Innovation is evolution
Innovations like the Smartphone and Generative AI seem revolutionary, but each is a natural evolution and a result of ongoing human ingenuity and cumulative knowledge. At Origin, we constantly remind our clients that true innovation emerges from a combination of existing ideas that are refined, reimagined, or applied in novel contexts. As organizations step into the AI age with both feet, separating themselves from the competition is dependent on a pragmatic and persistent approach to innovation.
Yes chef
“Mise en place” is a French phrase used in kitchens around the world which means “setting in place, positioning.” Before making the best Italian Beef sandwich or 5 course meal, the kitchen must be organized so that cooks can focus on flawless execution of the menu. Treat your AI solutions and the data that fuels them like a Michelin-star menu with a composable product approach.
Composability is not new
Composable architecture is a software design principle that enables developers to break down complex systems into manageable and reusable parts, fostering a modular ecosystem where each component performs a specific function. Composable architecture has been used to create modular and dynamic applications that can adapt to changing business requirements and scale effortlessly. By leveraging composable architecture, developers can accelerate development, reuse pre-built components, and facilitate seamless integration with other systems and services.
Within e-commerce, composable architecture has been used to create personalized experiences efficiently for customers. Having the ability to quickly respond to trends and introduce new innovative features quickly and seamlessly has led to tremendous results.
- Satisfy customers: According to a study, transitioning to MACH [composable] architecture improves customer experiences by 60%
- Better Business Outcomes: According to a survey conducted by Gartner, Inc., 63% of CIOs at organizations with high composability reported superior business performance compared with peers or competitors in the past year.
- Larger Budgets: In 2022, CIOs and technology executives at high-composability enterprises expect their revenue and IT budgets to grow, on average, by 7.7% and 4.2%, respectively, while low-composability enterprises only expect both to increase by 3.4% and 3.1%, respectively.
Composing AI solutions
Composability hinges on three fundamental principles: modularity, autonomy, and discoverability.
- Modularity involves breaking down a system into distinct, self-contained components to achieve a specific purpose.
- Autonomy means that those self-contained components are not dependent on other components.
- Discoverability means that components can be found and understood so they can be reused.
The path to composable AI begins by aligning AI capabilities with business outcomes, defining an AI product roadmap, and creating an AI factory.
Align AI capabilities with business outcomes
Organizations tend to jump into AI product development or get analysis paralysis determining the first step. Moving too fast leads to applications that don’t scale. Taking too long to begin delays any progress and let’s your competitors get ahead.
At Origin, we believe that answering the questions below provides enough analysis to create a meaningful AI product roadmap.
- What are the results that are important for your organization and your customers?
- What are the KPIs that show how well you are doing?
- What use-cases will drive results?
- What use-cases will have the greatest impact?
- What use-cases are most complex?
- What AI services will enable the use-cases?
By doing this, organizations have the information needed to define products and plot their development and iteration on a roadmap.
Define an AI roadmap
Once you have gathered and linked use-cases to AI services, you can discover patterns and design products that can address multiple use-cases. By having a comprehensive view of what is required, organizations can find an effective way of creating reusable components for multiple products. To finalize an AI roadmap, it is important to ensure components and products are launched along a maturity curve to provide consistent incremental value. Once your initial roadmap is complete (it will and should change), you need to set up a factory approach to building AI products & components.
Create an AI factory
MLOps, short for Machine Learning Operations, is a strategic and composable approach to building scalable and efficient AI models and applications. MLOps combines the principles of DevOps with machine learning, enabling teams to automate the end-to-end lifecycle of machine learning applications. This includes integration, testing, deployment, and monitoring, ensuring that AI applications are modular, autonomous, and discoverable. Of course, the work is not over because AI factories depend on composable and trusted data products to deliver value.
Decompose your data to enable AI composability
Data is the fuel of AI, but does not come out of systems trustable, consistent, or available to those who need it. To create an AI factory that can produce high-quality and reliable applications, organizations need to decompose their data into reusable and trustworthy components. This means decoupling your data from source systems, identifying and mastering your critical data elements, harmonizing transactional data with master data and building data products for maximum reusability.
Decouple your data from source systems
Every data consumer asks for “one source of truth” to be the building block of all Data & AI initiatives. To accomplish this, organizations have invested heavily in ERP systems only to be disappointed that it still does not serve all their Data & AI needs. ERPs handle many mission-critical tasks but not all tasks, which leads to multiple sources of truth.
Master your critical data elements
Once data is decoupled, organizations then turn to managing their most important data assets, master data. Master data includes customer, product, material, vendor, employee, and geography data. There are Master Data Management (MDM) solutions that can handle all master data and specialized ones like Customer Data Platforms (CDP) and Product Information Management (PIM) systems.
Harmonize your transactional data
When you write sentences to tell a story, you have nouns and verbs. Master data provides the nouns to tell a data story, but transactional data are the verbs. Transactional data includes sales, demand, costs, inventory, shipping & logistics, research & development results, and manufacturing events. Transactional data is very fragmented, so the work here is to not necessarily make it uniform, but to make it interoperable with your master data.
Build data products for maximum reusability
While harmonizing transactional data with master data is essential, it is merely the starting point. The true value lies in business-driven transformations—aggregations, integrations, calculations, and meaningful experiences. But here is the key: do not confine these transformations to dashboards or isolated data products. Instead, perform them at a broader level. Understand your organization’s capabilities and value chains deeply. By doing so, you unlock the full potential of reusable, impactful data assets.
Conclusion
Take a lesson from Michelin-star restaurants and e-commerce by using a composable approach to AI solutions. Create an overarching strategy to drive business outcomes with AI. Ensure your data estate is trustable and reusable to fuel AI solutions. And if you need any assistance throughout your journey, we are here to help at hello@origindigital.com.
.svg)