By: David J. Wardell

© 1988 By: David J. Wardell.  Reproduction or redistribution in any form without written permission is strictly prohibited.

These latest efforts to automate pieces of the travel business are usually stand-alone software packages (not part of a CRS, but possibly linked to one), that make decisions as to fares, routings, flight choices or traveler profiles in relation to the itinerary booked, and either report their findings or correct the passenger’s reservation.

How well do they do their job, and why would somebody use one?

While not quite the rage in travel automation, there are several products of this type in production or development.

They are essentially "electronic travel agents," in that they address all or part of the functions your agents perform and apply enough of a systematic process so that the computer can attempt some of the same decisions.

There are some fairly cosmic implications here as to whether or not computers can successfully make the types of decisions that an agent makes,

The premise for most computerization is simplifying work, if you can automata your accounting and have statements and reports prepared faster, with less effort, and more accurately than is possible by hand, then automation has been successful.

Most people don’t understand, however, that the next fundamental premise should be that "the best tool is no tool.

Rather than developing yet another system (where effort is expended and errors are possible), if the task at hand can be reduced so that no tool (human or computer-applied), is required, then all types of development and operation costs are avoided.

Computerized travel management systems, trip planners and fare-checkers go beyond simplifying systems into actually making decisions that translate into actions.

Some developers have drawn the line between tools for the decision process and true makers of the decisions quite finely indeed, so that it’s hard to determine exactly where the computer’s role begins and ends.

This is an area known as machine intelligence (or artificial intelligence). It seeks to use computers to supplement and ultimately replace human decisions for specific applications.

It includes various exotic computer disciplines such as rule-based programming, logic-based programming and so-called "expert systems."

The travel business has its fair share of these systems.

Travelmation, CompuCheck and Air Planner fall into this category. To successfully evaluate how well these, and other, systems are suited to their purpose, you should understand what, in a general sense, they’re trying to do.

First, all machine intelligence-based travel systems accept the premise that the non-sales activities performed by an agent are functions of technique, and not creativity.

Once you understand the chain of decisions that underlie the technique, you can transfer these to a computer fairly effectively, you can’t (yet) do that where a creative process (one involving flashes of insight or new ideas) is involved.

It all sounds pretty simple--until you try it. Then, you quickly discover how difficult it really is to even understanding the decisions that comprise planning and evaluating an itinerary.

Even with the most basic tasks, humans make literally hundreds of small, and often unspoken, decisions and evaluations in a fraction of a second, based on training, experiences and memories. All this information, and the process for applying it, has to be codified and transferred to a computer if it’s to do the same job.

This is why some computer programs are defined to in-dude only parts of the travel management and planning~ process--it’s just too difficult for some developers (and too time-consuming for the ma-chine) to address more than a few tasks,

Further, if the programmer didn’t think of every contingency, or made some wrong I choices, the program would have to be redone.

One possible shortcut for this problem is rule or logic-based programming. Rather than attempt to anticipate every decision and response in the chain, the programmer can define criteria (rules) upon which decisions are made, thereby reducing the bulk of the program and the time required to run it

Because this is new and not well-understood technology, people to write these programs are hard to find. There are almost no truly rule-based systems marketed for travel applications.

The second big problem, once the process is dealt with, is accessing a database against which to make the decisions. If the system is applying traveler profiles to Itineraries already booked (or being researched), then those profiles can be organized as the programmer wishes.

If fare research is involved, however, the developer must deal with numerous potential databases. Building your own, proprietary, database is difficult and expensive, as is maintaining the database once it’s built.

Therefore, most developers use one that’s already available, despite its imperfections, and also frequently search that database by using the process or function already set up by its builders.

So are these systems commercially practical? Most of them are not.

It’s so hard to build the type of system that could really do the job that most products either compromise performance to unacceptable levels (limiting productivity), or draw such a small box when defining the product’s features that it doesn’t meet very many real needs.

Some software is so fragile and based upon such contrived premises that supporting the tool (rather than minimizing it) becomes the focus of more activity than is involved in the real work of handling customers and making sales.

Most travel management and planning software products are attractive more as marketing showpieces than as productivity or accuracy enhancements that truly simplify workflow.

Beyond some specific applications, this is a promising, developing technology for the travel industry, but not yet a proven commodity ready for widespread acceptance. My next column will deal with several of the specific products available and how they can be used.