If you're of a certain age, you probably remember the artificial intelligence craze in the 1980s
If youre of a certain age, you probably remember the artificial intelligence craze in the 1980s. AI, the pundits predicted, would bring about major changes in most enterprises, revamping the way expertise was stored and used. If you remember that, you probably also remember that AI, for the most part, failed to deliver on those vaunted expectations.
Now AI is back and in a slew of new products designed to improve online customer service. But, experts caution, keep in mind that some types of AI technologies are smarter than others, at least in automated customer service applications.
In general, avoid rule-based AI approaches, say experts such as John Ragsdale, an analyst with Giga Information Group Inc., in San Jose, Calif.
Rule-based systems, like that used by the Web search engine Ask Jeeves, from Ask Jeeves Inc., in Emeryville, Calif., rely on a number of predetermined rules to provide answers when users can make natural language queries.
The problem? Rule-based systems are generally too rigid to work well in an environment where unsophisticated users may make mistakes such as spelling errors.
Thats because rule-based approaches tend to be more rigid than other types of AI. Nor do rule-based systems do well remembering users queries and adjusting subsequent responses based on earlier experiences, according to experts.
As a result, rule-based systems may not always give users seeking online service the right answer the first time.
More effective, experts say, are online self-service bots that rely on so-called case-based reasoning.
Found in online customer support products such as those of eGain Communications Corp., of Sunnyvale, Calif., case-based reasoning systems solve problems by cataloging them, then matching them with previously solved problems.
Not only can systems that use case-based reasoning do a better job of working out what a customer is trying to say, they can also put customer questions into context. This, experts say, is the key difference between a search engine that produces a huge list of imprecise hits and one that narrows the field considerably.
Such smarter AI systems are able to remember a previous right answer to a question and serve it up more quickly and prominently the next time a similar question is asked, according to experts.
The bottom line: Ask vendors who tout the smarts of their customer self-service bots for references and find out what percentage of queries are actually getting solved in the field.
"The best systems out there are going to capture as many data points as possible," said Mitchell Nitzan, an analyst with Gartner Inc., in Stamford, Conn. "Time, day, time of year, all of those things should be able to be factored in, and the systems should be able to partition data based on the key words."
Nitzan points to Banter Inc., of San Francisco, as an example of a leading-edge AI customer support provider.