Glossary
The A-Z of enterprise agentic AI and work context
Accounts Payable (AP) is the business function managing payment obligations to suppliers, covering invoice processing, approvals, and payment execution. Skan observes how AP teams handle invoices to identify manual touchpoints and exception patterns.
An action is the smallest observable unit of work captured by Skan AI, such as opening an application, entering data, or clicking a field. Skan's computer vision classifies actions automatically, transforming raw desktop activity into structured process data that makes previously invisible work measurable.
A group of related actions completing a defined step in a business process. Skan identifies activities automatically from desktop observation, replacing manual process interviews.
The divide between knowing what to do and being able to act autonomously. Gartner coined this to describe GenAI's limitation: informing but not acting. Skan closes this gap by grounding agents in operational ground truth.
An autonomous software system perceiving its environment, making decisions, and taking actions to reach goals without step-by-step instruction. Skan builds agents from operational ground truth and real observed work data.
The capability of AI agents to retain and build on information from previous interactions and tasks, improving performance over time. Skan's context graph provides persistent process memory.
Agent Operating Procedures (AOPs) are structured instructions derived from observed human workflows that govern how AI agents execute enterprise tasks safely and consistently. AOPs replace static standard operating procedures with dynamic, AI-ready frameworks built from real operational data. They define what an agent is permitted to do, when it should escalate, and how it should handle exceptions.
The practice of rebranding existing automation tools as AI agents without implementing genuine agentic reasoning. Gartner coined this term to warn buyers about inflated vendor claims.
An open standard enabling AI agents from different platforms to share context and coordinate tasks securely. Gartner pairs this with MCP as foundational enterprise agentic architecture standards.
An autonomous form of artificial intelligence that independently makes decisions, learns in real-time, and adapts to dynamic environments with minimal human intervention, enabling a shift from human-driven processes to agile operations.
An organization where AI agents autonomously handle routine decisions and workflows, freeing human workers for judgment and strategy. Skan serves as its operational intelligence foundation.
Agentic observability is the ability to monitor, trace, and audit the actions of AI agents in real time as they execute tasks inside enterprise systems. It answers the operational question: what is the agent actually doing, and is it doing it correctly? Agentic observability is the control layer that makes governed autonomy possible in practice rather than in theory.
A standardized vocabulary developed by Skan for classifying human and AI work across enterprise environments. Launched February 2026, it provides a common language for actions, activities, processes, and agents.
The coordination of multiple AI agents, human workers, and enterprise systems to accomplish complex, multi-step goals. Skan's process intelligence provides the operational context orchestration requires.
Enterprise infrastructure coordinating AI agents, human workers, and systems into a governed execution layer. Gartner identifies this as a $550 billion market and critical control plane.
Enterprise automation where AI agents independently plan, execute, and improve complex workflows without predefined scripts. Skan trains agents on operational ground truth for day-one accuracy.
A discipline combining continuous process observation with AI agent deployment into a closed loop, where observed work data powers, governs, and improves agents executing enterprise workflows.
The execution layer of an agentic orchestration platform, carrying out agent plans with speed and cost discipline. It allocates resources, selects optimal models per task, and ensures auditable actions.
Structured, process-level operational data used to train AI agents to execute enterprise workflows accurately. Skan's continuous desktop observation produces comprehensive first-party agentic training data.
The operational practice of managing the full AI agent lifecycle, covering evaluation, testing, deployment, monitoring, drift detection, and improvement. Skan supports this with a work telemetry baseline and observation data.
Continuous observation and analysis of how AI agents execute tasks within enterprises, producing behavioral intelligence used to audit performance, detect drift, and verify conformance. Skan delivers this through desktop observation.
Autonomous software systems perceiving their environment, making decisions, and taking actions to complete goals without step-by-step instructions. Skan builds enterprise agents grounded in operational ground truth.
The set of policies, controls, and accountability structures ensuring AI systems operate reliably and in compliance with regulations. Skan's framework ensures every agent action is bounded, traceable, and auditable.
The organizational structure, governance framework, and capabilities defining how an enterprise manages and scales its AI and automation initiatives. Skan supports this with accurate process baselines.
The application of AI capabilities including machine learning and autonomous decision-making to automate complex business processes that rule-based tools cannot handle. Skan delivers this through agentic process automation.
Actionable findings generated by analyzing data about how work is actually performed. Skan produces these by observing desktop interactions continuously, capturing patterns system logs miss.
The use of artificial intelligence algorithms to automatically discover, document, and analyze business processes, accelerating identification of process variations, inefficiencies, and improvement opportunities.
The process of evaluating and optimizing the collection of software applications within an organization, eliminating redundancies, improving efficiency, and reducing costs by retaining only value-delivering applications.
Documentation of how a business process currently executes in practice, capturing actual steps, exceptions, and deviations rather than intended design. Skan generates these automatically from desktop observation.
Attended automation supports human users in real time; unattended automation executes complete processes independently. Skan's observation identifies which steps require human judgment.
The process of identifying which tasks and workflows are best suited for AI or automation deployment. Skan scores every candidate automatically by observing real work across all applications.
A framework defining the stages organizations progress through advancing from basic RPA toward autonomous AI-driven operations. Skan anchors maturity assessment in process intelligence.
The end-to-end sequence for moving an automation opportunity from identification through deployment and monitoring. Skan compresses discovery and design phases by automatically observing and scoring candidates.
The quantified return from an automation investment, comparing cost savings and productivity gains against implementation costs. Skan's Unit Cost of Work metric provides an accurate process baseline.
An organization where AI agents handle routine decisions and workflows autonomously, freeing the workforce for judgment and strategy. Skan supports this with accurate, observed process data.
The ability of a system to perform tasks independently without human intervention. In business terms, it means automating repetitive tasks and workflows within processes.
The process of identifying stages in a workflow or system that slow down operations or create inefficiencies, helping organizations pinpoint and address obstacles hindering optimal performance.
A standardized visual language for modeling enterprise processes using defined symbols to represent steps, decisions, and handoffs. Skan can validate BPMN documentation against observed execution behavior.
Technology-enabled execution of recurring business processes with minimal human intervention, from simple rule-based scripts to AI agents handling complex decisions. Skan provides the process intelligence foundation.
The capability to analyze and generate insights from enterprise process data supporting operations decisions and transformation programs. Skan delivers this by combining desktop observation with AI analytics.
The discipline of designing, executing, monitoring, and continuously improving the operational processes that run an enterprise. Skan adds an observation layer replacing documented assumptions with real-time data.
A team that governs and scales an organization's automation and AI initiatives. Skan provides the operational ground truth CoEs need to prioritize investments with evidence.
Process compliance is the verification that business processes are executed in line with regulatory requirements and internal policies. Skan monitors compliance in real time by observing actual work execution.
AI technology that interprets visual information, including screen content. It is the foundational technology behind Skan's process intelligence platform, enabling observation of desktop activity.
The minimum certainty level an AI agent must reach before acting autonomously without human review. Above threshold the agent acts; below threshold the workflow escalates to humans.
Comparison of how a business process actually executes against its intended design, identifying deviations and compliance gaps. Skan extends this beyond ERP event logs by observing desktop-level work.
The discipline of designing and managing the information context that AI agents need to reason and act accurately. Skan automates this for process agents through work telemetry and context graphs.
A context graph of work is a dynamic, structured map of how tasks, people, applications, and decisions connect across real enterprise workflows, built from observed work telemetry rather than inferred from system metadata or organizational charts. It captures the actual sequence of activity across an enterprise, not the documented or intended version. This makes it the operational context layer that AI agents need to act accurately and consistently at scale.
A real-time, structured representation of how work actually happens across an enterprise, capturing the flow of actions, decisions, and handoffs rather than just system events. Skan builds this from continuous desktop observation.
Graphs that capture the directional flow of decisions, workflows, and actions across an organization, not just entities and relationships. AI agents use them to understand how work flows. Gartner predicts 50% adoption by 2028.
The practice of providing AI agents with accurate, current information about a specific operational environment before they reason or act, preventing hallucinations. Skan provides this through its context graph.
The ongoing effort to enhance products, services, or processes incrementally. It involves gathering feedback, making adjustments, and striving for excellence over time.
Skan's compliance monitoring feature applying an Identify-Map-Monitor-Report-Remediate methodology to regulated workflows and approval processes. It generates real-time alerts and audit evidence.
The capture and analysis of data generated by how work is actually performed. Skan collects first-party data directly from the desktop, revealing manual workarounds system logs cannot surface.
Measures the average days an organization takes to pay suppliers after receiving goods or services. Skan identifies specific P2P workflow bottlenecks extending payment cycles.
Measures the average days to collect payment after a sale, reflecting order-to-cash process efficiency. Skan identifies specific collection workflow steps extending DSO.
Records of the reasoning pathway through which an agent or human arrives at an action, including which data and logic contributed to the outcome. Skan captures this through process observation.
Answers the question: what actually happened? Skan delivers this by recording every process step in real time, producing an accurate picture of how work flows.
Real-time capture of user interaction data at the screen level, including application navigation, data entry, and task execution. Skan collects this through computer vision.
Identifies the root cause of a performance problem or process deviation. Skan performs this by correlating observed work data across applications and roles.
Measures how effectively employees use the technology tools an organization has invested in. Skan quantifies this by observing actual application usage and identifying feature utilization.
Continuously tracking how enterprise applications are actually used across the workforce, including detection of unsanctioned AI tools. Skan surfaces shadow AI inventory without requiring endpoint agents.
The shift to AI-powered, data-driven operations that let enterprises move faster and outperform competition. Skan provides the process intelligence foundation making transformation measurable.
A continuously updated virtual model of how an enterprise actually operates, built from observed work data rather than assumed process maps. Skan constructs this from continuous action capture.
The practice of observing and structuring exactly how employees interact with enterprise applications as the primary input for process intelligence. Skan's computer vision is purpose-built for this.
Building structured representations of how work flows within a specific enterprise business domain, capturing processes, decision rules, entities, and relationships. Skan constructs these from continuously observed work data.
The application of artificial intelligence technologies within organizational settings, including AI-driven solutions tailored to improve decision-making, operational efficiency, and customer experiences.
Strategic redesign of how an organization operates and creates value in the age of AI. Skan provides an accurate operational data baseline and evidence of where AI will deliver the highest return.
A structured record of timestamps and activities captured by enterprise systems, used as the primary input for process mining analysis. Event logs miss manual steps and decisions between system touchpoints.
Distributes AI governance across business units, with a central CoE setting standards while local teams execute. Skan supports this with a unified process intelligence platform.
Work activity data captured directly from an organization's own operations through real-time desktop observation, not inferred from third-party system logs. Skan produces this continuously.
Describes the shift to AI-augmented enterprise operations, where agents handle routine decisions and humans focus on judgment and strategy. Skan supports this transition through work telemetry.
A subset of artificial intelligence focusing on creating new content or data by learning patterns from existing data. It uses algorithms such as GANs and transformers with applications in various fields.
Governed autonomy is the operating model for enterprise AI in which human oversight is built into policy design upfront rather than inserted into each individual transaction. Instead of requiring a human to approve every agent action, enterprises set clear boundaries: what agents can do, when they can act, and what requires escalation. Governed autonomy requires a precise understanding of how work actually executes before policy boundaries can be set accurately.
The principle that AI agents operate independently within explicitly defined boundaries, escalating to human judgment when actions exceed authorized thresholds. Skan sets clear operational boundaries for agents.
Specialized AI agents that monitor, evaluate, and constrain the behavior of other agents, functioning as an autonomous oversight layer. Gartner predicts adoption by 2029.
A design principle of embedding specific human review points within automated workflows, ensuring AI acts with oversight where judgment matters. Skan implements this through confidence thresholds.
Combines system-based process mining from ERP event logs with desktop task mining from observed user activity, providing a complete operational picture. Skan delivers both natively.
A strategy combining RPA, AI, and process mining to automate as many business processes as possible. It has evolved toward agentic process automation where agents reason and adapt.
Combines AI with process automation to handle judgment-intensive tasks beyond the reach of traditional RPA. Skan provides the process intelligence foundation making intelligent automation genuinely intelligent.
Extracts, classifies, and validates information from unstructured documents such as invoices and claims forms without manual data entry. Skan identifies where IDP deployment will eliminate the highest-volume work.
Combines RPA with AI capabilities to handle complex, judgment-intensive tasks. It has evolved into agentic process automation. Skan bridges both stages by grounding automation in continuously observed data.
Enterprise activity occurring between system touchpoints and never captured in any system of record, including manual steps, data reconciliation, and exception handling. Skan makes this visible.
A structured representation connecting people, processes, applications, decisions, and outcomes into a queryable model of how an organization performs its work. Skan's context graph extends this with real-time decision traces.
AI foundation models trained on human application interaction data to understand and execute complex multi-step tasks within graphical user interfaces without explicit scripting. Skan creates training data for these.
Large Action Models (LAMs) are AI models trained to take actions inside enterprise systems rather than generate text. They enable AI agents to execute real operational tasks end to end, from navigating applications to completing multi-step workflows. The quality of the observed work telemetry used to train a LAM determines how accurately it can replicate or improve on human workflows.
AI foundation models trained on large-scale structured process data to understand and reason about enterprise workflows. Skan's observation creates first-party process datasets for robust LPM training.
Records how each decision or output in an agentic system was produced, documenting the data, models, and reasoning that contributed to it. Skan's observation provides the foundational data.
A continuously updated operational model reflecting how an organization actually functions today, not a static snapshot. Skan creates this by observing work at the desktop level.
Maps business processes through workshops, interviews, and manual observation, capturing only idealized snapshots of how processes should run. Skan replaces this with continuous desktop observation.
An open standard for how AI agents securely access enterprise tools and systems through a standardized interface. Gartner endorses MCP as essential to avoid vendor lock-in.
An architecture where multiple autonomous AI agents collaborate, each with a defined role, to accomplish goals too complex for a single agent. Skan's context graph provides shared operational knowledge.
Combines neural network pattern recognition with symbolic rule-based reasoning in a single architecture. Skan uses this approach to deliver accurate and auditable process intelligence.
Analyzes how multiple business objects, such as orders, invoices, and customers, interact within a single process rather than following one linear case. Skan adds a desktop observation layer.
Captures how users interact with multiple business objects simultaneously in a single work session, reflecting the reality of knowledge worker multitasking. Skan's desktop observation captures this.
Foundational infrastructure that captures and continuously updates an organization's real-time work activity, providing AI agents and analytics with an accurate operational picture. Skan constitutes this layer.
Continuous improvement of processes and systems to achieve sustainable business growth, efficiency, and customer satisfaction. It focuses on eliminating waste and optimizing resources.
Operational Ground Truth is the verified, observed reality of how work flows inside an enterprise, captured from actual system activity rather than self-reporting, documentation, or manual observation. It gives AI agents, process teams, and enterprise leaders an accurate picture of what happens at the task level across every system and workflow. Without it, AI agents operate on assumptions rather than facts.
An accurate, real-time picture of how work is actually performed, captured from desktop observation rather than inferred from system logs or interviews. Skan establishes this continuously.
Continuous, real-time visibility into how enterprise operations are actually performing. Skan delivers this by observing work at the desktop level, capturing telemetry system dashboards cannot.
Defines how work is structured, how agents interact, and which policies govern decisions across all automated workflows. It translates business intent into repeatable execution blueprints.
The end-to-end process from customer order through to payment receipt, covering order management, fulfillment, invoicing, and cash application. Skan's O2A platform takes its name from this transformation.
Uses data to forecast future process performance, such as when a bottleneck will form or an SLA will be breached. Skan applies predictive models to continuously captured work telemetry.
Recommends the specific actions that will produce the best operational outcome. In Skan, it identifies not just where a process underperforms but exactly what to change.
Captures process-level activity data without collecting personally identifiable information about individual employees. Skan masks all PII at the point of desktop capture.
Systematic examination of a business process to identify areas for improvement. It involves techniques like process mapping, process mining, and unit cost analysis.
Applies data analysis to process performance data to identify inefficiencies, measure cycle times, and quantify improvement opportunities. Skan delivers this from first-party desktop observation.
Uses technology to execute recurring tasks or processes in a business where manual effort can be replaced, increasing efficiency, reducing errors, and freeing workers for strategic activities.
Compares process performance within an organization against internal data across teams, industry standards, or peers. Skan enables benchmarking with observed performance data.
Integrated infrastructure connecting, harmonizing, and continuously updating process-level operational data from across an enterprise's applications and workflows. Skan builds this through desktop observation.
A structured specification of a business process including each step, decision points, application interactions, and exception paths, used as a reference for automation development. Skan generates these automatically.
A virtual model of a business process that continuously mirrors real-world execution using live operational data. Skan builds these from desktop observation, capturing human activity and workarounds.
Identifying and mapping out business processes using data and analytics. It helps understand how customer orders are processed, identify bottlenecks, and optimize workflows.
The gradual deviation of a business process from its intended design, typically invisible until performance has already degraded. Skan detects process drift in real time by comparing observed patterns.
The use of AI and advanced analytics to monitor, visualize, and analyze business processes in real-time, helping organizations gain an end-to-end view, identify inefficiencies, and optimize workflows.
A visual representation of a business process, outlining the sequence of activities, decisions, and agents involved. Often flowcharts or diagrams helping visualize and improve business processes.
Creates a visual representation of how a business process flows, documenting steps, decision points, and handoffs. Skan generates maps automatically from observed desktop activity.
An advanced process discovery technique that analyzes event logs to visualize and understand business processes in detail, identifying inefficiencies and improving customer experience.
Operational scenarios where analyzing event logs delivers value, including conformance checking, bottleneck analysis, and automation discovery. Skan extends standard use cases with desktop observation.
A more formal representation of a business process, often created using specific modeling languages or software tools. It can be more detailed than process maps and used for simulation and analysis.
Improving business processes to achieve higher efficiency, lower costs, and better-quality outcomes. It encompasses redesigning workflows, eliminating redundancies, and adopting best practices.
Coordinates automated tasks, AI agents, human activities, and system interactions into an end-to-end workflow that delivers a defined business outcome. Skan identifies optimal orchestration design.
Coordinates automated tasks, agents, and human activities from a central control point across an end-to-end workflow. Skan's process intelligence complements orchestration platforms.
Evaluates the current state of a business process, determines the next best action, and selects the appropriate agent to execute it. Skan functions as a process reasoning foundation.
Models a business process digitally and runs scenario tests to predict how changes would affect performance before production implementation. Skan provides a simulation foundation grounded in real data.
Full, real-time visibility into how work actually happens across an enterprise, not just how it is documented. Skan achieves this through continuous desktop observation.
The different execution paths a business process takes in practice. Skan captures all variants, including those occurring entirely outside system logs, by observing work at the desktop level.
AI agents trained on observed enterprise process data rather than generic instructions, understanding how work actually flows including exceptions and workarounds. Skan builds these from its context graph.
Skan's generative AI capability letting users query process intelligence data through natural language, with answers grounded in first-party observed process data.
Keeps process documentation and standard operating procedures (SOPs) readily available at your fingertips, improving efficiency and reducing errors by ensuring teams have the latest process information.
Covers all activities from identifying a purchasing need through to supplier payment, including requisition, approval, purchase orders, goods receipt, and invoice processing. Skan observes the complete P2P workflow.
Uses software bots to execute repetitive, rule-based tasks by mimicking human interactions with digital systems. RPA breaks when processes change and cannot handle unstructured inputs.
Specific processes suited for software bot automation, typically high-volume, rule-based, and stable. Skan eliminates guesswork by automatically scoring every observed process.
Workflows and agents that automatically adapt when underlying applications or processes change, without requiring manual re-scripting. Skan detects process changes in real time.
The agentic execution layer of the Skan AI platform, deploying AI agents built on process intelligence. Agents operate inside deterministic playbooks with embedded guardrails and a full audit trail.
The strategic discovery and assessment layer of the Skan AI platform, designed to give C-suite leaders a quantified business case for AI transformation. It produces a board-ready investment roadmap.
The operational intelligence layer of the Skan AI platform, providing CISOs and operations leaders with continuous visibility into how work is actually executing, capturing process variants and compliance adherence.
A documented set of steps required to complete a task consistently and correctly. Skan automatically generates and updates SOPs from observed work data.
Work activity data captured through real-time desktop observation and transformed into an analyzable format with defined schemas and classifications. Skan converts raw activity into labeled data.
A platform that actively directs work and executes decisions rather than just recording events. Gartner introduced this to distinguish agentic orchestration platforms. Skan's O2A is an SOA.
Captures user interactions with software applications to analyze and identify patterns in task execution. It provides a narrow, task-level view of processes, ideal for low-complexity RPA.
Analytics derived from observing how employees actually interact with enterprise applications. Skan captures application usage at the desktop level, revealing system utilization patterns.
Handles supplier invoices from receipt to payment authorization entirely through automated systems without manual intervention. Skan identifies exception patterns preventing straight-through processing.
Undocumented process expertise held by experienced employees, including informal workarounds, unwritten decision rules, and expert shortcuts never appearing in official documentation. Skan captures this.
The total elapsed time to complete a process from start to finish. Skan measures TAT from observed work data, identifying exactly which step or queue is responsible.
A single, end-to-end picture of how a business process flows across all applications, systems, and teams. Skan builds this through desktop observation, connecting every action.
Ensures AI agents and applications share consistent definitions of business terms and process concepts across an enterprise. Skan provides the process-specific semantic layer.
Determines the true cost of completing one unit of work within a business process. Skan enables precise measurement by observing actual time and steps at every stage.
A Skan metric calculating the observed cost of completing one unit of work, including labor time, application interactions, and process steps captured from the desktop, providing an accurate baseline.
Information that does not conform to a predefined schema, including documents, emails, and screen content. Skan's computer vision extracts structured process intelligence from unstructured activity.
Skan's desktop observation agent is a lightweight software component installed on a user's device using computer vision to observe workflow activity in real time. All PII is masked at capture.
The ability to model the impact of a proposed process change before implementing it in production. Using Skan's digital twin of operations, leaders can simulate redesigned workflows.
Continuous capture and analysis of data about how human work is performed across an enterprise. Skan observes every action across every application, producing structured insights.
Continuous capture of data signals from how employees interact with applications and workflows. Skan collects this through desktop observation, transforming interactions into structured intelligence.
Automates repetitive tasks and workflows using technology, such as AI and robotic process automation, aiming to increase productivity, reduce errors, and accelerate execution.
Data-driven analytics derived from Skan's process intelligence platform, providing valuable information about how employees work, measured through key metrics.
Uses process observation data to understand how work is performed across teams and identify efficiency improvements. Skan takes a privacy-first approach with all PII masked.