“Object in Mirror” by Sabda Renjana is licensed under CC BY 2.0
One of the hardest challenges you'll face with introducing AI-driven enterprise search is getting it in the door PHOTO: “Object in Mirror” by Sabda Renjana is licensed under CC BY 2.0
One of the biggest challenges of bringing artificial intelligence (AI) into enterprise search is simply getting it in the door. 

The people building information management solutions have little interest in conversational applications. The inertia of ongoing projects (and of the new AI snake oil salespeople who promise the world) will siphon all available resources away from approaches requiring a strong knowledge architecture foundation.

Because we have few examples of big successes in this area, executives can break out in a sweat when asked to back something lacking short-term ROI. At the same time, executives are justifiably concerned about being left behind. 

Where's the ROI?

To move ahead, executives need to know the limitations and when they are engaging in pure research versus experimenting with innovative approaches and business models.  

You will not realize the benefits from AI-enabled search immediately. Not only do the machines need to be trained, but the humans do as well. 

It’s not just about building a search bot, it’s about learning how to interact with these systems. This learning process is a form of discovery, socialization and communication: users need to learn what they can do and this is not as intuitive as the Amazon Echo TV ads would have us believe. 

Corporate information is complex, and knowledge work is nuanced. Work is hard, which is why they call it "work." The tools require some degree of training to make the most of them.  

ROI metrics can be “hard” and measurable or “soft” and less tangible. 

The former could include a significant impact on “time to access information” or “time to complete a process,” given accurate baselines and metrics. However, companies rarely measure things like “time-to-information” or “employee efficiency” unless that efficiency is carefully tied to a specific process. 

Knowledge workers’ gains and losses related to time spent looking for information must be tied to adjacent performance measures. Time-to-market is one measure that ties to business efficiency — again, the key is linkage. 

Some organizations see correlations between information system efficiencies and measures such as “rate of voluntary terminations,” and “employee satisfaction.” While these measures do not always come with hard dollars attached (though high turnover is costly to any business), most companies realize beating the competition to market requires retaining talent and corporate knowledge in order to create a competitive advantage. 

Where’s the Talent?

AI-enabled search applications requires new thinking and new teams. Your employees who are currently using traditional methods to solve the same business challenges are ideal candidates for this team, the trick is getting them to think in new ways around natural language conversation responses to queries. 

Some of the talent required may come from outside the traditional technology department roles. People with backgrounds in the liberal arts, creative writers, authors, theater/performing arts all help here — these people intuitively know the conversational language that occurs across many walks of life. User experience designers interested in a new user interface paradigm — the Conversational User Interface (CUI) — can also be important contributors.

Crafting scripts and responses in conversational language is a new skill set. Is there a “skills gap” to overcome? Demand is higher than the supply of the conversational language programming skillset. Amazon, Microsoft, Google are all hiring storyboard writers right and left, from Pixar and similar companies. 

Today's AI Is Tomorrow's 'Search'

We live in an “interrupt world.” When we can’t find the information we are looking for, we interrupt our colleagues in the hopes that someone knows where to find our data. 

This not only interrupts the requester, but also those targeted with the requests, and the impact may domino through several layers of the organization. Large enterprises suffer the most, battling through scattered information repositories and dispersed resources. 

AI-enabled search promises to put a dent in this problem. The hype engine promises more than that, but for the purposes of this discussion, let’s be realistic and be happy with a dent.

Knowledge is the currency of business today. We've all heard the mantra: “Knowledge is power.” But if you don’t build knowledge bases related to the “secret sauce” for how you do business, you will not lead … you will lose.

Deciding to rebuild an enterprise search paradigm means taking resources away from how things are done today. It takes strong leadership — maybe even a change agent hired to disrupt — to realize the status quo isn't working and act accordingly.

Finally, realize this: It’s AI when it involves knowledge or information previously thought to be only in the realm of humans. But as soon as it's automated, we stop calling it AI.  Spell check, grammar checks, then style suggestions were all once considered AI. Now we take them for granted as part of word processing. 

So maybe you’ll now ask Siri different types of questions or make travel arrangements conversationally. This is today’s AI, but soon it will just be regular old mundane “search.”

Editor's Note: Be sure to read the previous two installments in this series on the potentials for conversational AI search and the foundation required to deliver it.