Cognitive computing and AI have become a big part of business plans going forward. A recent survey finds one in three businesses say they already have fully implemented AI across their organizations. Another 20% have implemented AI at parts of their business.
It’s surprising to imagine that 33% of today’s organizations are already full-functioning AI enterprises. It’s not clear from the survey results, posted by EY, what respondents consider to be “AI” – it’s possible they are looking at advanced analytics and algorithms and branding that as artificial intelligence. It can be assumed that since the survey base was mainly AI professionals, it’s something close to cutting-edge AI. Still, “AI” is a very fluid term. Even the US Congress is on the case, with a proposed bill directing the Department of Commerce to define exactly what it means. (Sounds like a job for NIST.) In the meantime, David Gershgorn wrote up an excellent overview of the true meaning of AI in Quartz, defining it as “software, or a computer program, with a mechanism to learn. It then uses that knowledge to make a decision in a new situation, as humans do.”
Regardless of what it actually is – AI, analytics, machine learning – it requires skilled people to make sure it is capable of delivering at least some value. In my previous post, I explored some of the new types of jobs that are arising out of the cognitive computing and artificial intelligence space. For executives or managers, it means identifying what skills make a difference, then supporting appropriate training or hiring. It’s likely “Voice of the Customer analysts” or “automation economists” don’t even exist yet in most professional pools.
In the EY survey, a majority of senior AI professionals say they are hurting for such skills. Fifty-six percent say a lack of talent is the greatest barrier to implementation within their business operations. Skills is the greatest barrier, and managers also point out that AI technology is a barrier – 46% feel it has still not reached maturity, so it’s important to take a wait-and-see approach. Many of the leading vendors are bolting AI onto their existing offerings, so it may be more ubiquitous in the coming months and years.
Stakeholder buy-in – presumably meaning line-of-business managers and key corporate decision makers – also is an issue among one-third of the organizations represented in the survey.
Where to begin? The EY study points to IT itself as the place where most AI implementations are happening – cited by 46%. Customer service, along with sales and marketing, follow with about one-third of organizations.
It sounds like AI – however you define it – is off to a great start in its current form. So what are the keys to “selling” AI adoption across the enterprise. Here are some points to consider:
Always put people first. AI and its cognitive siblings can be either threatening or confusing to people. The key is to help people become comfortable and even skilled at handling AI systems. Businesses have been “hampered by a shortage of experts with requisite knowledge of the technology,” according to EY’s Chris Mazzei. “This serves to demonstrate that successful AI integration is not just about the technology, it’s about the people. Looking to 2018, organizations should prioritize talent acquisition and cultivation – both by recruiting individuals with strong technical backgrounds and investing in skills and training programs to help retain and foster leading AI practitioners.”
Simplify the science of AI. Right now, “AI requires very sophisticated human resources, such as data scientists to build machine learning models, and computational linguistics professionals to write knowledge extraction applications,” according to Sudhir Jha, senior VP at Infosys. “This restricts AI applications and innovations to a select few, and consequently limits the speed of adoption within the enterprise. But this scenario will not last long. Technology companies are building tools to automate tasks performed by these skilled individuals, thus enabling even a data analyst or business user to build AI applications.”
Increase the auditability and ‘explainability’ of AI. Right now, AI operates in black boxes, especially when it comes to deep learning, but Jha notes that there is early work going on with Explainable AI (XAI). But there needs to be better “auditability and basic visualization tools to take steps towards a system that doesn’t behave like a black box,” Jha says. “It would be hard to imagine wider adoption of AI within the enterprise without it.”