How Machine Learning and Adaptive Methods Are Revolutionizing Integration
Let's take a cue from natural language processing (NLP). NLPs make an educated guess about what a person is trying to say and what the resulting action should be.
Join the DZone community and get the full member experience.Join For Free
For years, those of us in the technology sectors have been building integrations between disparate systems. In fact, enterprise organizations often have specific resources dedicated toward building and maintaining integrations between mission critical systems. According to dictionary.com, the word integrate is defined as a verb meaning "to bring together or incorporate (parts) into a whole." For decades, technologists have been manually creating integrations between systems whose interfaces have been continuously changing. Those of us who were lucky got the opportunity to create an integration between a mainframe and a front-line database, thus ensuring that fewer updates to the integration were necessary. However, for those of us who had to create integrations between the multitude of SaaS services on the market, we determined that a frequent maintenance plan and update cycle were an absolute requirement.
As years have gone by, integration has often been treated as a synonym for synchronization. While keeping data in sync between multiple databases and platforms is a key part of the integration process, it's not nearly the totality of the integration problems organizations often face. Focusing primarily on synchronization leaves out the most important part: automation.
Automating different components of business has always been approached with trepidation. While there are a number of things that could easily become automated (those rote tasks performed 1,000 times a day), more core components of the business could not benefit from automation due to the analysis required in decision making. It's time that we as technologists take a cue from another part of the industry: natural language processing (NLP).
An NLP uses the concept of intents to make an educated guess about what a person is trying to say and what the resulting action should be. For example, when I speak to an automated system at an airline, I may say "I need to know if my flight is on time." That phrase will then get passed through the NLP and compared against several different intents in their system. Other examples of intents for an airline may include “Speak to a representative,” “Purchase a ticket,” or “Get updated flight information.”
In the specific case of my first example, the intent would map to “get updated flight information.” Once the intent is known, we can map specific actions. In this example, the action would be a response with the current status of the flight.
The next major step forward in integration technology will take its cue from those that have come before. Utilizing some of the new techniques being developed by iPaaS providers in concert with machine learning techniques such as the example given above, we can come close to the holy grail of integration requirements. An integration developer should be able to explain the ultimate intent for the integration and have an intelligent system create a blueprint for review. Following a successful review, the same system should be able to deploy the integration and manage any changes to the underlying systems.
Imagine being able to work visually with an environment that allows me to explain my intent. For example, every time a lead comes into Marketo with a certain lead score, I need to automatically contact the head of sales and initiate marketing automation. To give general intent and then receive a blueprint on the execution by connecting different API endpoints (from Marketo, Salesforce, and Microsoft Outlook) would be just the start of a world of possibilities for automation that creates and maintains integrations.
We’re currently looking down the barrel of the Internet of Things, but we’re still approaching integration in traditional ways. Artificial intelligence combined with new tools for building integrations can bring about that ever-elusive goal: adaptive and intelligent integration. Building a system that maintains integrations based upon the ultimate intent as opposed to some rigid constraints will be a necessity when attempting to seamlessly link platforms that manage thousands (if not millions) of devices. The reason for pushing automated integration creation and management is to ensure we're ready for the incredible amount of data to be processed from the oncoming IoT storm. However, it will have tremendous value elsewhere as well.
Opinions expressed by DZone contributors are their own.