Perform as smart as you are.

There is a widespread feeling among managers that their businesses are not performing up to their full potential. A common lament is that we are so much smarter than we routinely perform.

But why? To understand why, we need to shift from a traditional business point of view, such as product flow and dollar flow, and come to view our products and services as the documentation of decisions made by our employees.

So decision making is key. But decisions are dependent on not only information but on the knowledge of what to do with the information. I believe we underperform in business because we seldom have our best corporate knowledge available at the point of decision. We have invested 20 years of effort and expense in getting more information in the hands of the decision makers, but our methods of capturing, storing, distributing, and applying knowledge remains essentially unchanged. We still hire good people, train them in our procedures, and then put them on the firing line. This process of transfering knowledge is beset with shortcomings. It is here that knowledge-based systems can have a significant impact.

How much of an impact? I have found that you can get a 1,500-percent return on your investment in knowledge-based systems. That's a $15 return on every dollar you invest.

What's in a decision?

Three things are necessary to make a decision: relevant information at the point of decision; the knowledge of how to analyze that information at the point of decision; and enough time to do the analysis.

In today's corporations, oceans of data drown most decision makers. Eliminating irrelevant information requires the knowledge of what is relevant, the knowledge of how to access and select appropriate data, and the knowledge of how best to prepare the data by sorting and summarizing it to facilitate analysis. This is the raw material of decision making.

A knowledge-based system, which is nothing more magical than a computer program for capturing, storing, distributing, and applying knowledge at the points of decision, has only three parts: the user interface, a search mechanism, and a rule base. Of these three, the most critical is the rule base. The systems that are now on the marketplace have the user interface and the search mechanism, but it's up to the user to capture and store corporate knowledge in the rules.

The unit of storage in a knowledge-based system is an if/then rule, similar to a record in a date base. Most shells of knowledge-based systems today are general purpose and so supply no rules, although there are a few that are targeted at specific markets--for example, insurance underwriting or tax planning, which do contain some domain-specific rules.

There are four broad classes of rules a user typically would input in varying amounts, depending on the problem to be solved:

* Rules of physics (IF there is an action, THEN there is an equal and opposite reaction);

* Laws of government (IF you violate an IRS regulation, THEN you are subject to penalty);

* Corporate policy (IF this is a new operation, THEN a non-infringement ruling from the legal department is required prior to commercialization);

* Rules of experience (IF a sample is grossly out of range, THEN resample and test again).

The point is to apply all of these rules routinely at the point of decision and to have a mechanism for getting them there at the moment of truth. This is what a knowledge-based system does in spades.

When knowledge-based systems

should be used

Knowledge-based systems (some refer to them as expert systems) should be used any place and any time imperfect decisions are being made. The benefits of these systems are that they provide timely, adaptable compliance with regulations, consistent enforcement of corporate policy, and more thorough analysis of time-pressured decisions. Through these systems you can apply your best knowledge to your corporate operations--in assigning credit risk, for example, or determining sales use tax. For a big corporation such as du Pont, getting that last item right can make a difference of $4 million each year.

Expert systems also provide the best corporate knowledge of control logic to operations 24 hours a day. They plan and schedule operations with lower inventory, increased capacity, and lower cost. They bring the corporation's cumulative knowledge about how to extract the functionality from products, like dyeing a carpet or driving a car, which enhances the product, builds strong ties with the customer, and gives you significant competitive advantage. And, finally, they can provide your representatives in the field with a cumulative product knowledge at every sales call, helping the representative choose the best product to recommend to a customer when several are available.

How to make an expert system

Developing an expert system requires three elements: the skills, the tools, and the method.

* The skills are teachable and transferable. In fact, for most systems, a few days of training will be enough to get the necessary skills. (There are a few, however, that require more extensive training.)

* The tools are existing hardware and inexpensive software.

* The method is the important thing. It is in the method, the design, that expert systems challenge the corporation. The important thing is to design a system by successive approximation. Don't put all your energy into getting the design right from the start. Start building it and have it interact with the people who will use it until you get it right. At du Pont, it's not uncommon to have three versions a day as we're building something.

It is also important to have incremental deliverables. You don't have to have it complete to have it be valuable. In fact, most of the more than 50 planning and scheduling systems we've written at du Pont have more than paid for themselves by the time they were declared to be in service because they were delivered in increments and each increment was put into service at the time it was released.

No inventions required

The opportunities for expert systems are pervasive. They can be applied almost anywhere in the organization. The returns are compelling. And no inventions are required. It's simply a matter of will and focus. In today's world, I see over and over again, in my own organization and in others, that we are not able to execute as smart as we are. Expert systems go a long way in turning that around.