Digitalization for AI at Scale – Identification 2 of 8

Exploration on Digital Capabilities for AI at Scale continues with Identification part 2: Digitalization of products, services, customer solutions, business processes and business systems. For background, see article Introduction: Digital Capabilities for AI at Scale. In addition, getting familiar with Integration Requirements and Constraints is recommended.

Digitalization

The relationship between digital transformation and AI transformation is intimate. AI deployment wouldn’t even be possible without digitalized assets and capabilities. In effect, AI transformation is a subset of digital transformation – but a very substantial subset that calls for tailored analysis.

Digital characteristics of products, services, customer solutions, business processes and business systems create the foundation for AI use case integration.

AI deployment builds on five digitalization domains: products, services, customer solutions, business processes and business systems.

So, there are five digitalization domains to investigate. But why?

AI at Scale is about embedding AI in all aspects of value creation – for higher customer value, better customer experience and enhanced operational efficiency. Hence the five domains to consider for full coverage.

Missing digital elements – where ever they may be – emerge as bottlenecks or outright roadblocks to AI use case integration. In the Age of AI, such shortcomings constitute technical debt that is now to be paid.

In the AI-defined competitive landscape, digitalization related technical debt becomes payable. Time is running out.

Products

Products are either natively digital or digitalized. From AI use case integration perspective either variant serves the purpose but their digital characteristics do depend on the origin. In?general, natively digital products offer somewhat smoother path to full-blown AI utilization.

Natively digital products

Natively digital products are designed from ground up to function in a digital environment. Typically these are software-based solutions that users can purchase or subscribe to, such as applications, digital platforms or online tools. By definition, they need to have built-in computing, connectivity and UI capabilities.

Digitalized products

In terms of their origin, digitalized physical products differ fundamentally from natively digital products. In their case, digitalization is often an afterthought or added onto the physical core. Digitalization of physical products is a sum of many things, including:

• IoT sensors and actuators – Ability to capture real-world data for operations and AI and to be acted upon. For instance, sensors in manufacturing equipment to monitor wear and tear.
• Connectivity – Ability to connect to the Internet, and Operational Support Systems, and other devices.
• Edge computing – Ability to perform analytics or AI computing locally within the device or product itself.
• External computing – Option to perform computing tasks elsewhere, i.e. in cloud or on external server.
• Embedded software – Product enhancement and control through software. Small device might utilize firmware while larger products might have more substantial software capabilities, including update option.
• User Interface (UI) – Digital interface to use and control the product. For example, touchscreen or voice?activated controls.
• APIs for product integration – Capability to allow system integration based on Application Programming Interface with sophisticated functionality.

AI integration and constraints

AI builds on data. Subsequently, AI utilization depends products’ ability to source, process, exchange, store and maintain data. Specifically, processing of data builds on AI algorithms and architectures – realized with AI models. Depending on digital characteristics of products, AI utilization is more or less restricted. Understanding these restrictions is essential.

Integration contexts vary a lot depending on underlying technology, architecture, used methods and tools, and available resources. There’s no one-size-fits-all solution. Instead, every case is unique. Situational awareness becomes key success factor.

Every case is unique. Situational awareness becomes Key Success Factor.

In case of natively digital products, contraints to do AI integration are less common and less severe compared to digitalized physical products. Natively digital products are often designed to generate, access and utilize data to deliver value, and they tend to come with built-in connectivity, interfaces and API capabilities. Turning natively digital products into AI-enhanced products is about data and AI model integration within the pre-existing application. Effort required depends largely on software architecture and design: The more monolithic design the harder AI integration. Conversely, microservices architecture provides a good starting point for smooth AI integration.

Natively digital products offer easier path to AI-enhancements compared to digitalized products

Assessing digitalized physical products for AI use case integration involves careful analysis of their digital capabilities, including things like:

• Computing power – AI processing done on-device, locally or cloud, including trade-offs like latency and battery life.
• Sensors and data collection – Retrofitted IoT sensors and the amount and quality of data created.
• Connectivity – Bluetooth, WLAN, 4G/5G or wireline connectivity solution, including bandwidth, intermittent connectivity and latency related limitations.
• User interface – Access to product thru mobile app, on-device interface or similar.
• System integration – Ability to integrate with digital systems, platforms and solution components, including compatibility and standard interfaces with protocols, data formats and APIs.
• Integration framework – Software engineering related factors like technology platform, software architecture and SDKs needed to do data and AI model integration.

After constraints have been identified, it is time to assess their significance in terms of limitations on AI use case integration, and magnitude in terms of effort, cost and time required to eliminate each constraint.

Services

Like products, services too are either natively digital or digitalized. Natively digital services are designed for the digital realm, making them typically more scalable and versatile compared to digitalized services. Natively digital services tend to integrate better with other services. Due to missing physical legacy, constraints to do AI integration tend to be less severe.

Natively digital services

In the digital realm, the line between natively digital products and services becomes blurred. This is because fully digital environment caters to seamless blend of what constitutes a product versus service:

• Value Delivery – Both deliver value through on-demand accessibility with as-a-service mechanisms like SaaS, continuous updates and improvements, and cloud-based data storage.
• User Experience – Both are accessed through interfaces like web browsers or apps, making the user experience similar.
• Pricing Models – Pricing is typically based on subscription models, usage-based pricing, or freemium, which can apply to both products and services.

In the context of AI utilization, differenting between natively digital products and services is less relevant than understanding functionalities, user interactions and technical infrastructure. The key is to identify where AI can add value and what are the constraints for AI integration – regardless of whether the offering is labeled as product or service.

Line between natively digital products and services becomes blurred

Digitalized services

In contrast to natively digital services, digitalized services are typically hybrid in nature, containing both digital and physical elements of service delivery. Service digitalization involves things like:

• Digital service platforms – Platforms or applications that allow users to access and use a service. For instance, digital banking app or web store.
• User Interface (UI) – Web or application based interface for users to interact with the service.
• Service automation – Some parts of service delivered automatically with e.g. a chatbot without human involvement.
• APIs for service extension – Application Programming Interfaces that enable integration with other services, platforms or tools. For example, weather data retrieval from another service.

AI integration and constraints

In terms of AI utilization and integration, factors discussed in connection of products are largely applicable to services as well. For example, AI’s dependence on sourcing, processing, exchanging, storing and maintaining data.

As discussed, natively digital services are not that different from their product counterparts. Like products, natively digital products are designed to generate, access and utilize data to deliver value, and they tend to come with built-in connectivity, interfaces and API capabilities.

However, the picture changes with digitalized services. Due to their hybrid nature, AI use case integration calls for partial service redesign. That is, assessing AI integration constraints in the digitalized services context involves understanding how these services have been digitally enhanced and how AI can further improve service experience. Factors to consider include:

• Baseline digitalization – Level of service delivery digitalization considering the entire service delivery process; Digital interfaces and platforms like mobile apps and web platforms deployed to provide and manage the service.
• Access to data – Data generation, collection and usage in the service delivery process; Quality, volume and variety of data available; Limitations of data silos with data in legacy systems or formats that are not easily accessible.
• Connectivity – Connectivity with other digital systems, IoT devices, and external data sources; Reliability and bandwidth of connectivity solutions.
• System integration – Ability to integrate with digital systems, platforms and solution components, including compatibility and standard interfaces with protocols, data formats and APIs; limitations caused by legacy systems.
• Scalability and flexibility – Scalability of the service infrastructure to handle AI integration; Flexibility of the service model to cater to AI-enhancements.
• Impact to service design – AI integration impact to customer experience related to e.g. personalization and automation.

Like with products, identified constraints’ significance and magnitude are to be assessed next.

Hybrid nature of digitalized services calls for partial service redesign in order to assess AI deployment gains and needs

Customer solutions

Customer solutions are characterized by their integrated, problem-solving nature, offering cohesive package of products and services tailored to specific customer needs. They provide more than just the sum of their parts, delivering added value through deep understanding of the customer experience.

Examples across industries

Identifying customer solutions in various industries involves discerning whether the company's offering is integrated and systemic, solving specific customer problems and adding distinct value, as opposed to being a collection of isolated products and services. Some examples from different industries:

• Manufacturing – A manufacturing company offering a complete automated assembly line solution, including machinery, software for process optimization, ongoing maintenance, and training services.
• Utilities – Smart grid solutions that include hardware (sensors, smart meters), software for grid management, data analytics services for energy usage optimization, and customer engagement tools for energy conservation.
• Telecom infrastructure – A comprehensive telecom infrastructure solution might include network hardware, installation services, software for network management, cybersecurity services, and ongoing technical support.
• Telecom services – Bundled offerings that combine voice, data, television, and cloud storage services, often with added features like family plans, flexible data management, and integrated customer service platforms.
• Healthcare services – Integrated healthcare solutions providing telemedicine platforms, wearable health monitoring devices, personalized patient care plans, and data-driven treatment recommendations.
• Retail – An e-commerce solution offering a seamless online shopping platform, integrated supply chain management, personalized marketing and sales services, and customer analytics.
• Construction – Complete building solutions that include design and planning services, construction management software, building materials and equipment, and post-construction maintenance services.
• Financial services – Financial service provider with tailored investment solutions, combining personal advisory services, digital investment platforms, and portfolio management tools.

Customer solutions defined like that form a core element of company’s value proposition in the Business Model context. Something to keep in mind when assessing AI use cases as ways to boost value creation.

Customer solutions are at the heart of Business Model value proposition

AI integration and constraints

Systemic nature of customer solutions makes assessment of AI’s potential more challenging compared to individual products or services. Assessment is no longer only about digital readiness of products and services by themselves but also about how digitalization and AI enhancements impact the solution as a whole.

Here’s an outline of how to assess AI’s potential and constraints:

• Solution characteristics – What are the products and services that make up the overall solution and how they relate to each other? How does the solution connect to customer journey and how does it enhance customer experience thru multiple touchpoints?
• Digital characteristics – What is the digitalization maturity level of products and services that make up the overall customer solution? How does digitalization enhance the solution on system level, beyond individual products and services, e.g. thru enhanced customer experience across multiple touchpoints?
• Access to data – How seamless is data exchange between solution components? Is the available data sufficient for AI deployment in terms of access, diversity, volume and quality?
• System integration and interoperability – How do solution components (products and services) integrate and interact with each other? What is the level of compatibility in terms of standard interfaces, protocols, data formats and APIs? What would be the external resources like cloud to do AI computing?
• Scalability – Are there any apparent bottlenecks in terms of available computing, storage and data resources considering increased needs due to AI deployment?
• Regulatory compliance – What would be the regulatory compliance implications related to data privacy and AI ethics? To what extent would these restrict AI use cases e.g. related to sensitive data?
• System cohesion – Overall, how cohesive the solution is from architecture, functionality, flexibility, scalability, operations and regulation perspective? How does it appear as an environment to do AI integration?

Systemic nature of customer solutions makes assessment of AI’s potential more challenging

Business processes

In terms of overall productivity, products, services and customer solutions contribute mainly to customer value and customer experience. That is, to external factors. With business processes and business systems, the attention shifts now to internal factors contributing to operational efficiency. In reality, the division to externals and internals is not clear-cut: For example, the way business processes are run will have an impact on customer experience too.

A business process is a series of structured and related activities or tasks that produce a specific outcome. It's essentially the "how" of doing something in business. For example, the order-to-cash process in sales or strategy process in strategic management.

ERP dominance

Digitalization of business processes is predominantly about deployment of an ERP module of some sort. In case of fairly standard processes and corresponding ERP modules, evolution of AI support boils down to vendor roadmap. Conversely, in case of business-critical highly customized processes and modules, AI integration gets more complicated and depends on ERP module flexibility, scalability and customization capabilities.

Business process digitalization evolves around ERP modules where each module serves the needs of specific business processes. For example, CRM module having close linkage with sales and marketing related processes.

At its inception, ERP was synonym for massively complex piece of software that took care of every operational aspect within a business. Selected ERP vendor was happy to upsell multiple modules within the single monolithic architecture.

Introduction of Postmodern ERP broke down the monolith with the emergence of loosely coupled ERP modules, often delivered as Software-as-a-Service applications. This marked a decisive move towards distributed architecture and from single-vendor control to ecosystem play. The latest ERP framework is Composable ERP introduced by Gartner. Composability refers to highly modular approach where ERP components can be easily added, replaced or removed based on evolving business needs.

Composable ERP is about flexibility, adaptability and scalability. Because of these characteristics, it is useful benchmark in assessing potential and constraints to do AI at scale. That is, both functionality and scalability are of interest.

Postmodern / Composable ERP align well with AI at Scale. Avoid monoliths. Leverage ecosystem play enabled by distributed architecture and open interfaces.

ERP categories

ERP modules fall into two primary categories: Operational ERP and Administrative ERP.

Operational ERP – These are mission-critical modules that directly support the core operations of the business. These modules affect directly company's ability to produce products or services and deliver them to the market. A failure in operational ERP modules can significantly disrupt company's ability to function. Corresponding business processes are related to:

• Production and manufacturing
• Inventory and order management
• Supply chain management
• Distribution and logistics
• Purchasing and procurement

For example, CRM module would typically be part of operational ERP category. This is because CRM supports directly core business operations, especially in industries like retail, services, or any other sector where customer interaction is primary activity. CRM tools help manage sales processes, customer service, and marketing campaigns – all of which directly influence revenue generation and customer retention.

Another example of operational ERP module would be PLM (Product Lifecycle Management), especially for manufacturing and product-centric companies. PLM systems manage the entire lifecycle of a product, from its initial conception and design through to its manufacturing, service and disposal. Given that this module is so closely tied to the core operations of product development and production, it is typically considered operational.

Administrative ERP – These modules support the internal administrative functions of the organization. While they might not be directly connected to production and delivery of goods or services, they are nevertheless vital for the overall functioning and governance of the organization. Corresponding business processes are related to:

• Human resources (recruitment, payroll, benefits)
• Financials (accounting, financial reporting, treasury)
• Business Intelligence (analytics, insights, reporting)
• Projects (project management, resource allocation)

AI integration and constraints

In the context of business processes, AI integration and related constraints connect to three main factors:

1. The degree of business process digitalization overall,
2. Characteristics of deployed and available ERP modules, and
3. System integration related possibilities and limitations.

By assessing prevailing constraints, it is possible to gain holistic understanding on AI deployment limitations. This is essential to guide future investments in ERP upgrades and customizations needed to facilitate AI integration.

While Postmodern or Composable ERP provide ideal starting point, the reality can be quite different, e.g. partially digitalized processes running on a monolithic legacy ERP. In such a scenario, it is vital to build a realistic view of not only the constraints themselves but also of viable road ahead.

While Composable ERP provides ideal starting point, in practise AI integration assessment and planning may need to start from monolithic legacy ERP.

Here’s an outline of questions to assess AI’s potential and constraints in the context of business processes:

• Degree of process digitalization – To what extent all key processes have been digitalized? What are the key technical or non-technical factors slowing down further digitalization effort? Are there significant dependencies on legacy IT systems hindering transition? Are there any mission-critical applications that rely on legacy systems?
• ERP suite current state – What is the prevailing ERP architecture and technology maturity? What is the current state of the ERP modules in use? To what extent they support AI integration and deployment?
• Data integration – Is the available data sufficient for AI deployment in terms of access, diversity, volume and quality? How seamless is data exchange between ERP modules and other system components? Are there mechanisms to ensure data integrity across ERP modules e.g. related to standardized data formats? To what extent real-time data processing is supported to enable AI solutions that require real-time inference?
• System integration and interoperability – How robust and flexible are the integration enablers like APIs and related interfaces, protocols and data formats? What would be the key AI integration points in the current ERP modules and architecture?
• Modularity and scalability – Are there any apparent bottlenecks in terms of computing, storage and data resources considering increased needs due to AI deployment? How modular is the current ERP solution to facilitate scaling up with AI?
• Adaptability and roadmap – What is the ERP module suite roadmap in terms of planned AI capabilities? Are there any apparent vendor-specific limitations? Is there a roadmap to Composable / modular ERP with an investment and deployment plan? To what extent standard off-the-shelf ERP suite will be enough? What would be the ERP modules where customization seems necessary? How the necessary customization will be facilitated thru outsourcing and/or in-house development? How customization would serve AI integration needs?

Business systems

A business system is a coordinated combination of processes, methods and tools. For example, Customer Relationship Management can be considered a business system, encompassing business processes from lead generation to customer support, including the tools needed. CRM is not just a random collection of processes and tools but a carefully designed system – both as value-creating entity within a company as well as a product or solution by CRM vendor.

Business processes are subset of a business system

In practise, ERP modules like CRM bind business processes and business systems together. So much so that one might ask: Why do we need business system as a concept at all? Why to mirror externally relevant customer solution with internally relevant business system? Could we just deploy what ERP module vendor sells us?

It seems that the answer is included in the last question. It is a matter of perspective. More importantly, it is a matter of ownership and control: It is for us to define business systems we need, not ERP vendors.

It is for the company to define business systems – not ERP vendors

In normal circumstances the two perspectives align and result in high-performance systems – delivered as solutions by those ERP module vendors. But it is always worth remembering who is supposed to be in the driver’s seat.

Examples across industries

More often that not, business systems and their characteristics are industry-specific:

• Manufacturing – Supply Chain Management System used to manage raw material procurement, inventory and logistics. Production Planning and Control System used for scheduling, production tracking and quality management. Maintenance Management System used for predictive maintenance and equipment management.
• Construction – Project Management System used to manage construction projects, timelines and resources. Resource Allocation and Logistics System for managing equipment, materials and workforce allocation. Design and Architectural Planning System used to integrate CAD and Building Information Modeling tools.
• Retail – Inventory Management System for stock management, demand forecasting and replenishment. Customer Relationship Management for managing customer interactions, loyalty programs and personalization. Point of Sale System to integrate sales, payments and customer data. Omnichannel System for seamless integration of online and offline sales channels.
• Financial Services – Risk Management and Compliance System for risk assessment, regulatory compliance and fraud detection. Customer Wealth Management System for managing client portfolios, investment strategies and financial planning services. Transaction Processing System to handle day-to-day transactions, reconciliations and reporting. Credit Analysis and Loan Management System for credit scoring, loan processing and monitoring.
• Utilities – Asset Management System for managing and optimizing physical assets like grids, pipelines and plants. Customer Information System to handle billing, usage tracking and customer service. Energy Trading and Risk Management System for managing energy trading, demand forecasting and market analysis. Grid Operations and Network Management System to focus on grid integrity, demand-response management and distribution logistics.

By understanding the unique operational nuances of each industry, it is possible to identify where AI can deliver the most value and where constraints are likely to occur.

AI integration and constraints

Business systems’ AI potential and integration constraints assessment is highly synergistic with business processes. Synegies emerge from the common nominator: The ERP suite.

In addition to constraints assessment of business processes as discussed above, business systems call for additional assessment as follows:

• Business system identification/definition – Identification of business systems in use and defining their role and scope. Pointing out ERP modules in play and cross-checking their status and outlook.
• Involved business processes – Verifying business processes as system components, including their digitalization and AI integration status and outlook. Structural overview of the system components: processes, interfaces, data sources and IT systems (based on Enterprise Architecture, when possible)
• System integration and interoperability – Business system specific integration needs and limitations beyond business process assessment findings.

Conclusions

Digitalization of products, services, customer solutions, business processes and business systems is essential but not sufficient enabler for AI at Scale. Beyond Digitalization, there are seven other Digital Capabilities. All of them are necessary enablers. All of them are potential sources of constraints that slow down or prevent achieving AI at Scale. Because of that, Constraints Assessment is the very first no-regret logical step.

Constraints Assessment is the first no-regret step

To eliminate constraints is to enable embedding AI in all aspects of value creation. To integrate AI into business operations across the whole company – the way it creates and delivers value.

Achieving AI at Scale is a journey with three key characteristics: holistic, systematic and disciplined. Because of that, a Critical Success Factor emerges: Business executives’ explicit presence with hands-on ownership of AI Transformation. Attempt to delegate AI Transformation to, say, CIO organization, is doomed to fail as the necessary capabilities go far beyond any CIO’s sphere of influence.

Attempt to delegate AI Transformation to a technology department is doomed to fail

In effect, AI Transformation needs to become part of Management Team and Board of Directors standard agenda. That is the team to provide strategic clarity and push for change in order to achieve AI at Scale.

All that calls for significant amount of senior management bandwidth but the alternative is much worse: With the Age of AI, competitive landscape has become AI-defined with industry leaders at the Productivity Frontier. Falling too far behind creates existential risks that call for proactive mitigation.

Constraints Assessment as a Service

Constraints Assessment as a Service covers all digital capability areas from Strategic Management to Data Culture. See detailed Service Description.

AI at Scale Workshop

AI at Scale workshop is a compact one-day event for business and technology executives and managers. Workshop seeks answers to the question: What should we as a business and as an organization do to secure our success in the Age of AI?

Next article in the series

Exploration of digital capabilities continues with Computing for AI at Scale.