BLOG POST: This post highlights how current Machine Learning services enables integration to expand into the enterprise landscape whilst requiring developers to take reinvent themselves and the tools they use.


It is important to be aware of what Machine Learning is and what it isn’t. And why this term is preferable to the term Artificial Intelligence for the range of services that exist today. Current examples of AI, such as face or voice recognition, chatbots, predictive maintenance, self-driving cars and others, operate using complex algorithms with a theoretical basis in probability theory and statistics.

These algorithms are trained using massive amount of data working within a very narrow domain. Face recognition software does not recognize voices; in fact it does not even understand what a face is, because, unlike humans, it has no conceptual understanding of the data it is processing. Instead it uses a vast set of images to create a mathematical approximation of what a face might look like. From that it can then guess whether the next image contains a face or not.

Through evolution, a human being is born with the ability to generalize using minimum data. The DeepMind software, which won the game of GO, played GO an enormous amount of times before it mastered all the different strategies of the game. Artificial Intelligence today is concerned with the application of mathematical tools and engineering approaches to solve specific problems in specific domains.

Although a great deal of research is being conducted, we do not yet have a) software with the ability to generalize from a set of basic rules or b) software which can create abstract concepts and from those concepts create new ones. Enterprises will be greatly affected by Machine Learning services and will try to adapt them but will also need to understand the limitations, strengths and weaknesses of the services if they are to utilize them to their maximal potentials.


For Machine Learning and Integration to merge, the definition of “Integration” will have to be expanded so that we are talking about building services outside of what has normally been integration, i.e. linking systems. For example, we can move data from multiple databases to the cloud, find discrepancies in the data and create APIs to allow access. Or we can let services start other processes based on these deviations.

This is the future for us who are integrators. Certainly system-to-system integration will remain, but more of the new requirements, projects and technology will be about building new forms of services and there we need to broaden what we mean by integration


Even with the advent of machine learning, questions still need to be asked about why organizations are building connected systems. One must remember that machine learning such as predictive maintenance, which requires competence in IoT, is an investment that should result in value creation. It is about collecting data from environments, devices and users and turn it into information that is used to improve performance. The value is in greater reliability, faster time to market, reduced costs, better customer relationships and increased revenue. These outcomes are made possible when you know the state of your machines and the world around you at each moment in time. These are the outputs of the new Systems of Intelligence created by data-focused, flexible, secure, and scalable IoT solutions.

To be successful, business roles and integration developers working together should start with a specific problem in a specific domain. They should be crystal clear about the benefit the enterprise will derive from a cloud-based IT model. Applying these new skills and services will in the end require combining organizational, software delivery, and IT infrastructure transformation. Integration developers should be at the cutting edge of this transformation.

Back to community