It’s the rise of the machines!
Stories of robots endowed with Artificial General Intelligence (AGI) that take over the world have been with us for decades. Although it’s entertaining to contemplate and it’s theoretically possible, it’s not likely to happen in our lifetimes. That said, you can’t deny how rapidly artificial intelligence (AI) technology is approaching AGI or AI-complete.
And although systems and machines cannot fully match a human’s intelligence, they continue to make our lives easier as AI technology becomes more sophisticated. So, there’s no reason why AI can’t supplement or even replace some of the experts at your company–or is there?
SEE: Special report: How to automate the enterprise (free ebook)
In a previous TechRepublic article, I shared strategies for building an expert system, though I didn’t address the cultural challenges you’ll face when trying to deploy advanced expert systems; in a recent McKinsey article, Murli Buluswar, chief science officer at AIG, highlights this as a formidable challenge for the analytic leader. It’s important to deploy expert systems in your company; however, don’t forget about the human experts as you make the transition.
Subject matter experts don’t have to be humans
Subject matter experts are assumed to be human, but it’s not a strict requirement. The leader who’s moving his or her organization in a more analytic direction should embrace AI to whatever extent possible, and organizational expertise is not off-limits.
When I worked with Visa on its enterprise data strategy, there were certain people who were in very high demand because they were experts–sometimes the only expert–in their particular domain of knowledge. While that was job security for the expert, it was a great risk for the company, because important initiatives that required that expert’s knowledge were on hold until he or she was available.
But the dependency wasn’t on the individual as much as it was on the individual’s knowledge–something that could be stored in a system instead of a human’s grey matter. In fact, we’ve become accustomed to using systems as experts in certain scenarios. We build large data lakes where volumes of structured and unstructured data are harnessed and then layer-on elaborate analytic systems to bring us insights. How is this any different from offering a 30-year veteran of the company a penny for her thoughts? Although the two use cases are almost identical, it’s the human condition that forces us to treat these scenarios differently.
SEE: Inside Amazon’s clickworker platform: How half a million people are being paid pennies to train AI (PDF download)
Humans vs. systems
Humans have general intelligence and machines don’t. That doesn’t mean machines can’t outperform humans in certain mathematical challenges–a simple calculator can do long division a million times faster than a human. But I don’t think the onboard guidance system of US Airways 1549 could have landed safely in the Hudson River like Captain Sullenberger did.
Humans are also easier to deal with–we are social creatures. And although we might not get along with certain members of our species, as a rule, humans engage best with other humans.
And yet, expert systems are largely undervalued and underutilized. The base requirement for expertise in any given scenario doesn’t always extend beyond the limits of AI, especially a carefully constructed expert system. Remember, an expert system starts with human experts and is subsequently trained with their help. And although it’s no match for the collective wisdom of the humans that brought it to life, over time the expert system should be able to handle at least the basics. Add to that, much higher availability, reliability, and manageability; and it looks like there’s a nice spot for expert systems in your organization–if the humans don’t mind.
SEE: 6 ways the robot revolution will transform the future of work (TechRepublic)
So, how do you think the subject matter experts feel about being displaced by a system? By the way, these are the very experts you need to build the expert system in the first place. It’s like training an apprentice to take over your job–not very encouraging or motivating. In fact, I’ve been in situations where experts purposely withheld information (from humans and machines) for fear of obsolescence. Even if your experts are cooperative and forthcoming with their knowledge, you’d be remiss if you didn’t take their concerns seriously.
Your transition from subject matter experts to expert systems must include a well-developed organizational change management plan. This means, your objective must expand to include what the future looks like for the experts once the system goes live. I recently worked on a project with Chevron to standardize the way inspection is performed in all its operated refineries; The biggest challenge I dealt with was removing discretionary autonomy from senior inspectors. You’ll need to manage a similar shift from know-it-alls to need-to-knows. Your experts must shift their mindsets from answering questions to questioning answers. This is a difficult but required transition for a successful deployment.
SEE: Deep learning for customer service is overhyped: How one company did it faster and cheaper (TechRepublic)
Making the shift from subject matter experts to expert systems is a smart move for your company, but it might not go over too well with your experts. At best you’ll get overt resistance from experts who feel like they’re helping their way out of a job; at worst, it will be subterfuge that will insidiously sabotage your efforts.
I’ve outlined the case for expert systems, highlighted your key organizational risks, and suggested techniques for making the transition; the next move is yours. You can continue to be constrained by the vital few who hold the keys to your organizational wisdom, or you can make the move to an expert system. I hope you make the right choice, but don’t go too far–if your expert system ends up taking over the world, I’ll be very upset.