Features

Gregory Piatetsky: What makes a system smart enough ? Can a system be too smart?

James Taylor:

A smart enough systems is operational, real-time and customer-centric. It is easy to change as needs change (it is agile) and it "learns" from the data it collects and stores. It is compliant with the dizzying array of laws and regulations and it is modular enough (service-oriented enough) to work in a modern distributed enterprise.

A smart enough system is one that does not rely 100% on human operators for its "intelligence" and yet is not an example of AI. It is "smart" about narrowly focused things that help make the system more useful. It might, for instance, be able to make a precisely targeted marketing offer or to make an accurate assessment of risk before offering credit prices and terms. It might be able to validate a complex product configuration or price a very customized product or service.

A smart enough system makes more decisions. Instead of leaving decision-making to humans and, perhaps, supporting them, it makes good or good enough decisions automatically. Smart enough systems replace worklist and manual tasks with automated decisions, deliver higher rates of straight through processing and enable personalization and targeting in every transaction.

Smart enough systems balance the load between people and their systems more appropriately - they handle more of the algorithmic processing and data munching and leave people to focus on interactions with other people, with creativity and with other non-programmatic aspects of a transaction.

I think a system can be smart in unhelpful ways - taking too long because it is trying to perfect a decision or requiring interaction by a customer or user that is unrealistic - and I think a system can be scary if it seems to take over but otherwise no, I don't think so.