Here’s a reality check for AI in the enterprise


When Slack introduced its new Enterprise Grid product in January, it pledged to bring “much of the same day-to-day Slack experience that users have come to know and love” to large organizations. Similarly, CRM giant Salesforce unveiled its new Einstein artificial intelligence service this past fall to great fanfare, touting it as “AI for everyone.” But, as many enterprise leaders already know — and would-be disrupters are quickly learning — the promise of AI and its reality are, for now, two very different things.

While chatbots, predictive analytics, and intelligent search are all the rage these days, AI’s current business value is typically overstated. One analyst recently called Einstein “a great starting point,” while IT departments are “freaking out” over security concerns — such as phishing scams — due to bots’ potential to sound a little too much like real people. And that’s key — most things AI today are just that: potential. While a lot of companies are trumpeting AI as a competitive differentiator, the technologies are still in their infancy and are a lot more speculative than disruptive. That’s no doubt a relief for those frightened of the self-aware, revenge-seeking androids from film and TV.

A reality check: AI is beginning to take on the low-hanging fruit of the modern enterprise, such as handling critical time-saving tasks — like streamlining email inboxes, prioritizing/scheduling meetings, and creating data-driven daily to-do lists. Some solutions already use predictive analytics to mine the rich work graph of data within a company, adding valuable context around workflows.

As the technology improves, it will get much better at anticipating employees’ needs, as well. In the near future, voice recognition technology may even become a type of universal ID, allowing people easier access to information and experts from partner and customer networks, as well as their own companies. But to take AI further along the path from potential to practical, organizations must set aside the hype and get the right systems and processes in place. Here’s how.

Requirements for successful AI-powered collaboration

1. Overcome fragmentation

Data provides the brainpower for artificial intelligence. With the amount of data set to expand to a mind-boggling 44 zettabytes by 2020, the problem for machine learning systems is no longer a lack of information; it’s the potential for fragmentation. Without unrestricted access to a ton of data, AI can’t possibly live up to its promises — either real or imagined. Unfortunately, companies are adopting more and more disparate systems, and it’s not helping that stack vendors are continually adding more disconnected tools to their productivity suites, and emerging conversational apps are siloing information in ever-narrower message threads. Companies need their technology vendors to provide open APIs and connected hub solutions in order to make sure valuable data won’t get locked inside niche tools and to ensure that the “signal” doesn’t get lost in a clamor of extraneous noise.

Read the source article at