Enterprise AI - Building Systems of Intuition

Abstract

This paper discusses how creating an enterprise-scale AI platform is not like creating a traditional IT platform. While some of the methods and approaches from IT platform development apply, there are some essential differences which are crucial for a successful enterprise AI implementation.


AI platforms are inductive systems (different from rational systems that are built on logic and explicit rules like many traditional IT platforms) that learn from past experiences. In that sense they are “Systems of Intuition” that build their own reasoning through learning and in doing so exhibit human like capabilities, albeit for a very narrow set of outcomes. This wave of AI is different because it has all the characteristics of a “General Purpose Technology (GPT)”. Such technologies re-shape the entire socio-economic structure (just like what steam engine and electricity did and so did IT) as they span across the entire economy through direct and indirect linkages with growth and transformation.

Building an enterprise scale AI platform requires a set of pre-requisites. Treating them like a typical IT system that relies passively on back-end data would prove to be a recipe for failure but they do build upon some of the building blocks of a digital enterprise. Organizations that have invested wisely in building their digital muscle through a well-defined digital transformation roadmap are better positioned for embracing AI than the digital laggards. The four most important ingredients of an enterprise AI platform are:

  • Business use case selection
  • Labelled data for training the models
  • AI experience (AIX) design (AI in the loop)
  • Critical role of the top management sponsorship

Data has a dual role in this entire storyline – one that makes the AI agent potent by helping it learn from the past and the other part that flows through the AI agent as an operand in a more passive role. The central premise of this paper is to help the reader understand the critical components of an enterprise AI platform, how to go about building such systems of intuition and how treating these like any traditional IT systems (systems of record and systems of engagement) may not be prudent. As is always with new technology paradigms, such efforts will be laced with hits and misses rather than a path filled only with a series of successes. Hence, top management leadership would need to play a prominent and a decisive role in making sure that AI is embraced in the right and timely manner (right business use case) to sustain the company’s competitive advantage and that its business impact is measured using the right yardstick.