Automation in the tax practice beyond the '90s.

This month's column was written by C. Eugene Prescott, CPA, Farley, Prescott, Mizell & Co., Inc., Greenville, N. C. Mr. Prescott is the former chairman of the AICPA Tax Division's Task Force on Automation of the Tax Practice of the '90s.

Automation in the

Tax Practice

Beyond the '90s

During the 18 months that the Tax Division's Task Force on Automation of the Tax Practice of the '90s spent developing a guide for the future, its members became acutely aware that even short-term predictions were risky in the technological arena. Nevertheless, a scenario was developed that the Task Force believes is reachable by the year 2000.

Central to the Task Force's predictions is the notion of a Personal Super-computer Network (PSCN) (developed in Chapter VII of the Automation Guide). It encompasses powerful processors and data capacity that have the capability to be connected and interact. Some processors will be on mobile PSCNs, and some will be on stationary PSCNs or docking stations. Multiple processors working in parallel is one of the keys to future computing.

There are several areas critical to the collapsing of the size and weight of the PSCN devices so that unencumbered mobility within the work area can occur, 1. Power source. 2. Data storage. 3. Physical method(s) of connectivity to local area networks (LANs), hosts, peripherals, etc.)

Similarly, critical areas of information will have to evolve for the physical technology to be useful: 1. Software: a. operating systems (intelligence layer); b. applications. 2. Data: a. embedded intelligence; b. ownership. 3. User interface.

Power

Until the mid-1980s, most American technology providers ignored this issue. Companies focused on the Negative Metal Oxide Semiconductor (NMOS) or bipolar, instead of the Complementary Metal Oxide Semiconductor (CMOS). Stated differently, they focused on speed and ease of development at the expense of power consumption. There was an apparent assumption that power consumption using NMOS or bipolar processes would be reduced over time. So although CMOS processes used only a fraction of the power, the original difficulty in fabrication and slowness in execution made the technology initially undesirable. Ultimately, making CMOS perform faster has proven to be more achievable than making NMOS and bipolar consume less power.

It is critical that PSCNs consume minuscule amounts of power. CMOS, or something better, will be required for rcal, sustained work to occur without being plugged into an outlet. Full mobility while working at the PSCN will allow the tax practitioner to reach new heights in polyphasic activity.

Storage

Traditional long-term storage media include disks and tapes. Random Access Memory (RAM) and Read Only Memory (ROM) have been used for temporary storage of rapidly changing data (RAM) and permanent storage of small amounts of data (ROM). RAM storage often holds data while a process is running, such as a spreadsheet or word-processing program. ROM storage usually contains instructions the computer needs to operate, such as parts of the operating system. In recent years, Compact Disk ROM (CD-ROM) or optical disks have emerged. CD-ROM is practical for large quantities of data that do not change, such as the contents of a book. Optical disks use light lasers as an alternative to magnetic fields to distinguish whether a switch is on or off. They currently place much more data per inch of surface than magnetic media but with significantly slower access time. The PSCN will evolve in at least two ways as it relates to access to data: * Large, inexpensive, low-power RAM. * Noncabled connectivity.

While the methods differ, the results are the same; the user can be mobile while at work on a PSCN. As several gigabytes of RAM become available on a single chip that stays on (by battery, solar power, etc.) while the PSCN station is off, traditional ways of achieving long-term storage of data will change. While speed of access to data stored on magnetic or optical media will continue to improve, it is not likely that it can overtake the speed of access of data stored in ROM or RAM. So if large amounts of RAM/ROM data can be maintained indefinitely without significant power consumption in a chip that can be housed conveniently and unobtrusively, full mobility while performing computer-assisted work can occur. Why is it necessary to have several gigabytes (or the equivalent) of data immediately accessible to every tax practitioner? There are many reasons, but the focus here will only be on a universal one. Information currently on paper in files, libraries and at remote locations will be available instantly on PSCNs. Client files (including handwritten notes) will be stored digitally, and will be displayable on screens or printable to printers immediately. Responding to voice command, the PSCN will find the information needed (whether it comes from a client's files, a tax library or a client's office) and present it in the manner requested. The issue is not how many gigabytes are accessible, but how many will be available at a cost of storage that 50 million potential users can afford.

Physical connectivity

PSCNs need to be connected in such a manner that huge amounts of data can be rapidly moved between devices, locations, etc. Band width, which limits the amount of data that can be moved, will need to expand to that provided by fiber optics for effective use of the evolving technologies. A two-hour movie that is stored as bits (or in some other manner that can be converted to bits) would consume 151 gigabytes of (noncompressed) magnetic disk space. Moving that much data between two remote sites within a cost-effective time period will require extraordinary band widths through all of the mediums of transport. Depending on the distance, these are apt to include satellite transmission, microwave transmission, fiber-optic cabling and radio signals. Within a building and between physically close computers (such as a LAN), fiber optics, microwave, infrared light or radio transmission are likely to be involved; fiber optics may also be involved in longdistance hauls. Devices that do not need to move (perhaps docking stations for devices that do need to move) will likely be connected with fiber-optic cable. The most portable PSCN will likely be connected with the other PSCN (docking station) wirelessly, allowing the necessary mobility.

For this technology to be useful, one will have to be able to give the computer instructions regardless of how or to what it is connected. Traditionally, the keyboard and mouse have been used for these kinds of instructions; again, however, technology must keep pace. To gain full mobility, the PSCN will need to deal with the user's conversational voice. The PSCN will need to adapt to the user more than the user adapting to it; voice communications that require code-word phrases and abnormal tempo will not be universally accepted in the marketplace. However, early implementations will likely be forced to use these techniques.

Clearly, ff one can be wirelessly connected among the PSCNs one can also be wirelessly connected to printers, disk drives, etc., through docking stations or other means. Although many of the specific components to implement the PSCN from a hardware standpoint have yet to be invented or perfected, that is not particularly worrisome. Developing the software required to work in a seamless environment poses greater problems. It will be of little value to have diverse systems connected if the information that is on the systems cannot be seamlessly used.

Software

Operating systems: To date too much effort has been, devoted to trying to create and market a single operating system that is so much better than all of the others that the world (i.e., application software developers and end-users) will migrate to it. Most operating systems perform some of the same functions. Those functions allow information stored on the computers as application programs to create, change, add, delete, etc., information stored on the computer as data. A problem with making an operating system so robust that it can do everything anybody needs is that the core of the operating system, the kernel, then becomes very large and complicated. A problem with making an operating system too task-specific is that it may not be able to evolve as things change.

With the current progress in Artificial Intelligence (at least in expert systems), it may be possible to take a different approach that does not require a significant change on anyone's part; i.e., most people could continue to use their current programs and data while starting to use different programs and data on different, connected platforms. For this concept to work there would have to be intelligent code between the user's kernel and the disparate system. Intelligent input/output(1) could convert data from its current format to the format of the host, change it and convert it back. Conversions would occur only as the need arose, and the user's awareness that a conversion took place would be limited to the additional time required. (With the increase in speeds from the hardware side, this time may be negligible.) Similarly, intelligence could also be implemented in the data-storage scheme. Programs and files with infrequent access would automatically be archived to different media than those used by the main system. For instance, if primary storage is in RAM, these files could systematically be moved to optical disk, leaving RAM storage for current files. When a request for an archived file (program or data) is invoked more than a predetermined number of times in a predefined time period, the file would automatically be relocated back to RAM storage. With these types of intelligent programs sitting between the kernels and the next layer of a system's activity, life can go on while global connectivity becomes practical. If the operating system's kernels can keep their currently most desirable attributes while hooking into the next layer, current users' competence with their current systems can be leveraged. In addition, the best application products, regardless of the platform they are currently on, could be accessed by everyone who is connected. The potential productivity gains from these factors should be enough incentive for code writers to write code. This technique, of course, would extend the Iffe of all of the current operating systems (those that have the intelligence layer added), which would cause the current developers of operating systems to be relegated to long-term coexistence. This, in itself, would require a commitment by operating system developers to refocus.

Applications: There is a paradox in application software development. The best, most elegant products have focused on single (or a few) tasks. Products that have attempted to combine or integrate a broader base of tasks have not succeeded as well in the marketplace. (Spreadsheets, such as Lotus 1-2-3 [c], and word processing, such as WordPerfect [c], are examples of the former, while Lotus Symphony [c] is an example of the latter.) It seems that the best functional ideas across a broad spectrum of tasks cannot be brought to the end-user marketplace by a small group; i.e., the brightest spreadsheet mind may not be the brightest word-processing mind (or vice versa).

Perhaps the focus should shift. Rather than rolling an of the application programs into a single product, why not integrate them on the fly? If the new intelligence layer between the kernel and the applications programs and data can be aware enough to convert information as needed, this converted information can be merged into integrated programs and databases. The obvious advantage is that no one has to preordain what facets of operations should be joined. Endusers can determine their own levels of integration. A tax practitioner could combine word-processor, spreadsheet, accounting-data, tax-planning, tax-research and tax-preparation software into what appears to be a single program. The same practitioner could have different levels or degrees of integration for different clients or purposes. The only degree of conformity required of the creators of the various products would be to Write to a common standard. This standard would define the elements required for the on-the-fly integration. The vendor of a single-task product would have to certify only that the product meets the standard. The ability to integrate would make all of the best minds in application-software development available to the end-user, thereby providing greater flexibility.

Data: If intelligence can be interjected at any level of the technology process, it can be interjected at every level. If hardware can "know" what it is and operating systems can "be aware" of disparate connected systems and applications can "discern" each other, data can "recognize" its own characteristics. With an embedded propensity to understand what it represents, the concept of data may change. "Smart data" may be able to behave hke data being manipulated by a "smart program." How this data is organized may differ from the traditional methods of data organization.(2) Hierarchical file systems have traditionally provided the pathways to data. Records, documents, programs and other user-specific data may be incorporated in a user-specific database. Conceptually, the PSCN (which knows each user's identity) could then locate each user's data regardless of where it is situated.

In addition, the amount of data at the user's potential disposal will increase rapidly. The ownership of this data will become a perplexing issue. If data is from a public source but is organized in a special way, does the organizer own any aspect of the underlying data? Ownership would imply the right to charge a user fee to whomever uses the information. Ownership would also indicate the right to prohibit some types of use. Even as the issues of ownership and control are resolved, how to account (and bill) for this use will also become an issue. If all information workers have access to all information concurrently, should they be billed for the perpetual right of access? if information is an owned product and others access that internally developed information, do they need to be preapproved and prepay? If data has value, some method of compensation for the use and exchange of the data has to be devised. If each user has to subscribe to hundreds of data repositories, the administrative logjam may be counterproductive. Beyond the concept of compensation is the issue of confidentiality. Restricted-use data will need to be secure from unauthorized use or tampering. A series of well-designed, cryptic, unique passwords for each repository may become a little much. ff the PSCN can be aware of itself and its peers, it can also be aware of the user. Voice prints or retina patterns are likely to supplant other security identification techniques. However, if the "matching" pattern is digitally stored, access to this matching information will remain a security issue.

The user interface issue

One of the primary potential benefits of a Graphical User (GUI) is that applications will become more standardized as they relate to common housekeeping tasks; i.e., saving a file in one application will be done in the same exact way as saving a file in another. The notion is that the GUI and operating system are actually doing the saving and the application merely has a hook to the GUI's technology, which has a hook to the operating system's technology. The trouble is that there are about a dozen GUIS on the market (and in the field) today; while they have more similarities than differences from a user's perspective, they are still different enough to limit the potential benefit. With so many different GUIs on so many different platforms, there is still a learning curve for the same person to work on different systems. Additionally, application software developers are currently having to write for multiple GUIs or limit their product to a single GUI. So while GUIs are supposed to make using computers easier and more efficient (and they do), they are currently also adding yet another layer of complexity to the overall process. Most tax practices use a variety of application-software and utility-software products. If some of the software providers upgrade their product to GUI A and others to GUI B, does the user install two GUIs or perhaps run a GUI under a GUI? Of course, each GUI provider has a simple answer to this dilemma: just choose his offering and convert to whatever software is compatible. Multiple GUIs that are similar but different are causing confusion and in many cases result in a reluctance by the user to adopt any. Ironically, there are more possible GUI choices under the various flavors of UNIX [c] than under any other single operating system. While Windows 3.1 [c] (and its successors) is likely to be the standard GUI for systems based on MS-DOS [c], a UNIX[c]-like operating system is apt to be at the hub of many future systems. So GUI compatibility seems to be an issue. Therefore, a system that can seamlessly access programs and data on disparate systems may be the GUI foundation for many future systems. Supporting the notion that a UNIX [c]-like system will be an integral part of broadly connected (networked) systems is the fact that most systems vendors support a version of UNIX [c] (in addition to whatever else they support).

While the numerous variations (and inconsistencies) are still barriers, there is a thread of conceptual uniformity and adaptability with UNIX [c] that does not yet exist with other operating systems. Perhaps the GUI uniformity solution is voice communications. If the PSCN is responding to the user's conversational voice rather than to the pointing and clicking of a mouse, GUI representations will diminish in importance. Clearly, data will have to be represented on a screen in some manner; however, any inconsistencies in representation will be of less concern if the PSCN is reacting to the user's voice in a perceptive manner. The best lessons to be learned from GUIs may be that they are conceptually designed somewhat like the intelligence layer of software that will come to reside between the kernel and the application and data. By the year 2000, GUIs will be a nonissue.

Conclusion

The term that describes how disparate systems will work in harmony while maintaining their unique characteristics is interoperability. While interoperability already means different things to different people, it nevertheless is a key to the future. Powerful processors working in parallel over various methods of connectivity are going to require an extraordinary level of cooperation among currently competing technologies. By the year 2000, the providers will have cooperated or died.

(1) Broner and Powell, "Intelligent I/O Rule-Based Input/Output Processing for Operating Systems," Operating Systems Review (ACM Press, New York, July 1991). (2) Gagliardi, "The Past, Present, and Possible Future of Office Automation," UNIX Software Journal (FourGen Software, Inc., Edmonds, Washington, Vol. 4, Issue 2, 1991).