Modern enterprise operations demand unprecedented visibility into business workflows and process optimization capabilities. Process digital twins address this need through sophisticated virtual replicas that mirror operational activities, delivering efficiency improvements of up to 15% and cost reductions ranging from 20-30% across implementations. Current adoption patterns show 70% of C-suite technology executives at large enterprises actively exploring and investing in digital twin capabilities. The global market trajectory reflects this momentum, with annual growth rates approaching 60% and projections reaching $73.5 billion by 2027.
These market dynamics underscore the strategic importance of understanding process digital twin architecture and its technical components. The technology enables organizations to construct accurate virtual representations of business processes, moving beyond traditional monitoring approaches toward predictive, simulation-driven process management. This blog examines the complete architectural framework of process digital twin technology, from foundational data layers through integration systems and user interfaces. Organizations seeking to implement effective digital transformation initiatives will find practical guidance for building robust virtual process models that deliver measurable operational improvements.
A process digital twin functions as a comprehensive digital model that captures business operations rather than focusing solely on physical equipment characteristics. Traditional digital twins concentrate on monitoring manufacturing assets or individual products, whereas process digital twin technology constructs dynamic, data-driven virtual replicas of complete operational workflows, encompassing resource allocation, material flows, decision logic, and complex interactions between system components.
According to IBM, digital twins enable continuous monitoring, simulation and analysis over the course of a lifecycle, from design to decommissioning. The Digital Twin Consortium establishes a technical definition of this technology as an integrated data-driven virtual representation of real-world entities and processes, with synchronized interaction at a specified frequency and fidelity.
From a practical implementation perspective, process digital twins distinguish themselves through their data sources. Rather than relying on IoT sensors monitoring physical equipment, they extract operational intelligence from business systems through event logs and transaction records. For example, in an order-to-cash process, the virtual model captures every step—from initial order placement through final payment receipt—enabling real-time identification of bottlenecks and operational inefficiencies.
Process digital twins establish bidirectional data flow between virtual representations and their physical counterparts. This two-way communication capability distinguishes them from conventional monitoring systems that provide one-directional visibility. The virtual model extends beyond passive observation to actively influence decision-making processes and trigger automated actions within the operational business environment.
This distinction is critical because digital shadows—virtual models that mirror physical processes for monitoring purposes only, without the ability to feed insights back into operations—primarily enable monitoring and observation, while digital twins extend these capabilities by analyzing patterns, running simulations, and supporting decision-making that can directly influence physical operations. The addition of bidirectional data flow creates a continuous feedback loop, where changes in either the virtual or physical environment are reflected in the other in near real time.
McKinsey research demonstrates that organizations can increase decision-making speed by up to 90 percent using digital twin insights. System twins go a step further, modeling the intricate interactions between physical and digital processes across manufacturing operations, end-to-end supply chains, and complete customer journeys.
The temporal characteristics of data synchronization significantly impact process digital twin examples and their practical deployment scenarios. Real-time digital twins process incoming data instantly or within sub-second intervals, continuously reflecting current system states and enabling immediate operational responses. According to research in ScienceDirect, response times on the order of milliseconds to seconds are desired in many applications. Industrial manufacturing processes where real-time monitoring ensures safety and operational efficiency may target response times of 100 milliseconds or less.
Consider the difference in practice: A manufacturing line detecting equipment vibrations that signal imminent failure requires real-time processing to prevent costly damage. Conversely, a supply chain digital twin optimizing delivery routes can operate effectively with hourly data updates, as routing decisions don't require split-second precision. The key is matching synchronization frequency to business criticality.
Real-time systems prove essential for critical operations where processing delays create operational risk or financial loss. Industrial safety systems, autonomous operations, and traffic signal optimization require instant data processing and immediate response capabilities. Near real-time approaches work effectively for asset performance monitoring, urban planning systems, and maintenance scheduling where decisions don't require split-second response times.
A balanced approach often proves most practical for enterprise implementations, using real-time updates for mission-critical operations while employing near real-time processing for planning and optimization functions. This hybrid strategy ensures system scalability without incurring excessive infrastructure costs. Real-time systems demand high bandwidth, low latency, and scalable processing capabilities, which increases infrastructure requirements and operational complexity substantially.
Understanding these operational considerations provides important context for how process digital twins are architected. The ability to balance real-time responsiveness, scalability, and complexity directly influences how systems are structured across data, business logic, and infrastructure layers.
Process digital twin technology operates through multiple interconnected architectural layers that collaborate to create functional virtual representations. Think of these layers like a building: the foundation (infrastructure), the framework (data processing), the interior systems (business logic), and the visible exterior (user interfaces), all protected by security measures throughout. While different implementations employ varying architectural depths, the fundamental layer-based structure remains consistent across process digital twin platforms.
The presentation layer encompasses all user interfaces and visualization modules that enable human interaction with digital twin systems. Research published on digital twin architecture indicates this layer converts complex analytical outputs into interpretable visuals, alerts, and executive reports. Core components include real-time dashboards, KPI and performance views, alert and notification systems, and analytical reports that allow operators to comprehend process performance patterns.
Interactive dashboard systems present twin-generated insights in formats tailored to different organizational roles, from field engineers to executive leadership. The visualization module manages data delivery to end users, ensuring dashboard functionality and command execution operate reliably. This presentation layer essentially bridges raw analytical forecasts with human decision-making capabilities, establishing seamless interaction between virtual models and physical operations.
The business layer addresses user requirements and exposes core functionalities through structured API frameworks. This component interprets incoming data streams and converts them into actionable insights using business rules, simulation models, or artificial intelligence algorithms Pattern recognition, anomaly detection, and performance prediction capabilities operate based on specific business requirements and operational objectives.
Workflow systems and API structures facilitate data exchange across diverse sources, maintaining synchronization between virtual models and their physical counterparts. The service layer coordinates with access control components that regulate data flow through secure gateway systems. This architectural approach ensures process digital twins respond appropriately to business logic requirements while maintaining controlled inter-layer communications.
The data layer manages information collection and organization from physical environments, performing data cleansing, synchronization, and structural formatting for reliable system utilization. According to platform documentation, this layer incorporates data governance, integration, visualization, modeling, connectivity, and control capabilities.
Data storage employs hybrid approaches to serve dual purposes: time-series databases (such as InfluxDB) capture high-frequency sensor data and operational metrics, while data lake solutions (like Snowflake or Databricks) store massive historical datasets for trend analysis and machine learning training. This architecture maintains heterogeneous data repositories containing information about data sources, ontologies, user profiles, and essential metadata—essentially creating a comprehensive information ecosystem that supports both real-time monitoring and long-term analytics.
The infrastructure layer provides physical or virtual server environments and supporting computational resources that establish the foundational platform. Technical documentation indicates hybrid infrastructure configurations combine cloud and on-premise elements to support continuous integration and delivery workflows. This configuration enables machine learning module development and training while maintaining required performance standards.
Cloud services manage analysis operations and store substantial data volumes generated by process digital twin systems. Physical infrastructure components include network routers, actuators, IoT sensor arrays, and edge computing servers that form the hardware foundation. The infrastructure must accommodate high-frequency data transmission using messaging protocols such as MQTT, OPC UA, or Apache Kafka to stream telemetry into the digital ecosystem.
The security layer implements essential security measures across all architectural layers. Access control systems incorporate gateway components that regulate data flow between layers. Platform architecture documentation shows the governance and operations module ensures data structuring and on-demand availability while maintaining appropriate security protocols.
This layer addresses dual security concerns: protecting the digital twins themselves and enabling security through twin-generated insights. Cybersecurity hardening, logging, monitoring, detection, and response mechanisms form integral design elements throughout the system lifecycle. The security architecture safeguards sensitive process data while enabling authorized personnel to access digital twin insights for operational decision-making.
Process digital twin effectiveness depends entirely on four foundational data components that collectively enable accurate virtual representations. Data quality determines whether these sophisticated models deliver actionable operational insights or remain unused analytical artifacts within enterprise technology environments.
Event logs function as the primary data artifact for process digital twins, systematically capturing every activity occurring within business workflows. Process digital twins extract information through event logs and transaction records from business data objects, distinguishing them from physical digital twins that rely on IoT sensors monitoring manufacturing equipment.
Three core attributes define each event within these logs: case identifiers distinguishing different process instances, activity names indicating steps followed to achieve business objectives, and timestamps marking when events occurred. Additional optional attributes may include resource assignments for activity execution or departmental classifications for task performance.
Case notion definition requires careful consideration during the extraction process. Events sharing identical case identifiers constitute execution traces or business process instances. Traditional process mining extracts these logs from organizational databases and data warehouses. ERP systems generate timestamped event logs containing activity names, timestamps, case IDs, and optional attributes that reveal actual process flows, bottlenecks, and compliance deviations.
However, not all process data comes pre-packaged with business context. When organizations incorporate physical monitoring into process digital twins, they face a fundamentally different challenge.
Sensor data presents distinct extraction challenges because sensors lack explicit awareness of process concepts including activity names and execution instances. Organizations must correlate sensor data to specific process instances while maintaining alignment with business objectives.
Machine learning algorithms powering predictive capabilities require comprehensive historical data as their training foundation. Process digital twins accumulate extensive operational histories encompassing total operating hours, cumulative cycles, maintenance actions, failure events, and repair records. This historical foundation enables remaining useful life calculations and failure probability modeling.
Prediction precision correlates directly with clean, complete historical data spanning diverse operating conditions. Predictive models require adequate historical records covering various scenarios to establish reference points for accurate forecasting. Accumulated lifecycle data connects virtual model current states to established performance patterns.
Physical sensors provide real-time condition data streams that continuously update virtual model current states. Temperature readings, pressure measurements, vibration patterns, current draw, cycle times, and run-state data flow from IoT devices into digital representations. Individual wind turbines generate over 5 GB of data daily from hundreds of sensors. Facilities with 200 turbines produce terabytes requiring continuous collection, validation, and synchronization, often demanding sub-second latency performance.
IoT sensors capture diverse operational signals including GPS coordinates, energy consumption metrics, and equipment status indicators. Manufacturers embed sensing and connectivity capabilities directly into smart objects. Real-time data layers provide immediate answers about current process behavior and operational states.
The semantic layer that transforms raw measurements into meaningful information comes from comprehensive metadata. Contextual information creates connections between measurements and specific components, operational conditions, and historical performance trends. Asset identifiers, location tags, and documentation links become embedded within models to ensure proper interpretation of live data streams.
Data quality contracts establish specific requirements for freshness, completeness, and accuracy standards. Sampling rates, time synchronization methodologies, and missing-data policies maintain consistent measurement intervals across all assets and locations. Metadata schemas standardize twin definitions including lifecycle states, business ownership, temporal coverage, and control authority designations.
Digital twins achieve accuracy only through high-quality data inputs. Late-arriving data, poor quality measurements, or conflicting asset records across enterprise systems render virtual replicas unreliable, resulting in flawed insights and expensive operational decisions. Centralized, high-quality data spanning complete asset lifecycles establishes the single source of truth that creates valuable process digital twins.
Advanced modeling capabilities transform raw operational data into predictive simulations that enable organizations to evaluate potential changes before committing resources to implementation. Process mining establishes the analytical foundation for these capabilities, extracting actionable insights from business operations through systematic identification of bottlenecks, inefficiencies, and workflow constraints. These analytical tools have expanded their reach from traditional manufacturing environments into comprehensive business process management and workflow optimization applications.
Automated process mining software extracts operational parameters directly from uploaded event logs, capturing case arrival rates, activity durations, routing probabilities, and resource allocation requirements. This automation eliminates the weeks of manual configuration typically required for traditional business process simulation implementations.
Discovery algorithms uncover control-flow models represented through established frameworks including Petri nets, process trees, and BPMN models. Event data alignment with these control-flow models enables multi-perspective analysis incorporating temporal, financial, resource, and decision-making dimensions. Process discovery capabilities extend beyond analytical functions to simulate end-to-end process impacts across different operational scenarios.
Centralized knowledge repositories capture comprehensive system constraints, business rules, and operational logic within integrated simulation models. This unified reference approach ensures accurate replication of complex, mission-critical operations across individual facilities or distributed multi-site systems.
Intelligent data integration automatically collects, validates, and standardizes information streams from sensor networks, enterprise systems, and manual data inputs. This automated integration reduces implementation complexity while ensuring digital twin operations maintain access to complete, validated information. Process model design determines which operational functions and activities from physical processes receive inclusion in the virtual representation Design specifications encompass internal organizational divisions, system interfaces, material properties, spatial configurations, and overall system assembly requirements.
Performance benchmarking establishes baseline operational metrics for evaluating current system performance and generating accurate predictions for manufacturing facilities and supply chain networks. Organizations can monitor key performance indicators against established benchmarks to identify operational deviations and evaluate improvement opportunities immediately. Customizable metrics frameworks enable organizations to define and track performance measurements aligned with specific operational requirements and business objectives.
Discrete event simulation replicates system operations through sequential event modeling over defined time periods. Each simulated event occurs at specified timestamps, creating corresponding changes in system states. This mathematical modeling approach enables stakeholders to construct accurate representations of complex systems and evaluate potential modifications without disrupting actual operations.
Process simulation engines execute thousands of virtual process instances through established models, incorporating actual timing distributions, resource capacity constraints, and arrival pattern data. Teams receive quantitative predictions for cycle times, throughput rates, resource utilization levels, and operational costs before implementing real-world modifications. Simulation platforms support standard probability distributions including normal, exponential, uniform, triangular, and log-normal configurations.
AI-powered resource optimization models identify optimal allocation strategies, scheduling parameters, and throughput capabilities through continuous learning from historical and real-time operational patterns.
Unlimited scenario simulation enables testing of process modifications, equipment changes, and staffing adjustments before physical implementation[4]. Organizations can conduct comprehensive stress testing without disrupting production systems[4]. Scenario-based optimization develops and compares multiple operational approaches to identify optimal strategies for addressing changing demands and operational constraints[4].
Process digital twin applications enable risk analysis through virtual modeling of supply disruptions, inventory shortages, and market volatility scenarios[1]. Impact modeling helps organizations understand the specific steps required to increase operational resilience and maintain service delivery standards[1]. Process simulation applications extend across diverse business functions, including financial operations modeling cash flow scenarios, procurement departments predicting purchase price variations, and order management systems analyzing processing time requirements[3].
Enterprise data fragmentation presents the most significant barrier to process digital twin implementation. Enterprise data lives scattered across dozens of disconnected systems including CRM platforms, supply chain databases, payroll systems, and product usage logs. Traditional approaches to consolidating these information fragments required massive, bespoke engineering efforts accessible only to organizations with substantial technical resources.
Digital thread technology addresses connectivity challenges through bi-directional synchronization that maintains continuous alignment between physical and virtual environments. Enterprise system connectors serve as the foundational interface layer, collecting operational data from ERP, CRM, IoT, cloud systems, and operational sources. These connectors establish the primary communication pathway between existing business infrastructure and virtual process representations.
Data integration layers perform comprehensive collection, enrichment, and storage functions while ensuring process digital twins receive structured, historical, and real-time information streams. Knowledge graphs process both structured and unstructured data sources, extracting information from databases, IoT devices, APIs, and business systems to create semantic understanding. This intelligent processing layer transforms raw data inputs into meaningful, interconnected knowledge models that enable relationship mapping and reasoning capabilities across enterprise systems.
High-speed data pipelines support real-time ingestion requirements, providing the transport mechanisms necessary to move operational data from physical assets to virtual models with sub-100ms latency specifications. Gateway authentication protocols become essential under these performance requirements. Authentication gateways connect to IoT domains and manage data collection or routing for devices such as HVAC units, boilers, and dehumidifiers. Direct device connections transmit their own telemetry streams, while indirect connections route both telemetry and commands through gateway infrastructure.
API-based architectural approaches enable organizations to implement digital twins using existing system infrastructure while providing unified data access through single interface points. Process digital twins operate through modular software blocks that implement specific features and collaborate through defined interfaces. Each functional block provides services through APIs accessed via Data Link components, which function as message brokers between blocks and enable Internet-accessible service availability.
API gateways provide the practical implementation framework for these broker functions, forwarding HTTP messages to appropriate software blocks. This architectural pattern follows microservices principles, where individual software blocks implement discrete digital twin features. API gateways serve as centralized orchestration points for managing multiple APIs through authentication and routing protocols that ensure secure microservice communication.
Workflow management and API coordination facilitate data sharing across diverse sources for digital twin construction, maintaining synchronization between virtual representations and their physical counterparts. These modules handle operational workflows, processes, and event-driven flows that sustain continuous connectivity between virtual and physical operational domains.
Effective process digital twin implementations depend on centralized data fabric architectures that manage complete information lifecycles across heterogeneous environments. These components consolidate telemetry from IoT devices, historical records, and enterprise resource planning systems into unified information sources. Data integrity and consistency protocols enable organizations to maintain persistent digital threads that track assets from initial design through operational deployment and eventual decommissioning.
Bi-directional synchronization capabilities ensure connectivity extends beyond one-way data collection. Advanced digital twin architectures enable virtual models to transmit instructions back to physical assets through closed-loop control mechanisms. When digital twins identify operational inefficiencies, they trigger parameter adjustments to physical systems, creating direct pathways from digital insight to operational action. Continuous data integration protocols enable persistent, real-time communication between digital twins and their physical counterparts through automated exchange processes.
Effective visualization components translate complex process digital twin analytics into actionable operational intelligence. The most sophisticated virtual models deliver limited value when their insights remain inaccessible to decision-makers who need real-time process visibility.
Dashboard design requires careful selection of components that enable organizations to extract meaningful insights from digital twin outputs. Effective implementations incorporate six essential elements: process visualization displaying dynamic workflow states, performance indicators reflecting predictive insights, alert systems enabling proactive management, scenario comparison tools for evaluating alternatives, interactive controls for exploring what-if questions, and role-based customization tailored to stakeholder responsibilities.
Advanced dashboard configurations can display confusion matrices for each digital twin, integrated with asset management systems where KPIs receive continuous monitoring. Custom reports track confusion matrix evolution for continuous twin improvement and identification of process drifts as production environments change. These monitoring systems enable end users to assess performance accuracy and determine when virtual models require recalibration.
Dynamic visual representations communicate workflow states, resource utilization, and constraint formation through real-time displays. Color-coding, animation, and size variations provide immediate process health indicators. Manufacturing digital twin dashboards might display production lines using green for smooth operations, yellow for emerging bottlenecks, and red for critical constraints demanding immediate attention.
Comprehensive overview displays consolidate multiple control screens into unified monitors, delivering enhanced operational visibility. Operators gain the ability to monitor various processes simultaneously while identifying connections between different process stages. Research documented one case where operators identified front-end changes causing back-end contamination issues through comprehensive trend displays.
Enhanced metrics communicate process behavior and predictive insights rather than simple current-state KPIs. Modern dashboards present "throughput trend with 4-hour forecast" or "capacity utilization with bottleneck probability" instead of basic "current throughput" measurements. These advanced metrics enable users to understand performance trajectory and risk factors beyond present conditions.
Early warning mechanisms facilitate proactive management through identification of emerging issues before they impact operations. Alert hierarchies differentiate between routine variations, developing concerns, and critical situations requiring immediate intervention. Progressive alert escalation employs subtle visual cues for minor deviations, prominent warnings for developing issues, and urgent notifications for critical situations. Role-specific alerts ensure stakeholders receive appropriate detail levels, with field technicians accessing step-by-step repair instructions while senior managers view high-level summaries containing cost and scheduling details.
Sensitive operational data and intellectual property within process digital twins naturally attract cyberattacks, with up to 79% of organizations considering cybersecurity a key concern when implementing this technology . Process digital twin technology requires robust governance and security frameworks built directly into the architecture rather than added as afterthoughts.
Role-based access control (RBAC) restricts system access based on specific user roles, ensuring accountability across digital twin environments. According to Esri, organizations can classify data by sensitivity and control access through granular role-based approaches. Viewers receive read-only access for policy review, contributors gain limited editing rights for departmental tasks, field workers access task-specific mobile data, and creators build maps within assigned limits.
Blockchain technology offers decentralized access management through programmable smart contracts . These self-executing contracts automatically enforce access rules based on user roles, ensuring only authorized individuals can access or modify process digital twins . Smart contracts also automate access revocation when roles change or users leave projects, eliminating manual errors that leave unauthorized access points open .
Organizations must implement multiple security layers, specifically end-to-end encryption for all data transmissions and multifactor authentication (MFA) for accessing digital twin systems. Regular security audits coupled with penetration testing identify vulnerabilities before attackers exploit them.
Compliance with data protection regulations proves particularly challenging in healthcare applications processing personal data under GDPR or HIPAA requirements. Organizations need to balance comprehensive data collection against privacy mandates. Data encryption applies both in transit using TLS 1.2+ and at rest using AES 256. Adherence to standards like ISO 27000 and FedRAMP provides structured approaches to information security.
According to KPMG, governance frameworks must define clear roles, responsibilities, rules, and processes to ensure controlled and accountable operations across digital systems. Organizations should integrate regulatory requirements into architecture, data models, and interfaces early rather than retrofitting compliance later at considerable cost.
Every action taken on a process digital twin gets recorded, creating immutable audit trails that provide transparent histories of access, modifications, and system changes. This traceability maintains accountability and protects digital twin integrity. Organizations can generate detailed reports on user access, policy enforcement, and security incidents derived from the digital twin environment. These comprehensive records prove essential during regulatory audits and compliance verification processes.
The progression from architectural understanding to practical implementation demands systematic methodology that mitigates costly missteps and resource waste. Organizations must approach process digital twin development through structured phases that build upon foundational data components and integration capabilities previously established.
Process selection represents the critical first decision that determines implementation success. Organizations should target processes with clear objectives and measurable outcomes. The initial focus should establish specific achievement targets, whether minimizing downtime through predictive maintenance or optimizing patient flow in healthcare facilities. Beginning with monitoring simple components or single IoT devices provides essential hands-on learning before expanding to larger, more complex systems.
Successful process selection requires evaluation of data availability, stakeholder engagement, and business impact potential. Processes with well-defined boundaries, existing data capture mechanisms, and clear performance metrics offer the highest probability of successful digital twin implementation.
Data architecture forms the foundation upon which accurate virtual representations depend. Organizations must establish frameworks that collect, validate, and integrate information from sensors, operational systems, and environmental monitors. This phase requires careful planning for data preprocessing, cleaning, and synchronization protocols. Real-time data streams often contain noise, missing values, or timing inconsistencies that frameworks must address before feeding information to virtual models. Successful digital twins depend on thorough data collection from multiple sources to maintain accuracy.
The data requirements phase should establish quality contracts specifying freshness, completeness, and accuracy standards. Sampling rates, time synchronization methods, and missing-data handling policies ensure consistent measurement intervals across all monitored assets and locations.
Technology selection should prioritize interoperability, enabling different systems to exchange data without information loss. Component selection must support actual business goals while integrating with existing enterprise systems such as ERP or CRM platforms. Organizations should plan for ongoing training in machine learning, IoT integration, and advanced analytics capabilities.
The technology stack decision requires balancing current operational needs with future scalability requirements. Cloud-native architectures often provide the flexibility needed for iterative development while supporting the high-frequency data transmission necessary for real-time digital twin operations.
Model validation against historical data establishes accuracy baselines for virtual representations. Simulations should predict actual system behavior within acceptable margins of error. When discrepancies exist between virtual and physical performance, organizations must refine models until they reliably mirror real-world operations.
The validation phase requires systematic testing across various operational scenarios to ensure robust model performance. Continuous monitoring and adjustment mechanisms enable ongoing calibration as business processes evolve and new data patterns emerge.
Process digital twin architecture represents a sophisticated yet approachable technological framework when organizations understand its fundamental building blocks. The interconnected layers - from data foundations through modeling components to user interfaces - work in concert to create virtual representations that deliver quantifiable operational improvements.
Successful implementation begins with targeted process selection and clear performance objectives. Data quality serves as the critical foundation, as virtual models can only provide reliable insights when fed accurate, well-structured information streams. Technology selection should prioritize compatibility with existing enterprise systems, enabling organizations to build upon current infrastructure investments rather than requiring wholesale platform replacements.
The architectural components examined here - spanning data ingestion, processing layers, simulation engines, and security frameworks - provide the blueprint for organizations ready to capture the 15% efficiency improvements and 20-30% cost reductions that process digital twins enable. Understanding these building blocks positions decision-makers to make informed technology choices that align with operational requirements and business objectives.
Organizations that construct process digital twins following these architectural principles will unlock the predictive capabilities and operational insights that drive competitive advantage in today's data-driven business environment. As the technology matures and integration frameworks become more standardized, the barrier to entry continues to lower - making now the ideal time to begin your digital twin journey.
To help you get started, we've created a comprehensive guide that takes you beyond architectural theory into practical implementation. Download "Process Digital Twins Simplified with Simio" to access detailed implementation frameworks, real-world case studies, and actionable strategies for launching your first pilot project and scaling across your enterprise.