It's a peculiar admission to make in a computing magazine, but I generally prefer books to computers. So when I wanted background information on artificial intelligence, I overlooked the encyclopedia grooved into my CD-ROM and picked up my faithful Britannica. Unfortunately, my encyclopedia is now thirty-something and so makes no mention of artificial intelligence. It skips adroitly from articulation to artificial respiration. No mention of artificial sweeteners, artificial food additives, or artificial insemination for that matter. Times have changed, and "artificial" is a uniquely modern modifier.
The CD revealed that artificial intelligence (AI), as applied commercially, is little more than clever programming that uses the speed and volume-processing capabilities of the computer to solve problems, identify patterns, or predict results. I prefer thinking that AI is the result of being artificially inseminated with MENSA sperm, but that's another column.
AI has been around for a while but businesses have been slow to take advantage. The resistance may be partially due to a misunderstanding of how AI works and whether it can be trusted. As one businessman told me: "It sounds too tricky. I mean, a faulty chip, a bad line of code, and what have you got? Art Official Intel Agents." Sounds ominous.
But not menacing enough to deter Mary Lamphier, vice president of Information Systems at Veratex-a Troy, Michigan-based medical and dental supply company. Back in 1990, after leading a successful migration from the S/38 to the AS/400, she pondered how Veratex might leverage the new platform's capabilities.
Veratex is a telemarketing-dependent company. Although it posted a tidy $58 million in sales last year, it has no outside sales force. It is therefore critical that the 60 people flogging the phones know as much about the com- pany's customers and extensive product line as possible. For new employees, that was problematic. Veratex has a customer base of 106,000 medical professionals, and its product offerings support five highly specialized markets: medical, dental, veterinary, pediatric, and institutional. The jargon alone (what on earth is a digital Wig-L-Bug mixer?) would be daunting.
Lamphier was reading an article about AI in her office one day when she began to see the possibilities. But there was such scant business-partner support for AI applications that she despaired ever being able to implement an AI solution. But Lamphier lives right. Serendipitously, she receives an unsolicited call from business partner Churchill Systems, rich in AI application experience and located-get this-just around the corner from Veratex. Kismet.
Using Knowledge Tool/400 and Neural Network Utility/400, Churchill developed both a knowledge-based system and a neural network for Veratex. Unlike conventional programming, neural networks are not rule-based, but are trained, fed. They digest data samples and discover relationships. Once trained, the network can use known patterns to make predictions regarding new data.
The process is called "pattern recognition" and is one of six commercial and scientific applications for AI (the other five being problem solving, game theory, natural language processing, cybernetics, and machine learning). Pattern recognition is favorably suited for a medical supply company since it mirrors the procedure doctors use to classify medical problems on the basis of symptoms.
Applied to sales analysis, the system first absorbed customer demographics and buying patterns. Customers were ranked A though D, with A customers generating higher sales and profits. The neural network could then predict which D customers had the highest potential of becoming A customers based on matching demographics, much as a fingerprint identification system finds sufficient matching patterns in partial prints to identify a suspect with some assurance. The system does not use just any A customer as the archetype, but the best A customer. Such information allows telemarketing agents to concentrate on prospects with the highest purchasing potential rather than calling cold.
The impossible challenge of mastering a product inventory of 19,000 items was solved by a knowledge system that not only understands doctor-speak but can suggest complementary and substitute products. To assist telemarketing trainees unfamiliar with medical nuances, medically correct software modules, which reside on the AS/400, supply product information for a variety of specializations-from periodontics to oncology. If an ordered product is not in stock, the system suggests an alternative product. And to generate additional sales, when a particular item is ordered-needles, for instance-the system will automatically suggest related products the customer might need such as cotton balls, or Band-Aids, or duct tape to keep the patients strapped down.
Given systems that provide telemarketers with comprehensive account information, correct medical terminology, product recommendations, and credit history, the training cycle for sales reps was reduced while their productivity increased. Within a month of installation, sales of one product line swelled 14 percent. Over a two-year span, AI proved to be a helpful tool, making its own discreet contribution to a wider effort which yielded 13,800 new accounts with sales of $7.8 million.
Lamphier has expanded the system to take advantage of the AS/400's imaging and fax capabilities. By law, Veratex must provide material safety data sheets to its customers for certain products. The sheets are scanned and stored on the AS/400, and they automatically print and ship with the first order. The second time a customer orders the same product, the ordering history is available to the sales rep, who verifies whether the customer has the data sheet on file. If not, a copy is immediately faxed by the AS/400.
Conversely, customers must provide Veratex with proof of federal and state licenses for controlled substances. The AS/400 keeps track of customers with such requirements and notifies sales reps when licenses are due to expire within 60 days. As a courtesy, the rep can notify the customer, and a renewed license can be faxed to the AS/400 or mailed and scanned.
The company further plans to use imaging to give sales reps instant access to its direct-mail catalog. When customers call in orders, the telemarketer will be able to view the same page of the catalog as the customer. Imagine, two confused people staring at a picture of a digital Wig-L-Bug mixer.
The installation of the AI applications was completed in just six months. But the company has been continually expanding the system for the past three years. "Here at Veratex, we are constantly moving," says Lamphier. That, perhaps, is a reflection of her own restless background. The daughter of a serviceman, she lived in France and England, attending no fewer than 13 schools and finally graduating in Tokyo. Prior to becoming the chief information officer at Veratex, she played a little hardball with the Detroit Tigers, automating their ticketing, financial, and scouting operations.
Her success at Veratex hints at a promising midrange future for AI. The technology, which was something of a gamble when Lamphier first hitched the Veratex wagon to it in 1991, now joins so many other dizzying innovations on the retail shelf.
Still, there is an awkward mutual-exclusivity about the words "artificial" and "intelligence." Like military intelligence, it is not without irony. And there is an irony in the Veratex implementation of this technology. Here we have artificial intelligence selling medical supplies to treat people whose bodies are perhaps ravaged by artificial toxins. It brings to mind another uniquely modern concept: synergy.
Victor Rozek has 17 years of experience in the data processing industry, including seven years with IBM in Operations Management and Systems
LATEST COMMENTS
MC Press Online