Bringing your AI coworker up to speed

The web is full of stories detailing all the wonderful things artificial intelligence (AI) can do, and all the terrible things as well. But now that the deployment phase is well underway, one salient question remains: What is it like to work with AI?

For most people, the AI experience has been limited to consumer releases like Siri and Alexa, which, at the beginning at least, did not exactly shine. Yes, they could name the capital of Albania and direct you to the nearest coffee shop, but beyond that, the broad impression has been that AI is not all that intelligent. In fact, it can be downright stupid.

Training the newbie

In the workplace, the first thing most people will likely notice is that AI won’t simply take over all the tedious, unpleasant jobs right out of the box. It must be told what to do first. This is a radical departure from past generations of software in which users had to be trained and retrained with each new release. Going forward, the software will change on its own, but the user must do the training.

For this reason, said Turker Coskun, group manager at software developer C3 AI, AI apps will require a lot more care and feeding than traditional enterprise programs. The performance of any number of operating models will not remain consistent over time, due to AI’s ability to ingest data and alter its own operations as circumstances and objectives change. To accommodate this, many leading AI adopters are implementing MLOps frameworks (an intelligent form of DevOps) to continuously monitor performance and kick the AI back on track if it starts to drift beyond accepted parameters.

Working with AI also requires us to acknowledge the relative strengths and weaknesses of both human and artificial intelligence. Enterprise tech advisor Chris Gale cautions users against deploying AI in a haphazard way. Instead, it should be laser-focused on solving key problems in the data chain, preferably when the amount of data in play is simply too vast for the human brain to grasp. And while it is tempting to use AI to solve today’s problems, its true value comes in crafting the next-generation platforms that will foster new business opportunities and sources of revenue. The ultimate goal, after all, is not just to empower the workforce but to create a stronger, more productive enterprise.

Refining the UX

Perhaps the key element in the AI-human relationship is the user interface, which by and large defines the overall user experience (UX). According to Stephanie Donahole, of business consultancy eTatvasoft, AI is already helping to make its own UX “thinner,” as in less complex and burdensome for the average user. With access to historical data and other sources of information, AI becomes more intuitive over time when divining user intent, and can provide assistance much quicker and more accurately than traditional automation.

AI can also make UX systems more modular and efficient, reducing overall costs for tools like self-help desks and interactive data stores, and even streamlining the analytics engines that power their own operations. In the end, we can expect to see AI deliver more personalized user experiences that should tamp down much of the fear, uncertainty, and doubt surrounding the technology at the moment.

This will be a critical challenge for software developers over the next decade, said Kaj van de Loo, CTO of UserTesting, a platform that specializes in gleaning insight from user experiences. The first task is to replace traditional interaction models that merely react to user requests in favor of a more proactive approach that seeks to anticipate what they want. To get there, however, developers will have to stop looking at AI as a separate entity and figure out how to incorporate it into familiar platforms. And perhaps the trickiest dance of all, they will have to utilize AI to radically change the way software operates and how it interacts with people, but not too quickly or to such an extent that users are put off by the technology.

AI is not the first technology to face the call to be all things to all people, of course, nor will it be the first to fail at this impossible task. Experience will most likely show that the rosiest projections of what AI is and what it can do were wildly unrealistic, but so were the gloom-and-doom predictions of robots running amok.

AI will be helpful, convenient, and valuable, but also frustrating, disruptive, and contentious. But the world, and all the people in it, will adapt.