Digital Technology
Jun 19, 2017

AI for the enterprise: Beyond fun and games

Artificial Intelligence (AI) has garnered much media attention lately. Most of the media coverage has focused on what the five big tech titans—Apple, Microsoft, Google, Amazon, and Facebook—have been doing with AI in the B2C space, like developing conversational user interfaces for their home products or beating the best humans playing Go, poker, or video games.

Much less has been written about how AI is being used—with significant transformational impact—by businesses. Although still very early in its overall adoption life cycle, AI use cases within the enterprise have been accelerating over this past year. Although perhaps not as interesting to the general public, this early AI usage within enterprises is providing significant business impact. That said, successful implementation of AI-based solutions in the enterprise creates different requirements and challenges, including:

  • Regulatory compliance requirement: For B2C-focused AI offerings, consumers don't really care to know exactly how an answer was determined; they seem to accept that AI has a “mind of its own." This is NOT the case within the enterprise where companies must be able to document the rationale for why their AI solution provided the particular decision made (i.e., Why was this loan application rejected? Why was this supplier rejected?). Enterprise requirements for ensuring compliance with industry regulatory audits add another hurdle complexity for AI-based solutions.
  • Lack of huge datasets needed to train AI systems: Most AI-based solutions require access to VERY large data sets in order to be able to effectively train AI systems. The large B2C tech titans playing in the global consumer space have access to this needed magnitude of data; however, on the B2B side, most enterprises do NOT have this access. This challenge necessitates looking at different approaches to be able to effectively train AI-based solutions within the enterprise.
  • Smaller R&D budgets/faster project development cycles: Unlike the large B2C tech titans who now are the five most valuable global brands , most enterprises do NOT have similar sized R&D budgets and/or the luxury of longer development timeframes. Most enterprises need to see successful proof of concept (POC) results before receiving funding to achieve scale. In addition, enterprises need to ensure their AI investments do NOT stretch out into multi-year projects. ROI and time-to-value are driving the need for faster AI-based project implementation in the enterprise.

Meeting enterprise-specific AI needs

 Although these enterprise AI requirements appear challenging, there are new approaches for AI solution development that are well-suited to meeting enterprise needs. To solve the regulatory compliance requirement, unlike mathematical algorithm approaches used in the B2C world, there are advanced computational linguistic approaches, which power AI systems and support the trackability of the answers provided. Computational linguistics is better suited for enterprise usage because much corporate “data" lives in the form of some type of document (e.g., PDF, Word, online form, etc.) which is better suited for AI (versus, say, photos of a cat on the B2C side). In other words, these advanced, yet pragmatic, AI solutions for the enterprise are not a “black box"; instead, their interpretations and answers are fully trackable to enable compliance with regulatory audits. 

Another benefit of leveraging computational linguistics is a MUCH reduced need for a large data set to train the AI system. This approach means that an AI system in the enterprise can learn enough to be effective—without requiring zettabytes (i.e., a huge quantity) of consumer data. This smaller data corpus requirement significantly increased project implementation speed, an important need within the enterprise. 

Lastly, new zero-code and model-based development approach for AI systems are well-suited for the enterprise. When designed with this architectural model, these advanced AI systems are well-suited for enterprise budgets and project deployment timeframes. In fact, some AI POC projects—which leverage the above-described technologies and approaches—can be up and running in enterprises within 90 days; this is a long way from yesterday's multi-year project development efforts. 

Moving forward

Some business-focused publications, like Harvard Business Review and Fortune, have picked up on the significant transformational impact which AI systems are expected to have within the enterprise. We see our enterprise clients investing on digital transformation journeys, typically starting with more fundamental automation and advancing into AI solutions.

Today's AI solutions which are able to meet enterprise requirements enable companies to dream big, start small, and scale fast. Increasingly, this mindset and operating approaches are key fundamentals for emerging and disruptive business models. 

Back in 1993, William Gibson famously said: “The future is already here—it's just not evenly distributed." AI solutions for the enterprise are similarly “already here." Companies that are getting started now will become tomorrow's leaders… and others—if they are still around—will get started in the future.

About the author

Dan Glessner

Dan Glessner

Vice President, Digital

Follow Dan Glessner on LinkedIn