing rate

STP rate equals transactions completed without human touch divided by total transactions; it is the primary throughput KPI for hybrid automation programs.

Supervised learning prerequisite

Supervised learning requires labeled training data mapping inputs to known outputs; without sufficient labeled examples, classification accuracy will be unreliable.

Document understanding pipeline

A chain combining OCR, layout analysis, and classification to extract structured fields from invoices, forms, and contracts for downstream processing.

Process mining vs task mining

Process mining reconstructs actual process flows from system event logs; task mining observes desktop events to identify candidate automation steps.

Exception handling pattern

Retry transient failures with exponential backoff, then route persistent failures to a queue with structured error context for human or bot remediation.

Queue-based decoupling

Producer bots enqueue work items and consumer bots process them, enabling horizontal scaling, load leveling, and isolation of failures per item.

Attended vs unattended automation

Attended bots are triggered within a user session; unattended bots run on servers via schedule or trigger; hybrid models hand off between them using queues.

Stable selector strategy

Prefer automation IDs and accessibility names over screen coordinates or index positions, which break when UI layout or ordering changes.

Idempotency requirement

Any bot that may retry must use a transaction ID or dedupe key so reprocessing the same item does not duplicate downstream effects.

Process selection scoring factors

Candidate processes are scored on volume, standardization, exception rate, cycle time, and regulatory risk; high-volume, low-variation, rules-based processes rank highest.

Cognitive automation pipeline stages

Ingest, extract via OCR, classify, validate against business rules, post to system of record, and log confidence and outcome for each transaction.

Layered bot architecture

Entry trigger, orchestrator for queues and workflow, reusable component library, connectors for UI/API/database, and logging telemetry layers.

Bot state model

States such as queued, in-progress, succeeded, business exception, and system exception persisted outside the bot process so restarts are safe.

Secure configuration practice

Credentials, endpoints, and thresholds belong in a secure config store; secrets go in a vault referenced by key and are never hardcoded.

Required log fields

Each log entry captures timestamp, transaction ID, step name, outcome, and error code; these feed operational dashboards and audit evidence.

Deployment promotion path

Dev to test/UAT to production with versioned packages and a rollback path; promotion requires passing regression test suites.

Environment parity risk

Screen resolution, locale, OS version, and application versions must match between test and production or selectors and timings break.

Reusable component regression

Shared components are unit-tested and versioned; changing one triggers regression testing across every bot that consumes it.

Orchestrator core capabilities

Queue management, credential vaulting, scheduling, bot health monitoring, and centralized audit logging are the baseline orchestrator functions.

CoE operating models

Centralized (CoE builds all), federated (business units build, CoE sets standards), and hybrid (CoE owns platform and reusable assets, BUs own pipelines).

Governance gate sequence

Intake and assessment, design review, build standards compliance, UAT sign-off, production release, and periodic health review.

Bot inventory registry

A registry capturing owner, criticality tier, dependencies, and last-reviewed date for every bot; the minimum control for a growing estate.

Production change control

Any change to a production bot requires a change request, impact assessment on dependent bots, and re-test evidence before release.

Criticality tier monitoring

Tier-1 bots that are customer-facing or financial get real-time alerting and on-call coverage; lower tiers get periodic review.

Scaling metrics set

Track pipeline conversion rate, bots in production, STP rate, exception rate, hours saved, and cost per transaction to measure program scale.

Role separation in CoE

Developers build, business analysts define requirements and process maps, and control functions review compliance; avoid one person owning all for critical bots.

Capacity planning for scale

Plan concurrent bot execution capacity, peak queue volumes, and orchestrator licensing before expanding the automation pipeline.