The most common approach to AI implementation is to approach it as an exercise in automation. Companies look to the processes in their organization and decide which ones can be automated, thus removing or reducing the human content and saving money.
While this approach can be useful, and perhaps even a way to start, Gartner, in their release "Building the AI Business Case", points out that it is a mistake to stop there. They argue that many organizations are missing the best of what AI has to offer. That the way to approach AI is to find ways in which human effort can be augmented. That involves looking at decisions that need to be made and considering how AI can help that rather than looking at processes that can be automated.
There are several levels of AI - Reactor, Categorizer, Responder, Learner and Creator.
The reactor level involves simply automating existing processes, eg. filling orders. It's the most basic level. The categorizer level , as the name implies, is AI that can identify categories of transactions and apply algorithms to enhance those decisions.
A good example of responders is driverless cars. That level of AI can identify and react to a number of particular situations. This is quite a sophisticated level of AI. Which raises the issue that one of the considerations in implementing AI is the risk appetite of the organization, or its risk appetite in particular interactions, some being more sensitive or critical than others.
Learner levels can learn from experience and then use that experience to augment future decisions. One example of a learner level application is medical diagnosis - obviously a critical application.
An organized approach to AI implementation is critical to gaining the benefits and also to avoiding unnecessary risks. The Gartner paper offers some useful guidance.
No comments:
Post a Comment