Artificial intelligence implementation: Getting past barriers

Suhit Gupta, CIO for asset management company Carlyle Group, maps out the impediments to artificial intelligence implementation -- as well as ways to convince the C-suite to invest.

As CIO for the solutions division at Carlyle Group, a $200 billion global equity and asset management company, Suhit Gupta is tasked with making sure his company can respond to and compete against disruptive forces in the marketplace. In any battle between the disruptors and disruptees, he wants to be among the former.

One disruptive technology on Gupta's radar these days is artificial intelligence (AI). Having studied it as part of his Ph.D. program at Columbia University, Gupta is very interested in the topic, as is Carlyle Group. Indeed, the company is exploring the idea of artificial intelligence implementation for a number of purposes, Gupta said recently at the AI Summit in New York: to determine how world events affect financial portfolio risk assessments, to assess potential deals the company is considering and to improve "deal flow" -- the rate at which investment companies receive business proposals -- by identifying opportunities.

Artificial intelligence is particularly disruptive now for two reasons, Gupta said: First, the proliferation of technology has vastly increased the amount of data available to businesses, and data analytics has become a "known field that people have started to trust. This is aiding why AI can move faster." Second, whereas in the past AI technology was considered a black box -- with hard-to-decipher logic and opaque inner workings -- over the past two to four years, machine learning models have become more transparent.

"The systems that are available are much more user-friendly. Data can be curated more easily," Gupta said.

Josh Sutton, head of the AI practice at consultancy Publicis.Sapient, also speaking at the AI Summit, echoed Gupta's warning call regarding disruption. Citing an upcoming survey-based report by Publicis.Sapient, he said, "The people who are most advanced from the digital transformation point of view and have done the best to integrate the silos across their organizations believe at a 3-out-of-4 ratio that their industry will be led by a disruptor in the next five years."

Why is it that if it's so obvious to the likes of us that disruption is coming across the board ... that the C-suite tends to not invest in it quickly?
Suhit GuptaCIO for the solutions division, Carlyle Group

With strong evidence that disruptive technologies can upend entire industries -- for evidence, look no further than the oft-cited Airbnb and Uber -- Gupta has been perplexed by the typical slow response to such threats by C-suites in general. "Why is it that if it's so obvious to the likes of us that disruption is coming across the board -- it's coming fast and furious, and the cost of ignoring it can be quite severe -- why is it that the C-suite tends to not invest in it quickly?" he asked.

The answer, he suggested, is linked to business' need to maintain the status quo. "Organizations are generally set up to manage risk, not so much embrace risk," he said. And, he added, companies are naturally predisposed to take action against change, citing a concept put forth by Salim Ismail, a co-founder of Singularity University, a Silicon Valley think tank: "When you try to do something disruptive, the immune system of the organization tries to attack you. That's a sad place to be."

In addition, there are other impediments to artificial intelligence implementation, Gupta said.

  • Cost. In addition to the initial outlays in software -- as well as in hardware to support the software if a business hosts the software itself, rather than in the cloud -- another big cost is the continued training of AI systems when business processes change.
  • Culture clash. "[You're] talking to people in the C-suite [who] have been doing their jobs for 30-plus years [who] have been really successful at it. And you're effectively telling them that you can partner with them and help them out. [Their reaction is to ask], 'Who's this hotshot coming in and telling me he can do my job better than I can?'"
  • Overabundance of technology options. There are so many potential services that narrowing down the field to find the right one on a limited budget is difficult.
  • Inability to precisely predict ROI. "Because AI is in some ways experimental in nature" it's difficult to say exactly how much of an improvement it may bring to a project, he said. "Getting people to understand that concept ... is a potential blocker."

To overcome these obstacles to AI, Gupta offered up two possible paths to take in convincing company management to invest in AI. He characterized the first path as the "shock-and-awe, scare-the-hell-out-of-them approach," in which IT leaders point to the perils of not investing in emerging technologies, citing advice from business gurus like David S. Rose, an entrepreneur and angel investor who said, "Any company that is designed for success in the 20th century is doomed to failure in the 21st."

The second path to artificial intelligence implementation eschews scare tactics in favor of education and persuasion. It calls for IT leaders to discuss the potential of AI with company management, to make that potential relevant to them and to make the idea resonate. Gupta advised starting out by describing how AI could bring specific improvement to a business' operations. At Carlyle Group, for instance, "We're trying to explore ... investor relations automation. Like many of you, we have an investor portal, people can log on, [and] they can get access to their information. But if they want custom reporting on their assets that are typically worth billions of dollars and spread across different instruments, it can be difficult and we often have to construct a custom report for them," Gupta said. AI could automate such reporting for Carlyle. "We can get better reporting, faster reporting and generally increase customer satisfaction. Those things resonate for the C-suite."

Once an AI investment has been given the green light from management, Gupta advised: 

  • Start small. "Small proofs of concept can be more impactful than selling them on the larger vision," Gupta said.
  • Pick "fail-fast" projects. There are two reasons for this, he said: "You'll have the opportunity to pivot much faster if things are going poorly. And [fail-fast projects become] bite-size chunks that you can actually examine the success of."
  • Engage stakeholders often. Tell them about both your successes and your failures, to bring them along on the project journey.
  • Get help where possible. "Don't be the person in the boat by yourself when there are other people [who could] help you," Gupta said.

In the end, he said, proponents of innovative technology "need to recognize the opportunity, even though there are quite a few blockers ahead of us, that one can actually convince the C-suite." If the C-suite doesn't pay attention, progress stops.

Next Steps

More on artificial intelligence

Helping the C-suite distinguish between BI and AI

AI and potential job loss

Looking ahead to AI in 2017

Dig Deeper on Enterprise artificial intelligence (AI)