Deploying AI-Driven Systems

By on

Artificial Intelligence (AI) will continue to transform and disrupt most industries but, to really stay competitive, organizations need to deploy AI-Driven systems quickly. However, as Ramesh Mahalingam, CEO of Vizru, an AI-based autonomous applications company, said this is not an easy feat:

“AI adoption has proven quite a challenge. It takes enormous time, cost and resources to build and deploy scalable AI systems. Incorporating AI processes within an existing legacy infrastructure seems insurmountable. Add to this a chaotic environment filled with a recent surge in ill-conceived AI products/solutions, that raises concerns around data compliance and security, and you have a critical need for a reliable cohesive AI framework.”

Mahalingam has spent his career digging into the challenge on developing a reliable, cohesive AI framework. He enjoys teasing out turbulent technologies and sees a lot of disorder with AI capabilities, so he began figuring it out from a macro-level view.

Two Issues with AI Driven Systems

Mahalingam breaks down the confusion about AI Systems into two major issues: privacy and a lack of understanding about the AI systems trade. About security, he said:

“Companies get to the first stage of just trying and piloting an AI proof of concept. However,when the idea becomes serious and needs due diligence, most AI projects fail because security officers, for instance, are worried about what happens if these systems go rogue. AI systems are hungry and are fed large chunks of competitive information shared among a variety of people and machines, leaving costly compliance and security breach vulnerabilities.”

Because of this concern, Mahalingam finds that companies seeking vendors’ assistance for AI implementation want complete control and autonomy over their data, environment, and infrastructure. Organizations want to “partner with providers who can offer creative autonomy, IP, and flexibility in owning AI assets that they create.”    

Additionally, AI systems have a fuzzy environment that suppliers are misreading, noted Mahalingam.

“These vendors want twenty- or thirty years-worth of a customer’s data to come up with Data Modeling that detail how to build better products with AI systems. Instead enterprises look for applied, consumable, manageable, and quickly deployable AI capabilities, without having to compromise on large amounts of data. They want a specific artisanal AI model for a use case that caters to their DNA. This is a nuanced difference from yesterday’s AI rules-driven system that is the same for everyone.”

Mahalingam stresses that artisanal AI models are crucial to providing customers hyper-specific services that scale. Furthermore, he said users want to own the artisanal AI model.

To achieve this, Mahalingam and Vizru are continuing to envision a “snap framework”— technology that supports intelligent communication and work between AI systems or bots. This scaffolding provides specific models and pieces tied to desired functionalities, are plugged into the customer’s system, and allow the buyer exclusive ownership. As a first step, Mahalingam believes in creating a cohesive outer shell to “deploy and deliver AI systems.”

Starting a Framework to Launch AI Systems

In designing a framework to deploy AI Systems, Mahalingam conceives the long-term business environment.

“The way I look at it, you either focus on the weather or the climate. We wanted to target the strategic shift, analogous to the climate. This means building the framework in the right way to adapt to changing weather patterns. For example, people build chatbots very quickly, but that technology hype cycle will go away and become outdated every couple of years. A proper framework makes it much easier to deploy newer capabilities, and newer AI systems.”

What does the “right” mechanism for AI bots look like? It has a guardrail that determines where bots could be deployed and scaled and provides governance, so enterprises can oversee them. This ensures adherence to data privacy concerns and compliance with regulations. Particularly, Mahalingam identifies three main framework foundations.

  • A No-Code Development Platform
  • Stateful Network
  • Means of Evolving Applied AI Systems

A No-Code Development Platform

Mahalingam believes fundamentally in a no-code environment. He said:

“The next generation of applications will not be determined by simple ‘if statements’ and for statements that will be written in the future, but will be defined by complex and advanced AI systems.

This requires ensuring a framework does not have a legacy coding environment within, to completely eliminate any coding within the platform when incorporating these AI capabilities—a snap-in and play mechanism, like a LEGOS®.

When Mahalingam talks about a no-code capability he means no algorithms whatsoever are needed to deploy AI driven systems. This no-code platform “offers the flexibility, extensibility, and adaptability critical for AI technology,” he emphasized. He sees that no-code development platforms can “reduce time to market for AI systems by 75 percent, and its high cost of change management can be slashed by as much as 60 percent.”

A Stateful Network

This snap-in, no-code development must have a stateful network, stressed Mahalingam. AI bots need this feature to ensure they do not go rogue but can communicate when expected. He described it this way.

“Picture a large enterprise with $36 billion in revenue who is using 18 bots, all of whom need to talk intelligently with each other. These may be an HR, IT systems, press or aftermarket bot, these AI systems need to have dialogs without worrying about data breaches or compliance challenges.”

Means of Evolving Applied AI Systems

In addition to rules and regulations, Ramesh Mahalingam acknowledges that Applied AI Systems must adapt to changing corporate capabilities. He uses the example of OCR (Optical Character Recognition) functionality.

 “Let’s say, this year, companies eye Microsoft because it has better OCR capabilities than Google organization’s defaults. Well, businesses should be able to evolve and switch to Microsoft’s OCR potential and yet preserve the conversations between the bots that had been successfully used with the Google OCR platform.”

Mahalingam is a pragmatist, noting that “there will always be better players in the market.” We just know them when they become big. His philosophy makes sure enterprises can evolve without being locked into a vendor’s solution and having to graveyard it because it does not work well with newer technologies.

AI-Driven Systems Deployment to Scale

An adaptable, no-code, stateful framework to deploy bots turns out to be central in addressing security concerns and customizing functions. Mahalingam further explains how the technology framework described in this article will be used by one of the largest European banks who is deploying AI systems across 26 countries. He describes the bot as dealing with claims in 26 different languages. Using the AI deployment framework, claims can be managed, in an omnichannel environment, within 45 seconds. And the clean AI systems process for the bank, takes two months.

Caution and patience in building a framework to deploy bots has paid off, according to Mahalingam. This allows Vizru to  provide services for their customers via a truly scalable-models.

By closely collaborating with large enterprises Vizru has developed over 40 AI-driven use cases or stencils. Some of them are horizontal stencils such as HR, IT, and sales, while others comprise of vertical ones, specifically focusing on claims auto-adjudication, underwriting automation and policy administration, as Mahalingam clarified. “This is one of the big reasons why we’ve been able to see such tremendous traction, especially in large enterprises.” 

Particularly, where these Fortune 500 companies face an imperative to digitally transform, and incorporate AI within their ecosystems urgently, they see a snap-on, AI systems framework as attractive, and timely to gain a competitive edge.

The Future of AI-Driven Deployment

Mahalingam sees large enterprises completely transforming their core systems to human-less, back office operations. He believes hundreds of vendors will compete for edge use cases like natural language algorithms to provide support; a small piece of the AI systems pie. That type of business is likely to cause lots of noise, including concerns.

After industries cross those hurdles of security and customization, AI systems will disrupt all sorts of industries to be usage based 24/7, not just within a standard work week. It also means an explosive productivity growth of 250 to 300 percent, thanks to the introduction to AI systems.

Once core systems adopt AI, Mahalingam excitedly pointed out, we will be in “a completely AI-driven environment. It will be like the transformation Automatic Teller Machines (ATMs) made to the teller process in the 1980s.”

Image used under license from Shutterstock.com

We use technologies such as cookies to understand how you use our site and to provide a better user experience. This includes personalizing content, using analytics and improving site operations. We may share your information about your use of our site with third parties in accordance with our Privacy Policy. You can change your cookie settings as described here at any time, but parts of our site may not function correctly without them. By continuing to use our site, you agree that we can save cookies on your device, unless you have disabled cookies.
I Accept