ArtsAutosBooksBusinessEducationEntertainmentFamilyFashionFoodGamesGenderHealthHolidaysHomeHubPagesPersonal FinancePetsPoliticsReligionSportsTechnologyTravel
  • »
  • Business and Employment»
  • Business Management & Leadership

tools for business management

Updated on October 7, 2016

Artificial intelligence (AI) has gained a reputation, earned or not. To say it mimics “intelligence” is rather redundant. What do we mean by intelligence? We mean the power to reason, to come to conclusions, recognize patterns or acquire knowledge without it being pre-programmed. There’s the challenge.

Let’s take the easiest form of AI – acquiring additional knowledge. We are back to the WWW spiders that track every httprotocol they find, look for words which are used more than once on a web page and register that page in a database with the repeated word as an index. A search engine then uses this database for searches, tied to the indexed word. Of course, for this to work, there is also a table of ‘ignore’ words, such as “the”, “and”, “we” and “you”.

Another type of acquisition – let’s say that MFAP was the only DSS (decision support system) for a dairy. MFAP could search the network looking at all stored files for ones containing the text “cheese”. With this methodology, MFAP could widen its knowledge base, altering the decision support results.

Reasoning is a little tougher, perhaps. But not if we put it in the if…then context: If there are more than 150 people in a department, then we need to review the organization chart. For those of you who do programming, if/then/else statements, case statements, when and while statements are all examples of reasoning.

If you add a consequence to the reasoning statements, you come to a conclusion.

Have I lost you yet?

Now … pattern recognition. That’s a doozy. If you have ever taken calculus, you’ve seen questions where they take a series of numbers and you have to figure out the next number. Sometimes these can be quite daunting. But a computer can do it by testing all sorts of scenarios, so eventually a computer would figure out the pattern, by trial and error. To program it, one would tell it to first take the difference of each pair. You would visualize this quite automatically if the numbers were small enough, but the computer isn’t so bright. If that didn’t work, the computer may be told to factor each one and compare the factors. A whole series of possible ways numbers could be manipulated would be set up in a rules set for the computer to try.

Ah, but that is something only IBM would bother to do. In real life, when we are looking at sales and productivity, the numbers are not neatly broken down. So how is the computer going to recognize a pattern of loss or gain? It would do it somewhat like a sort program. First it would determine the difference in sales between January and February and store that number, perhaps in an array. Then it would calculate the difference between February and March and store that number, and so on throughout the year. The result is a series of 11 numbers. Now it should look at those numbers and analyze them. If two or fewer numbers are negative, they should be dropped from the calculation. Conversely, if two or fewer numbers are positive they should be dropped. What this does (and it’s a normal practice in statistical analysis) is compensate for ‘odd months’ or anomalies. Then the numbers are plotted on an x-y graph for a visual representation of the trend. Numerically, the differences are calculated on a bell curve, to find the mean and median.

This method can be applied to hiring rates, raises, sales, product production, anything that incorporates a rate.

Expert systems usually include a bit of artificial intelligence, so that not only do they know “everything there is to know” about a particular industry or system or organization, but they can make recommendations based on this knowledge. You wouldn’t pay a Subject Matter Expert (SME) if all he could do is regurgitate stats – you would want him to be able to advise you on ways you should manage your business.

Let’s take an example – you have an IRA and you want to know how to re-invest the funds. An expert system could look at the performance of the funds and stocks you currently own. It would then compare those to such agencies as the Dow (after all, everyone lost money in 2000, but only certain types of stocks dropped in October 2001). It would then poll you on your preferences – are you the ‘safe and secure’ type? Or do you want fast growth? How long will you allow the stocks to stay in the funds before cashing them out? Do you have extra cash for front-loaders? Will you stay in one fund long enough to go for a back-loader? The expert system would then look at possible recommendations to see not only if they fit your needs, but if they performed better than the ones you already have. The result – a recommendation for your IRA investments.

I haven’t seen a “shell” to which some texts refer, but I can certainly understand how they could be built. As long as there is room to plug in the product that you manufacture, and/or your goals, etc., the process is the same, no matter what it is you manufacture. It’s akin to writing a system up in pseudo code – then you can code it in Pascal, Fortran, Visual Basic, whatever. However, I suspect that such a generic application would not offer suggestions which really are applicable to a particular user.

Natural language processors…well, that’s pretty darn impressive if you ask me. “Ask Jeeves” works on a very simple premise – it screens out unnecessary words and searches on the words that are left as individual indexes. I think it also does a Boolean “AND”. If you try to figure out how to program a computer to understand natural language, it’s quite a challenge. Languages such as LISP and ADA help. Then to top it off with the ability to handle spelling errors…wow. There’s a big difference between them saying “lights” or “open” on Babylon Five and saying “Computer, show me all possible languages’ interpretations of this writing”, like Star Trek TNG. The industry has a long way to go…

An example of a reasoning demon – select all possible locations meeting the requirements, add up the costs of each location. If the costs reach 1.5 million dollars, mark the location as negative and stop adding up costs for it. These are used in MISs (management information systems) and DSSs (decision support systems) if needed by the user.

Slots and frames…these are sometimes called classes and subclasses. An example – “dog” is a frame. Its slots include 4 legs, fur, tail. Another slot may be “breed”, which would have its own slots. The breed of Dalmatian would have slots for spots, 50-70 pounds, white, black. If an object has the characteristics of 4 legs, black, white, fur, spots, 65 pounds, and a tail, it would fit the Dalmatian dog frame. The subslots for chow-chow would have black tongue and mane. This type of logic is used for natural languages and other forms of artificial intelligence.

If you look at an organization as a whole, you will more than likely find MIS, DSSs, Expert Systems and a bunch of combinations thereof. Unfortunately, organizations tend to think within their respective budgets, which results in a great deal of redundancy. A system analyst would have a ball coordinating them.

An example of groupware – I used to work for a company called Vanstar. They issued laptops to each consultant with Lotus Notes on them and configured for the individual consultant. With these, the consultants entered their timesheets, sent and received e-mail, generated reports, and received ‘broadcast’ messages from corporate headquarters and the branch. Since they met only once a month in person per branch, this was the virtual ‘family’ within which the consultants worked.

At Bristol-Myers Squibb we regularly had teleconferencing wherein we used computers for all visual aids – slide shows and graphs, etc. We also would send in notations that we had questions for the speaker. The switching between New Jersey display, Connecticut display, and Europe display was all run by computer.

English teachers get all fired up about the collaborative authoring systems, but they have a practical use as well – reports which are a combination of people’s input. At General Dynamics, every Friday, each person wrote up a ‘sig events’ report (significant events) of the work s/he did that week, milestones achieved, and plans for the following week. The supervisor melded them together in another report; the department manager melded all the direct reports’ reports and generated his own report, and so on. Bristol-Myers does the same thing on a monthly frequency all the way up to the divisional level, where they were published in e-mail for all other divisions to read. With collaborative authoring, this could all be one document, to be edited by the successive layers of responsibility.

Rational Rose is a very good example of multiuser technology. Several developers can work on the same application. Each develops his/her own modules and puts them into a library/workspace, where they can be swapped, combined, etc. We used it at BMS so developers could actually be working in different states (or countries) and still get the work done without the cost of rental cars or hotel rooms.

So-called Executive Information Systems. This is a unique term; usually the people don’t categorize such systems since they are built custom for each executive (by request of the executive). So usually they are known by their own unique name such as BINGO (bytestream Information for Negotiation within the Governmental Organization). At best, they are referred to as super-MIS systems.

We have all seen drill-down investigation in action on TV and in the movies – an investigator gets a photograph, and drills down to a detail like finding the killer’s image in the reflection of the victim’s eyes. OK, so they take it a bit far… But the concept is the same. For instance, I may look at a graph of my manure spreader production, and I want to know why the dip in January. So I can drill down to the factors in January, and find the temperatures were below zero in my biggest market (Wyoming), so manure spreaders were of little use with frozen manure!

I once worked on a series of daily reports for Pfizer. The information was taken from an ACD system (tracks phone calls) and an Oracle database which was in use 12 hours a day. I generated the reports in MS Access. The ACD information was converted to Access by its own data keeper every morning at 5 AM. To avoid slowing down the Oracle database performance, I wrote macros to download the information from Oracle each night at 8 PM into Access tables. As a result, the reports generated each morning were based on the ACD tables and the Oracle tables, and were always at worst 12 hours out of date, and it was in those 12 hours that no business was being conducted anyway.

The limitations of Executive Information Systems is not in their power – they are virtually omnipotent because of the customization and constant upkeep – but rather in their impact: executives relying too much on them, too-fast reactions to reports that destabilize business, and so on.

One of the things that we've discovered is that color is more than a 'prettying' of the interface -- red exit buttons, yellow help menus, colored Excel lines to keep the eye on the results -- all of these make the DSS (or any other application, for that matter) easier to read, use, and understand.

While the decisions and formats may be different, the process is always the same:

  1. Interview the prospective user. Find out exactly what decision the user wants to make, what his/her value system is, and what information s/he wants to base the decision on.
  2. Collect the data for the knowledge base – what format would it be available in; how would you acquire it and convert it into usable data for your DSS or MIS; avoid manual input; what information is date-sensitive and how can you get it regularly updated; what information would be useful to give the user that is not really of importance in evaluating the data
  3. Determine the processes to be performed – how to evaluate and rank the information to cull out the choices that meet the user’s needs
  4. Presentation – how to present the choices and peripheral information so that the user can make a decision.

Comments

    0 of 8192 characters used
    Post Comment

    No comments yet.

    Click to Rate This Article