Forget ‘mobile first’ and ‘cloud first.’ Modern applications being built today need to be ‘cognitive first.’

That’s according to Progress CEO Yogesh Gupta, who said intelligent applications need the capabilities to predict and to anticipate, and thereby help businesses become more successful.

And he’s not the only one. I would say a solid six of 10 calls and pitches I get each day involve some aspect of artificial intelligence or machine learning. I’m told that in the not-very-distant future of the Internet of Things, back-end systems will have to understand data, because the stream will become too thick for humans to be able to handle the load. Machines will have to learn what data is critical to the business and what doesn’t have to be dealt with right away, what will make customers happy and what will drive them away.

AI will be used more widely in software testing, as systems learn about themselves – and  to catch errors introduced during builds that might break the software. Test automation with AI will enable the system to order tests according to what it learned from prior defects, and suspect that something is wrong if that data appears again.

Applications will have to become intelligent, to understand which device I prefer to use to interact with them, to know my preferences and deliver to me information that is has determined from my actions is most relevant.

“We see a whole world of business applications going smart, able to predict, to anticipate and therefore to help businesses be more successful,” Gupta said.

In this issue, we look at what it takes to create a fantastic user experience. Gupta pointed out that the user experience is much more than the user interface. “The experience goes beyond to cover the interaction itself,” he said. “The interfaces have to understand context.”

He further noted that in a fairly short time, no-UI apps will become the norm. “Take the thermostat,” he said. “It should know how I want the house when I’m home. Data drives that learning. We see that in spades in industrial applications, which can do predictive maintenance” based on data showing in real-time how the machine is performing, instead of doing maintenance on a timeline, when it might not be needed.

To bring this world to life, Gupta said developers will need tools in three categories to create an app AI architecture:

  1. Machine learning engines
  2. A rules engine. ‘Let’s say we have a prediction. What the are business rules that define how to deal with it.” These, he said, or rules and business policies more than code writing.
  3. Modern user experience. “The interface could be conversational, like a chatbot, or AR/VR, or a mobile device.. whatever,” he said.

“All of this,” he added, “has to tie into a back-end platform to run your business. You need to run the business apps in a scalable, secure environment, with data connectivity and front-end tooling. We think that’s the architecture” for modern, smart applications.

Jeffrey Hammond, research analyst at Forrester, told me that “AI at its core is proactive, not reactive, inferring real-world connections based on data patterns,” and triggering actions based on predictions. So, for instance, a system monitoring water pipes in a city might detect a drop in flow from one pipe, which could indicate a leak. It automatically shuts down that section, reroutes if possible, and sends an alert to the utility repair crew to get out and fix it. This prevents those massive street floods we see from burst pipes, saving money for both the city and those that would be damaged by the huge water spill.

We’re even now starting to see cognitive services moving beyond language and speech recognition into gestures, empathy, vocal tone analysis and sentiment analysis.

“In the cognitive era, we’ll see people developing against cognitive capabilities and coupling that with data science. It will be a big responsibility of skilling folks, to articulate what’s being done and making use cases available” for developers to learn from, said Willie Tejada, chief developer advocate for IBM Watson. “We need to create on-ramps for software assets, tool chains and code and show how I design a retail chatbot or how do I do data science against a Twitter feed?”

As Progress’ Gupta said, “We still have a long way to go.”

About David Rubinstein

David Rubinstein is editor-in-chief of SD Times.