Build a Viable IT Architecture for AI and Analytics

I recently visited with the CIO of a Fortune 500 company. He was touting the advances they had made in IT and corporate culture regarding the use of artificial intelligence and analytics, but he had one major concern: How do you fuse AI and analytics into the rest of your transactional line of business IT infrastructure?

It hasn't been that way in his enterprise.

His IT organization had started its analytics initiative with an internal Hadoop group that was responsible for processing big data internally. Meanwhile other departments in IT supported transactional data processing on an assortment of mainframes and servers in the data center. Regular IT and the Hadoop groups were somewhat siloed from each other because the parallel processing and storage management needs for big data and AI were notably different than what they were for transactional data and processing management.

The original strategy worked for a while because the AI and big data capability was only being used by a company scientific research group. But now the potential of AI and big data drives more data and insights into everyday processes in marketing, finance, and customer service, extending AI and big data’s appeal. As new business processes were being defined, more of them depended upon the ability to exchange data between transactional and big data systems. That meant blending them into seamless business processes that not only pushed work out the door, but that also lent a hand in decision making.

The message was clear, and it is being heard in almost every organization today: It’s time to consider re-architecting IT into an overall framework that contains the processing and data of both big data and regular transactional data.

The challenge now on the plate is getting there.

First, Conceptualize an Architectural Framework

IT already has schematics that show the deployment of networks, systems, and data repositories for transactional data. On a system by system and application by application basis, there are data and process flows that show how the dispersal of data processing supports the execution of everyday business processes. That includes functions such as booking orders, paying bills, or inputting an entry into a customer service log.

If AI and big data processing has been functioning as an island, it’s time to bring that island into the overall IT infrastructure. This job begins by using the same type of process that’s utilized to document data repositories and process flows for traditional data. You look at where the AI and big data is being used, and you document the business processes that use it.

Regression testing should also be done. For example, if an AI system that is supporting edge-based manufacturing automation fails, what other IT systems and end business processes are affected?

Second, Figure Out How Cloud Fits In

It’s no secret that many companies that originally invested into on-premises big data processing are now opting to replace this internal AI and big data processing by moving it to the cloud. One reason is the ever-present lack of affordable AI and big data talent, which a cloud provider can step in to offer. A second reason is the sudden (and unforeseen) need to scale AI/big data storage and processing. It’s much easier to increase your cloud spend with the click of a button or two than to go back to management to request additional hardware, software, and bandwidth for the data center.

The takeaway is that your IT architecture, spanning both big data/AI and traditional data, is going to have to account for all these systems, processes and data that are cloud-based in addition to on-premises systems. It’s not enough to just document what you have on prem.

Third, Think in Terms of Sustainability

The Oxford Dictionary defines sustainability as “the ability to continue or be continued for a long time”. This is the definition that most CIOs think about when they contemplate orchestrating an overall IT architecture that encompasses every process and piece of data, whether data is big and unstructured or transactional and structured.

An IT architecture should also be malleable because technology and business continually change. Consequently, any all-inclusive IT architecture should be drawn up with expansion in mind (meaning it is easy to revise an IT architecture when new applications, technologies and business needs come along).

Fourth, Appoint an IT Futurist

Architectural changes and IT directions should be anticipated. This assures that new developments are entered onto an IT roadmap. It also gives the CIO an opportunity to brief the board and the CEO, at least annually, on which new technologies and trends are likely to impact the company over the next 3 to 5 years. This assures company technology and business readiness, and it paves an easier way for getting budget approvals for new tech.

One way to build a futuristic awareness is to appoint an individual in IT who is forward-thinking and who can look into future business and technology trends, understanding how they are likely it to affect IT and the company. This person can hold other responsibilities and doesn’t need to be full time in that role, but he or she can provide meaningful insights that will ultimately benefit both IT architecture and business direction.

Now is the time to think about how the IT organization of tomorrow can blend the needs and capabilities of both traditional transaction processing with AI and analytics.

What to Read Next:

Why the Mainframe Is a Mainstay of Hybrid IT Infrastructure

Sam's Club CTO Talks Infrastructure Modernization, AI

Turning Points in AI and ML