AI, everyone pretends to know but in fact we don’t

AI, aka artificial intelligence, is the most trendy word to pay lip service in business nowadays. Yes, we mean lip service, which is just validated by a new report co-authored by BCG and MIT based on a survey of 3,000 business executives in 21 industries across 112 countries. Below are some very wholesome takeaways. 

 

High expectation and low action

In the report they say there is a big disparity between ambition and execution:

  • four in five executives would say AI is a strategic opportunity for their organization
  • but only one in five has adopted AI in some offerings or processes. Even the level of adoption we have to put into question mark, as according to a senior executive at a large Western insurance company, his company’s AI program initiatives are simply limited to “experimenting with chatbots”…dah

 

The active adoption seems to be a long way journey.

 

-Barrier to adoption

So what really stopped those companies? In a nutshell, the hurdles are:

  • Lack of AI talent
  • competing for investment priorities
  • Unclear or no business case for AI application

 

Or maybe the fundamental reason in plain language is we do not have a clear understanding of AI. The misconception might be AI is a plug&play, automated, and ready-to-use magic wand. It is said most companies represented in the survey have little knowledge of the need to train AI algorithms fed by their own data.

 

In fact, many current AI applications were born more like naked babies, with one or more crude algorithms that become intelligent only upon being trained, predominantly on a large amount of company specific data. Successful training depends on having well-developed systems that can pull together relevant training data and integrate findings over time.

 

But data is also a vexing issue for many companies. Some data is proprietary and not allowed to be open to others; other data is fragmented across different channels, requiring consolidation with multiple divisions inside the company.

 

In house or outsource controversy

For some companies, they believe the power of self-reliance by developing internal AI skills through education or hiring. But for others, successfully leveraging AI will need partnering with third parties.

 

This somehow triggers another debate surrounding the costs. Some argued that because of the barriers involved, employing an in-house strategy can be extremely expensive and difficult. But others suggested that because the AI vendor usually offers a very young product which requires tons of data and effort to allow the algorithm to learn, in the end, the juice is not that worth the squeeze.

 

Again there is no cookie cutter choice for all.

 

The learning curve

Even if companies expect to outsource AI service, they probably still need their own people to know how to structure the problem, handle the data, and supervise the evolving process. The consensus has been reached that the old dogs must learn the new tricks.  And the report reiterated the fact that the traditional approaches of tech tour in Silicon Valley to marvel at robots, autonomous cars, or attend design thinking workshop and more are useless for learning AI. Business professionals should take some time understand few basics, such as how AI algorithm learns from data, through some courses, online tools etc. 

 

At this moment, the hype of AI has been met with an unprepared generation and legacy issue. If we ask ourselves below questions: 

  • Do we understand how AI product can benefit specific business unit? 
  • Do we understand the development cost of AI based products?
  • Do we understand the theory behind building specific AI algorithm?
  • Do we understand the data training process of AI algorithm?
  • Do we understand how to get relevant data for AI training purpose?

Then how many out there can truly provide sufficient answers?

 

Author: Cecilia Wu