The person that knew this has long left the company!

It is a typical situation in a company – a certain project needs rework, so the project team gathers the available data and analyses the available reports and interpretations. At a certain point, however, they realize that the people that had dealt with this project before and who had built up knowledge on the subject left the company some time ago.

Another very typical situation is when an expert announces that he or she wants to leave the company. This can pose a threat to the whole organization. Although it’s quite a common situation, there’s a lot to be taken into consideration. How important is this person to the company? Do they need to be replaced and if so who can replace them? How can the company make sure that knowledge remains within the organization and how will changes such as the above affect the organization and its performance?

Knowledge within an organization is based on its people. Their capabilities, expertise, experience, know-how and interconnection form a complex system, which cannot be transferred and stored easily. As a matter of fact, today, companies spend billions to maintain knowledge within their organization, having realized that this represents their biggest asset. But only four types of data can be stored and communicated, as there are numbers, texts, voice or sounds and images. Therefore, every organization has to develop strategies to prevent the loss of its knowledge and furthermore to enlarge it and make it accessible for the future.

Knowledge management – don’t turn a blind eye
According to the expert willing to leave the company, most of the time that company will try to make him hand over as much as possible of his work and knoweldge to the people assigned to the task – through proper documentation of his work, structured databases, etc. But, in practice, there is only limited time available for a smooth transition and it often becomes obvious that the quality and amount of transferable knowledge will not meet expectations.

Redundancy, in terms of more than one person at a time, creates the same issues – there is often an element of wishful thinking, considering that in most organizations the people represent the highest cost-factor. Consequently, there will be fewer and fewer people who have to knowledge, and who will therefore have to each learn more and more to keep the organization on top of business.
What is left over are a lot of files
Let us try a different approach to the problem of maintaining knowledge within an organisation by asking the question ‚What knowledge remains in the organization after the expert has left?’

The answer is mainly paper and files, as reports, interpretations, analyses, data bases, results and decisions. All these express the knowledge of this person and will be available even years after they have left. But try to reproduce any complex interpretations, for example, done by someone else and you are likely to quickly run into difficulties. Very often, it takes much more effort to try and re-interpret and understand what this person did than re-doing the whole interpretation from the beginning.

Redoing something from the beginning seems to offer some advantages, because in this process the organization will again accumulate knowledge, will be able to incorporate new data and will be more confident in defending its work – the organization as a whole will improve. The major drawback, however, is that a lot of time has to be spent to aquire that knowledge, knowledge that at some stage was in the organization. There is also still a high risk that this knowledge may get lost again sometime in the future.

Artificial intelligence – a science fiction approach?
Considering all this, we need to find a way to break the vicious circle. After all, we cannot yet store the whole human brain with all its capacity. But there appears to be some logic in the idea of transferring human intelligence into artifical intelligence.

We would probably consider this approach to be science fiction, and we wouldn’t be wrong. Technologically, there are still too many limitations to speed, capacity and strategy to be able to make the whole knowledge of one person accessible at reasonable costs.

It may be worth looking a little closer into the process of problem solving and its benefit for an organization: a) data or information represents the basis. B) Human know-how, experience, intuition, research, iteration, etc – then leads to c) a solution, in the form of a conclusion, interpretation, decision, report, etc.

Point a. and c. are still within the organization, and can be stored and made available easily, whereas point b. represents the human factor, the human capacity to solve problems in certain ways based on the experts knowledge. This is the part of the process that might get missing with time. This constellation, where there is input data (a.) and output data (c.) available, brings artificial intelligence into the game.

Let’s assume that the main goal of an organisation is to achieve good results. Let us further assume that a given input (information, data, etc.) results in a good output (good interpretation, good decision, etc.) performed by a good expert. And, finally, let us assume that this same expert with his knowledge would treat a similar problem in a similar way. So, all we got to do is find the relationship between the input and the output.

Neural computing – capturing the brain
This is exactly what neural computing can do! Neural computing is one part of artificial intelligence that offers the possibility to model the relationship between input and output based on training. Neural computing represents a simplified human software brain, with brain cells and the interconnecting synapses – which is able to learn, validate and store knowledge from training data. So, when an expert has left the outcome of his knowledge with the organization (in the form of reports, interpretations, decisions, etc.) it can come up with a storable model that relates the input with the output for a specific problem. The model does not exactly tell us how the expert did it, but it offers a way of reproducing his way of thinking.

The concept of neural networks was developed about 40 years ago but it is only today, with sufficient computer power and optimized algorithms, that it can be applied to these kinds of problems.

The main conditions for a successful application of neural computing to knowledge modeling are:
*Enough training data must be available.
*The formulation of the problems needs to be clear.
*The expert needs to be more or less consistant within his way of solving problems.

If these three points are fulfilled, a neural model can be derived for every expert and every given problem. This model can be used by others and can be applied with new and similar data by simply feeding the input data to the model.

This is an innovative approach to access human experts know-how, best applied and validated in cooperation with them while they’re still with the organisation. It may be worth considering this a strategic issue for organisations – all models can be saved within an experts database, which, if carefully maintained, will represent a powerful instrument to solve problems in the future.