Embrace machine learning to power your apps

 A blog post by Ram Ramdattan, DC Specialist Leader, Cloud Engineering US, Deloitte Consulting LLP

Designing, building, and deploying custom enterprise applications are key to tailoring and digitizing business processes. Distributed systems and the cloud era have ushered in widespread DevOps practices to enable better use of application build and deployments. Cloud-native technologies and applications accelerate deployment velocity significantly. With easier access to machine learning (ML), in all public clouds, we are now on the cusp of another transformation.

A new driving force for targeting user preferences

Machine learning (ML) has brought new possibilities and capabilities into the enterprise. Data engineers are being tasked with finding, cleaning, and transforming batch and stream data sets for ML use cases. Archived data is suddenly more useful to train machine learning models. Data scientists are discovering new algorithms that provide more accurate predictions for different facets of the business process and end users. The Deloitte 2021 holiday retail survey highlights how 62% of spend in stores is expected to be online while curbside pickup and buy online, pick up in-store (BOPIS) remain popular. Think of popular grocery apps that save customers time and elevate the user experience by recommending groceries based on past purchases or other users’ selections—ML is at work here. 

The average enterprise relies on hundreds of applications—internal and external facing—to support specific businesses and business processes. Enterprises have invested millions of dollars to build next-gen applications, and countless developer hours support these applications. While ML may open use cases for building new applications, it will be hard to ignore investments in existing applications. The effectiveness of ML for an enterprise could be determined by the adoption within existing applications. In other words, the convergence of ML and these existing custom applications is inevitable.

One can already see this trend in action within digital-native, born-in-the-cloud organizations. One can easily relate to global video and music platforms that provide content recommendations to users. Under the hood, a recommendation engine works through multiple ML models. It does this at scale, tailoring these recommendations across millions of users and petabytes of content. Without this ML capability, user engagement on these platforms would suffer, as the right content would be hard to find and consume, causing frustration and fatigue to content consumers. A point to note here is many of these platforms were started at a time when ML was either nonexistent or in its infancy.

Machine learning advancing the scope of current applications

Apart from recommendations, existing applications could benefit from several ML capabilities—accurate inventory forecasting, image recognition, language translations, and personal assistants are a few examples. The benefits range from higher efficiencies for organizations with smarter applications to better growth through a higher number of customers. We are now on the cusp of the new paradigm within the enterprise. This is the intersection of application development, security, operations, and ML. We could call this DevSecMLOps or MLOps. They all deal with the same theme: Consumer- and B2B-facing applications that the enterprise uses today will be more effective with embedded ML technologies delivered at scale.

It represents a change in the way that enterprises build and deploy cloud and ML native applications. There are five strategic areas that will need to be addressed:

1. Key ML use cases – Assessing and prioritizing which business processes/applications will start to benefit from embedded ML use cases.
   
2. Core architecture – Architecting, designing, and enhancing applications with ML built in and performant at the core.
3. New user experience – Evaluating how existing and new users will interact with not only apps but ML models within those apps.
4. Model and app development, testing, and deployment – Reengineering the process for at-scale continuous integration, testing, and deployment coupled with continuous learning.
5. Operational readiness – Conducting initiatives for teams supporting ML native apps and systems in production to generate awareness, impart relevant knowledge, and instill essential skills.

Changing the future of experiences for the better: For both consumers and business processes

Embedding ML within the application development life cycle is a big change moment. Years of systemic practices and organization structures may need to be revisited. The latest edition of the State of AI in the enterprise survey from the Deloitte AI institute provides a starting point for organizations.

Achieving this level of change, even within a single application, will be a huge milestone for an organization and provide tremendous benefits. This integration between ML and the app world is a question of when, not if. Businesses that can realize this transformation sooner will accelerate their growth curve faster than those who are still in the early stages of consideration.