RECENTLY, it was announced that Malaysia would be setting up                                                                                          
the National Artificial Intelligence System Development                                                                                               
Laboratory (AISDEL) to co-ordinate the use-of AI technology in                                                                                        
the commercial and industrial sector.                                                                                                                 
                                                                                                                                                      
The Japanese government is set to provide assistance worth                                                                                            
about RM7.8 million. Joint venture with local companies to                                                                                            
harness the technology is also being looked into.                                                                                                     
                                                                                                                                                      
Increasingly, AI technology is being explored as a strategic                                                                                          
business tool. Management guru Tom Peters warned that "any                                                                                       
senior manager who isn't at least learning about expert                                                                                               
systems, and sticking a tentative toe or two into AI's waters,                                                                                        
is missing a significant opportunity."                                                                                                           
                                                                                                                                                      
What is the nature of this AI technology? And how may this                                                                                            
technology be useful in the business context?                                                                                                         
                                                                                                                                                      
Knowledge-based systems are sometimes called classical AI, as                                                                                         
they were the first technology to come out of the research                                                                                            
laboratories and to the marketplace.                                                                                                                  
                                                                                                                                                      
Expert systems fall under knowledge-based systems.                                                                                                    
                                                                                                                                                      
Expert systems (ES) being the oldest and earliest AI                                                                                                  
technology has a long history of commercial use, though its                                                                                           
penetration level is not as high as predicted by its                                                                                                  
enthusiasts. But it has been steadily seeping in the                                                                                                  
commercial market, as the technology becomes cheaper and more                                                                                         
familiar (it is now routinely taught to business                                                                                                      
undergraduates) as it loses its high-tech sheen.                                                                                                      
                                                                                                                                                      
There are two important uses of ES, namely knowledge                                                                                                  
dissemination and data sieving.                                                                                                                       
                                                                                                                                                      
Expert systems allow efficient accumulation and dissemination                                                                                         
of knowledge by providing easy and rapid access to employees,                                                                                         
eliminating time and space constraints. The use of expert                                                                                             
systems by frontline customer service employees is an example.                                                                                        
                                                                                                                                                      
                                                                                                                                                      
The use of expert systems in customer service is appropriate                                                                                          
be cause of the ability of expert systems to generate                                                                                                 
customised responses.                                                                                                                                 
                                                                                                                                                      
With an expert system installed as an aid to selling and                                                                                              
cross-selling, shorter training time would be required for                                                                                            
front-line employees. Training can be switched from a                                                                                                 
"product-driven" approach to a "client-driven" emphasis.                                                                          
Instead of focusing on the product, more effort can be placed                                                                                         
in maintaining effective client relationship.                                                                                                         
                                                                                                                                                      
Another interesting area would be customised client                                                                                                   
correspondences.  Form-letters have been generally used to                                                                                            
address this need, but the need to generate high-quality                                                                                              
individualised letters cannot be adequately resolved.                                                                                                 
                                                                                                                                                      
Mailing form-letters to your clients seem to lack the personal                                                                                        
touch and they would feel like being treated as a cipher.                                                                                             
                                                                                                                                                      
One solution would be to have as many form-letters to address                                                                                         
the common problems-and this could number up to thousands.                                                                                            
                                                                                                                                                      
However, in practice, most of these letters would never be                                                                                            
used as nobody can remember which ones to use and would                                                                                               
usually settle for 10-20 familiar letters that fit closest to                                                                                         
the situation at hand.                                                                                                                                
                                                                                                                                                      
To resolve this need a major credit card Organisation has                                                                                             
contracted for an Intelligent Correspondence Generator (ICG)                                                                                          
to be developed.                                                                                                                                      
                                                                                                                                                      
The result of the implementation was a reduction of turnaround                                                                                        
time for a handwritten letter from an average of three days to                                                                                        
five minutes for ICG. The Quality Control Department judged                                                                                           
the ICG output to be 95% error free as compared to 80% for                                                                                            
handwritten letters. The workflow (Fig 1) has also been                                                                                               
simplified as a result of ICG.                                                                                                                        
                                                                                                                                                      
The other basic use of the expert systems would be in the area                                                                                        
where the data volume and rate is overwhelming, making it                                                                                             
physically impossible for any individual to monitor it or make                                                                                        
effective use of that knowledge.                                                                                                                      
                                                                                                                                                      
John MacGregor of MacGregor & Associates (Australia) recently                                                                                     
spoke on the use of data mining in Executive Information                                                                                              
Systems. Traditional Executive Information System (EIS) offer                                                                                         
trend analysis, drill-down and exception reporting.                                                                                                   
                                                                                                                                                      
However, this approach is data passive and user-driven. Users                                                                                         
have to discover the patterns themselves, and may miss unusual                                                                                        
valueS and trends. E IS also do not address the issue of data                                                                                         
volume.                                                                                                                                               
                                                                                                                                                      
MacGregor recommends a data active approach using intelligent                                                                                         
processing to augment executive analysis. Areas where this                                                                                            
would he useful would be where process control is used as in                                                                                          
manufacturing or during/after business process re-engineering.                                                                                        
                                                                                                                                                      
                                                                                                                                                      
Pattern recognition tools have a wide variety of uses in the                                                                                          
commercial or industrial sector. They are also used in the                                                                                            
implementation of expert systems knowledge base to find                                                                                               
patterns in the data. Pattern recognition tools are used where                                                                                        
repeated patterns need to be recognised expert systems can be                                                                                         
used if the data are symbolic, ie consisting of descriptions                                                                                          
not numbers.                                                                                                                                          
                                                                                                                                                      
Pattern recognition tools are also used where it is needed to                                                                                         
discover the pattern (ie the relationship between data points)                                                                                        
itself; this technique has no counterpart in expert systems.                                                                                          
Machine learning techniques called conceptual learning are                                                                                            
used for symbolic data learning.                                                                                                                      
                                                                                                                                                      
Business Week (October 1991) reported that researchers at                                                                                             
Georgia Institute of Technology (Atlanta, US) have developed                                                                                          
computer sensors that can detect wear or breakage in the                                                                                              
sewing-machine needles before they can damage the apparel. The                                                                                        
sensors detect the distinctive sound emitted only by the                                                                                              
broken or worn needles. While it was not reported specifically                                                                                        
that automatic pattern recognition techniques are used, this                                                                                          
is precisely the situation where it would be useful.                                                                                                  
                                                                                                                                                      
Artificial neural networks generally have 80-90 per cent                                                                                              
success rate of recognising a pattern.                                                                                                                
                                                                                                                                                      
The two key AI technology, expert systems and pattern                                                                                                 
recognition/machine learning systems are closely related so                                                                                           
that they can be combined. Sometimes expert systems are called                                                                                        
pattern-directed inference system.                                                                                                                    
                                                                                                                                                      
The power lies in their ability to recognise patterns, trends                                                                                         
or relationship between data-be it customer credit resume,                                                                                            
Foetal Phonocardiograms, travel expenses claims, airline                                                                                              
passenger trip or credit card spending data and act on those                                                                                          
trends or detected pattern automatically and rapidly.