Changing Rules in Decision Support Systems Automatically

Changing Rules in Decision Support Systems Automatically

By Paul Haley

Your company has made major improvements to a core product that should catapult it far ahead of the competition. Now it抯 time to update the customer support software. The IT department tells you it doesn抰 have the manpower to make all the changes and test the software in time for the product release, and there抯 no budget for outsourcing.

What if you could get all the changes made without the time and expense of reprogramming and troubleshooting the software?

New technology enabling non-technical business people to make changes to automated decision support systems is now commercially available. Knowledge management enhanced with natural language processing technology now enables customer relationship management (CRM) professionals to both create and modify the logic that supports their decision making ? without programmer intervention.

A company derives its business knowledge from its collective experience and its policies and procedures. Customer service representatives put business knowledge to work everyday when they decide:

• What questions to ask to help a customer solve a problem.
• What to tell a customer with a complaint about a purchase.
• What products to cross-sell to a customer.
• Whether a customer qualifies for a discount or a waiver of finance charges.

Automating decision making boosts CRM productivity and increases customer satisfaction by reducing errors, improving consistency and speeding turnaround. For example:

• Sales efforts can be personalized to consumers? individual needs and buying habits.
• A Web site can automatically ask customers pertinent questions based on answers already provided and act on the responses.
• The rules that determine how information is processed can be changed without the intervention of the IT programming staff.
• Functions such as screening credit applications or offering frequent buyer rewards can be automated and readily modified.
• Troubleshooters at help desks can quickly access prior references and other knowledge they need to respond to difficult questions.

A serious roadblock to automated decision making has been the difficulty of translating business knowledge expressed in human language into rules that a computer can use to support decisions. The fact that knowledge continually changes as the business environment evolves further aggravates the problem.

Procedural computer programming has at its tactical objective to express human knowledge as a series of on/off switches. Procedural programming has forced IT to turn business into processes, i.e., flowcharts, at each branch of which there are only two choices: yes or no, left or right. Today virtually all decision support software reduces decision making to a series of yes/no alternatives that programmers have to encode.

Because decision making often hinges on several variables, the flowchart of even simple business decisions becomes large and convoluted. And because a change at one point in the flowchart ripples through all downstream decision points, the flowchart will be hard to change without a considerable outlay of programming resources to recode, retest and release the software.

By contrast, rule-based programming greatly simplifies the process of encoding knowledge needed to automate decision making. Rules-based programming begins with translating business knowledge into short and simple statements of two types:

• Declarative sentences of fact, and
• Imperative sentences based on a condition.

Take the following simple sentences:

• People living in the 90210 ZIP code have a high net worth.
• People with a high net worth are good prospects for top-of-the-line products.
• Offer high-end products to Web site visitors living in 90210.

With knowledge-based programming, a computer fed these statements could generate a decision to prepare a personalized catalog of luxury products and e- mail regular offerings to qualified customers. If a company feeds a series of similar declarative and imperative statements into the computer, it can essentially automate much of its CRM operations.

That does not mean that representing a business model using knowledge-based technology eliminates all other IT. IT still implements and maintains the corporate information structure, including its relational databases, object- oriented programs, and legacy and enterprise application integration. It defines the initial nouns, verbs and other parts of speech that business people use in their statements. IT is, however, relieved of the onerous task of threading dynamic business logic throughout the infrastructure. Consequently, the infrastructure becomes more stable and reliable, and IT becomes more productive in advancing its architecture and functionality.

Although still in its infancy, significant advancements in natural language processing ?the analysis and breakdown of natural language statements into functional units that can be converted into formal logic ?now make it possible to enter business rules into a knowledge management system in normal spoken or written sentences.

As a result, once the system is in place, customer service professionals can input their business knowledge, policies and practices, and the system will automatically make the changes necessary for the knowledge to work automatically. That means managers can change the knowledge that supports CRM decisions at their desktop without having to ask the IT department to launch a project. All managers have to do is add or change a sentence by dragging their mouse or dictating it to a speech recognition package.

In either case, inputting these statements into a computer system becomes routine. It doesn抰 require creating a customer support model that accommodates every aspect of a decision and then coding it for the computer. Instead of reducing all CRM knowledge to yes/no choices, natural language knowledge enables people with firsthand knowledge of a company抯 products and processes to manage that knowledge in plain English ?without programming. The people who actually run the operation can automate complicated business decisions and thereby lower their cost of doing business dramatically It makes changing the rules as easy as Simon says.

Paul Haley is the president and founder of The Haley Enterprise which makes AI-based software that enables companies to bypass time-consuming, expensive programming and modify the software that automates or supports business decisions by writing or dictating new decision rules in plain English. Haley抯 products accelerate and simplify business application development and deployment. Haley can be reached at paul@haley.com.