TonyWi3

Artificial Intelligence’s (AI) new use case: Integration?

Blog Post created by TonyWi3 on Aug 9, 2018

In the last year or so we have seen a great deal of increased publicity about AI. AI is an umbrella term that includes multiples technologies such as computer vision, machine learning, natural language processing, and others. It’s useful to think of AI as software that has the ability to see, hear, reason, and learn.

Recent publicity has typically been focused on use cases such as fraud detection, voice-enabled assistants, data security, image recognition, and driverless cars; but what about systems integration?

Integration is becoming increasingly important for organisations wanting to stay competitive by leveraging data and capabilities provided by third parties. Currently, integration is mostly the domain of technical specialists, undertaking their work cautiously and slowly. But therein lies a problem; businesses now want integrated capabilities on demand. AI has the potential to turn these wants into reality.  

At the device level, automated integration is already here; the challenge is to provide complete systems that will automatically connect together. So how can you design for integration between ecosystems, platforms and things you cannot predict? The answer is using AI. 

Let’ consider a practical example: A bank needs to transfer funds to a new trading partner. The banks AI will know which API's to connect to, and just the right information to exchange and complete the transaction, in a secure manner. No integration project, no technical team, and no big cost.

Evolution of Integration

Up until very recently, integration has mostly been something businesses have either outsourced or used in-house technical teams for. The problem, however, is that the demand to connect to things digitally is outstripping the ability for human integrators to deliver; whether they are specialist in-house teams or outsourcers.

To meet this demand, integration delivery needs to be re-balanced in other ways. Enter the digital integrator. The digital integrator uses metadata analysis for improved data mapping and provides process analysis to suggest better integration, all packaged as part of a product. This approach reduces the need for integration specialists and opens up integration to a process of human-initiated needs fulfilled by AI enabled systems.

Digital integration using AI can, therefore, enable everyone to perform the integration. Humans can drive integration by using AI services to request digital outcomes. The AI engine listens to requests, then connects them together with the data and systems needed to support the requirement. AI-based integration potentially shortens the learning curve of both specialist and less / non-technical integrators to manage data and business flows, enabling almost anyone to perform integration tasks.

 In many business ecosystems we now have the following actors:

  • Integration specialists - The traditional integration of people who generally focus on enterprise projects, bulk data processing and governance.
  • Ad-hoc Integrator's - Tend to be in the line of business roles and use fit for purpose tools for Mobile and API development
  • Citizen Integrator's - Are focused on personal projects, want instant gratification and integrate using data sync SaaS apps often on mobile devices
  • Digital Integrator's - Are AI and ML (Machine Learning) driven, support digital business projects and deliver automation using deep learning

Challenges

This new and rapidly evolving environment is not without its challenges. The first challenge, obvious to the technically minded, is the need for common standards for AI/ML integration, as without a common standard for interconnection AI/ML will struggle. The challenge of standards, however, is a challenge in decline as increasingly the world is creating API's which in turn fuel the development of standards. 

The next challenges relate to change management, security and data privacy. With the rapidly increasing use of productised API management systems, these areas are also becoming less of a challenge. Change management is enforced by API management platforms ensuring the integration actors work to guidelines and processes enforced by the API owner. Security and data privacy are also managed by policy ensuring that API consumers are not only authorised but also mandated to control access to data at a very granular level. So the technology provides the policymakers with the tools they need to enforce a policy.

How does a digital integrator know which API to connect to, to deliver on the task the business owner has directed them to fulfil? For a long time, API's have operated with levels of discover-able metadata. Meta data is like digital sign posts that direct data consumers to APIs. With the right systems and processes in place, the API owner retains full control over what data, and services, are made available.

The most controversial challenge is; how much do we trust an AI? Most of what has been described so far are current technology, and some product vendors are actually starting to incorporate AI into their API management and integration products. So why is this whole area a candidate for wallowing in a trough of disillusionment? Basically, it’s us humans and our inherent distrust of AI. However, AI and ML, driven by citizen integrators, is now gaining momentum in terms of social acceptability.  

Enablers

  • AI and machine learning techniques applied to integration are emerging as digital integrator technologies.
  • Growing criticality for a business to leverage AI to support automation, insight and engagement.
  • Emerging technology innovations in AI for reduced time to integration and enabling citizen and ad-hoc integrators.

Inhibitors

  • Organisations lack awareness for these emerging offerings of AI in integration platforms and the benefits of applying such new technologies.
  • Business is cautious about exposure caused by high-value application and data flows used by AI.
  • Concerns about how to bring AI systems into the realm of reliable mission-critical systems.

I suggest it is now time to investigate the benefits of AI by experimenting with integration platforms. However, temper that with a good dose of scepticism and be directed by integration specialists. Even with new AI technology integration specialists still, have a key part to play. 

Outcomes