You can't turn anywhere without bumping into artificial intelligence, machine learning, or cognitive computing jumping out at you. Our cars brake for us, park for us, and some are even driving us. Our movie lists are filled with Ex Machina, Her, and Lucy. The news tells about the latest vendor and cool use of technology, minute by minute. Vendors are filling our voicemail and email with enticements. It's all so very cool!  

But cool doesn't build a business. Results do.

Which brings me to the biggest barrier companies have in adopting artificial intelligence. Companies are asking the wrong questions:

  • What is artificial intelligence (or insert: machine learning or cognitive computing)?
  • Where can I use artificial intelligence?
  • What tool can I buy?

These questions put artificial intelligence into the traditional analytic processes and technology adoption box. These questions assume you will begin from the same starting point as you did for big data. You are wrong: Artificial intelligence starts with the problem to solve and works backward. 

To succeed at artificial intelligence you need to ask the right questions:

  1. What is the business problem I am trying to solve? AI is not an exploration for the meaning of life. AI has a purpose in life. It observes, interprets, reasons, learns, acts, and adapts. Thus, companies need to begin with how they want to make people, systems, processes, and experiences intelligent.
  2. What will I teach the artificial intelligent system? A purposeful system infers a result. Let business metrics or customer journeys guide where AI will be deployed. AI systems are taught through experience what decisions and suggestions are appropriate and which are not. AI trainers course-correct the system.
  3. What will the artificial intelligence system do? Don't be caught in the trap that AI is only about the algorithm. To solve a problem, the AI system utilizes a variety of technologies and algorithms. The human-to-machine experience or the autonomous machine experience will determine how you leverage existing systems, devices, and sensors as well as the data and algorithm. 
  4. How do I get started with my AI effort? Think crawl, walk, run. Focus on a proof of concept first that has a meaningful measure. For example, introduce an intelligent assistant with speech and natural language processing that can answer customer questions for a product purchase and be further trained to be more intelligent through these customer interactions and reduce human agent interaction. Next, extend the system to include other products and recognize cross-sell and upsell opportunities. Futher, transition the system to an advisory service to recommend or assist sophisticated purchasing scenarios such as investing or personal shopping.
  5. How will I continuously mentor the artificially intelligent system? Contrary to traditional data systems and analytics, AI systems need constant supervision. As the system hits a plateau, it may mean you need to consider additional information sources, add new senses, or add new response methods.
  6. What should I have experience with before I start my AI journey? The ability to curate data is critical to starting the AI journey. Quality data will produce trusted results. In some circumstances, baseline concepts and classifications are required as first imputs to support recognition and rules. Additionally, experience with linked data educates AI mentors and trainers in the way an AI system will think and learn through comparison and association. Semantic technologies such as graph, cataloging, and text analytics/mining provide a strong AI foundation.

In the end, to succeed at AI, don't buy or build a system and think the business will come. It didn't work when you initially rolled out the big data lake and looked for the business opportunity and adoption. It definitely won't work with AI. AI is not a standard definition for all purposes; it is purposeful for the business or product objective.