Best Practices for Enterprises Planning to Adopt AI

Enterprises that want to use artificial intelligence to solve specific business problems need to carefully study the technology to ensure that the AI system they select is the right tool for the right problem.

Remember, your goal is to solve real business problems, not simply putting in place an AI strategy. You must understand the challenges your business is facing, how you wish to solve them and whether there is an AI technology that aligns with the process.

Be aware of the landscape and do your homework to understand the length and breadth of the available systems. AI has become a buzzword, so the landscape is crowded, but early work is being done to map out the available solutions. Do your research to know all the options available.

Focus on your data. You may have the data to support an inference or prediction, but no system can think beyond the data that you give it.

Know your needs and let those drive the technology solution. Rather than start with the question, “How do I use AI?” start by looking at the problems you need to solve. Then find the technologies that can help.

Don’t confuse machine learning with artificial intelligence. Bernard Marr, a big data consultant, stated it well when he said, “Artificial intelligence is the broader concept of machines being able to carry out tasks in a way that we would consider ‘smart,’ and machine learning is a current application of AI based around the idea that we should really just be able to give machines access to data and let them learn for themselves.”

Always remember the trade-off between depth and breadth. Any system that is broad is proportionally shallow. For example, if a system understands all languages, it doesn’t understand them deeply.

Understand how your system will fit into your workflow and who will use it.

AI won’t be successful independently just because it is “smart.” It needs someone to help it learn. As such, when you acquire AI technology you are entering into a human-computer partnership.

Always think about how your systems are going to communicate internally and externally. For example, if you require that your system’s output be auditable or provide an explanation of its conclusion, machine learning alone may not be a suitable solution. Consider other AI technologies that are traceable and auditable and automatically explain analytic decisions, such as advanced natural language generation.

Remember the story of the emperor’s new clothes? Don’t be bullied by vendors or technologists who tell you their products are “too complex” for you to understand. If you don’t see how a solution works, it is because a vendor cannot explain it, not because you cannot understand it.