A practical approach to introducing AI into corporate environments
By Willem Conradie, Principal Consultant: Big Data at PBT Group
Gartner estimates that the global business value of Artificial Intelligence (AI) will reach $1.2 Trillion, this is more than the GDP of most countries in the world. Gartner categorises the sources of AI business value as enhanced customer experience, new revenue, and cost reduction. One starts to wonder why the world is not falling head over heels for AI. In short, AI is about relinquishing control to an autonomous entity that acts and makes decisions without any human intervention. One might argue being cautious of the “unknown” is the major reason companies are hesitant.
The rise of Data Science is leading the introduction of AI into the corporate world. While Data Science is not AI itself, Data Science is top of mind for every “data” driven company because it is these Data Scientists that “teaches the artificial engine to become intelligent” through statistical descriptive, predictive and prescriptive models.
For many corporates the difficulty is finding skills in the market. The “unicorn” Data Scientist is rarely found, and if found, unaffordable to most. This is forcing companies to employ graduate, or lesser experienced, Data Scientists straight out of universities. The challenge then becomes in implementing the Data Science in an operational environment for sustained benefit, often referred to as the ‘last mile’ in Data Science.
Very few companies succeed at deploying “sustained” Data Science. This is where the so called “unicorn” comes in. Graduate Data Scientists, even though they are highly educated for their trade, tend not to have the necessary skills and experience to deploy the Data Science operationally. They are educated on the science side, but unfortunately not on what it takes to sustain Data Science in a corporate environment.
For sustained Data Science solid governance, architecture and data engineering are just as vital. The sustained value of Data Science lies in the complete end-to-end life cycle, which includes the “last mile”. But what does the ‘last mile’ entail? The ‘last mile’ is a lot about good Data Science governance: the data science process, business and technical architecture, model management, model performance monitoring, systems monitoring, business continuity strategy, and last but not least, trust.
Trust is the most difficult one. Data Science, at the moment, is for many an acceptable way of introducing AI into the corporate world as it is typically performed by a person. Trust entails the Data Science process itself, and even more so, trusting the AI to make decisions on your behalf. For many, this is the biggest risk and the most difficult to accept.
Another option might be to ignore AI, and not face the “trust” risk at all. However, the lost opportunity cost can be substantial. Companies who introduce AI and succeed will get the first bite at the $1.2 Trillion pie. This is where good Data Science governance practices play a vital part. It’s difficult to eliminate all the risks of AI in corporate environments, but it is very possible to manage it.