THIS IS A DRAFT OF MY ARTICLE ON AGENDA, IN THE MAY 1992 ISSUE OF THE ATLANTIC. THIS DRAFT IS 95 PER CENT SIMILAR TO WHAT WAS FINALLY PUBLISHED IN THE MAGAZINE. THE ONLY DIFFERENCES ARE SOME LAST MINUTE COPY CHANGES WE MADE, PLUS THE PHONE NUMBER OF EYE-ON ASSOCIATES. THANKS FOR YOUR INTEREST.

JIM FALLOWS

Lotus Development Corporation released the first version of a program called Agenda in 1988. When it appeared, Esther Dyson, publisher of an influential computer newsletter, gave it an ecstatic review. The techniques of "artificial intelligence," which were used for chess-playing computers and systems to simulate a doctor's diagnostic skills, had been gestating on large-scale computers for years. With Agenda, Dyson said, they would be available to the ordinary personal-computer masses for the first time.

Agenda was created by Mitchell Kapor, the founder of Lotus and at the time its chairman, and two other men: S. Jerrold Kaplan, who had been head of his own artificial intelligence firm, and Edward Belove, a senior Lotus official. After two years of refinement, the Agenda was ready -- but by that point, Kapor had decided to leave the company. (Kaplan also left, to develop "pen-based" computers.) Since then, Lotus has often acted as it were not quite sure what it has on it hands.

Agenda 1.0, the original version, was in a sense too powerful for mortal men. Artificial intelligence systems involve routines called "inference engines," which are sets of rules for assessing data. For chess-playing systems, these rules test the power and desirability of various moves. In medical systems, they fit a list of symptoms against patterns of known disease syndromes. Devising these rules is of course the hard part of the job. This is why there are not yet any satisfactory computerized language-translators or even grammar-checkers. Real human languages involve so many quirks and special cases that no one has reduced them, yet, to a manageable set of rules.

Agenda 1.0 provided the engines for assessing a wide variety of rules, but it did not give the user much help in understanding how the rules should be drawn. When you turned it on, it offered a blank screen. Word-processing programs generally offer blank screens, but most people can connect that with the concept of a blank piece of paper. Agenda's first screen didn't connect to any familiar image. The early word of mouth about Agenda was that it was "hard to learn." This complaint missed the point. The program was easy to operate -- once you understood what it was supposed to do.

In 1990, Lotus released Agenda 2.0, which included several sample systems to perform familiar tasks, like scheduling appointments or managing to-do lists. Lotus advertised Agenda for a while, usually presenting it as a "Personal Information Manager," or PIM. This is a grab-bag category of software, which includes calendar programs, Rolodex-like systems for phone numbers and addresses, "ticklers" to remind you of obligations, and other programs that are generally more straightforward, and much easier to explain, than Agenda. About a year ago, Lotus stopped advertising Agenda altogether and it has not released another version. Although the program is still for sale, and Lotus offers active technical support to Agenda users, the company barely mentions Agenda's existence in public, not even in the big advertising campaigns that present Lotus's across-the-board software array.

I hope Lotus's silence means only that the company is puzzled about how to present the program most effectively, or is secretly working on a new release -- not that it has decided on euthanasia. Esther Dyson was right; Agenda really is something special. The evolutionary struggle in the computer world has only a rough connection to real merit. Everyone who has watched the business has a story of a wonderful program that was overlooked or orphaned, or of a mediocre design that somehow became the standard.

Lotus presumably has enough muscle to keep Agenda program from dis- appearing. I hope it decides to put its influence to work. Of all the computer programs I have tried, Agenda is far and away the most interesting, and is one of the two or three most valuable. It cannot replace the truly indispensable work-horses of the computer age -- the word-processing and telecommunications programs. But the more I have used it over the last year and a half, the more I have come to rely on it and to admire the wit that went into its design.

Now I face the problem that has apparently buffaloed Lotus's advertising staff: describing just what Agenda does. The possibilities are easy to demonstrate on a computer screen but a little trickier to explain.

To start with the big picture, Agenda has three operating components: items, categories, and views. Items are the basic units of information. If you are using Agenda to keep a schedule, each item would be an appointment or task to do. If you are using Agenda for accounting, each item would be a transaction. If you are using Agenda for what is, in my view, its most revolutionary and valuable purpose -- manipulating very large amounts of research or reference data -- each item would be a quotation, a piece of data, a chart, or anything else you want to find later on.

Categories are the attributes you ascribe to each item. Time can be a category, and in a scheduling system events that happen today would be in the "today," "this week," "this month" and "this year" categories. Categories can be priorities -- high, medium, never. They can be the people in charge of various tasks. If you were cataloguing baseball cards, the categories could include team, position, player, and year. Categories could include zip-codes, for addresses, or yes/no listings for whether something has been done. They can be substantive themes -- "Meiji restoration," "Mississippi politics during Reconstruction" -- for research material. There is no built-in limit on the categories you can define.

Views, finally, are presentations of the information in your items, arrayed and selected according to the categories you specify. This may sound similar to what a normal data-base program does. With Paradox, dBase IV, RBase, and so on you can retrieve pieces of information, through a "query," according to the criteria you choose. ("Show me the last name, first name, and phone number for all families whose addresses have a zip code from 10001 to 10292.") The difference is that Agenda eliminates the need for queries. In most data-base programs, there is one bed-rock chunk of data, the mother lode, from which you request samplings from time to time. In its fundamental technology, Agenda also has one mother-lode of data, but -- in ways that are, again, easier to appreciate on the screen -- it creates the illusion that the information exists in small, pre-customized chunks. You can create an Agenda view called "New York City," comparable to the zip-code query above. Whenever you flip there, with one key, it can show you all the dealings you've had with anyone in New York.

The "artificial intelligence" part of the system comes from the interaction of items, categories, and views. The simplest illustration involves dates. If you type in an item saying, "Have lunch at noon next Tuesday," Agenda will automatically schedule it for the proper date and time. If you type, "Lunch meeting at noon every Tuesday," it will generate Tuesday bookings on into forever.

The rules for converting "next Tuesday" into "May 26, 1992" are built into Agenda, as they are into a number of scheduling programs. What makes Agenda unusual is that you can develop unlimited numbers of other rules. Any item including the words "Call Mom" can automatically be sent to the "High Priority" category. Anything including the words "file estimated taxes" can come up on the appropriate quarterly dates. Some obligations become more important the longer you delay them, like paying bills. Others become less important, like buying tickets for a concert that's already occurred. You can arrange for Agenda to move events of the first type up the priority list, and discreetly eliminate the second from your file. If you type in, "See Sue next Wednesday about Hmong families in Fresno," the item can automatically appear in views displaying "topics to discuss with Sue," "events next week," "immigration from Asia," "California sociology," or any other grouping you choose.

Most of the examples I've given involve one of Agenda 2.0's built-in sample applications: Planner, which is an appointment and to-do system. This is where most people start with Agenda, because its functions look familiar. Unfortunately, the very familiarity of the system undersells Agenda's potential; everyone has seen scheduling programs before, and initially this just looks like one more of them. Indeed, if your only goal were to keep track of appointments and business contacts, programs designed for that purpose and no other, such as one called Act!, would prove faster and more efficient. I use my own version of Planner, with "Call Mom"-type rules built in, and after many years of using Paradox to keep track of accounting information, I've switched to an Agenda application I developed.

But what I value most in Agenda is a function I didn't know it possessed when I first bought it. I use it every day to organize and keep track of large quantities of research information. I don't know of any other program that can do this as flexibly as Agenda can.

Suppose you had collected information on your computer in dribs and drabs, knowing that eventually would need to put it together in some organized way. Perhaps you had been copying out notes and citations for a thesis, or typing up interviews for a legal case or for a report, or collecting information from an on-line database, or in any other way storing facts in your computer for later use. (The information does obviously have to be on your computer disk, rather than on a printed page, before Agenda can deal with it. This is a fundamental barrier for most computer programs, which will be breached when Optical Character Recognition systems become reliable, and you can "scan" a passage from a book or report into your computer.) Agenda can then sort the information into usable categories, based on the rules you specify. If your project concerns the history of China, you could specify that any paragraph containing the words "John King Fairbank" could be assigned to categories such as "Harvard scholars," "long-term impact of American missionary families," and "Who Lost China controversy." If you were research world leaders, you could assign all names including "King" to the "royalty" category, except those also including the words "John Fairbank," "Martin Luther," or "Kong." Thinking up appropriate rules is time consuming, but no more so than with any other cataloguing system, computerized or manual. Once you have laid them out, Agenda can retroactively apply them to information already on your computer, and automatically categorize each new bit of data you add.

Then you have the pay-off: being able to switch from view to view and see exactly the information you are looking for. If I want to see all items concerning relations between America and China, I bring up one view. If I want to see all citations I've collected from the writings of John King Fairbank, including some that were also in the "America and China" view, I can switch there. Then, if I want to see notes on the influence the children of American missionaries in Asia, I can switch to another view and see comments from Fairbank, and Edwin Reischauer, and Henry Luce. Through a process I would love to describe, but which would seem incomprehensible to the normal civilian reader, I have used views and categories to make Agenda into an extremely powerful outlining system. Agenda's categories can be arrayed in hierarchies, like a normal outline. Chapter 1 can be a category, with sub-categories like 1-A and 1-B, each of which can have several generations of subcategories of its own. If I know that one of Fairbank's comments should show up in Chapter 7, section B, subsection 3, I can assign it to that category -- and then switch to a view to show me the material I mean to use in each section. I can use another set of categories to classify items as "Used" or "Not Used." When I am done quoting from Fairbank, Agenda can remove the item from the screen, and put it in a different view, showing when and how I used it. Neither Agenda nor anything else in computing can make up for bad organization or sloppy thought. But, like a word-processor compared to the old typewriter- and-whiteout approach, it reduces the practical barriers to writing and thinking.

What are the drawbacks to this program? It can consume a lot of time. Some programs are "hard to learn," in that you must master a lot of details before you can make the program do anything at all. Agenda is, by this standard, easy to learn but hard to master, like the Asian game of Go. It was only after I'd used it for about a year that I felt fluent in Agenda, able to tailor it precisely to my purposes and apply it in ways no one had told me about. There is no substitute for simply using the program, but along the way there are two excellent sources of advice. One is a Compuserve forum, in which Lotus technical experts respond immediately to questions about Agenda, and other users discuss what they've learned. (To reach the forum on Compuserve, type GO LOTUSB.) The other is an organization called EyeOn Associations, which publishes an expensive but very practically-minded newsletter on Agenda. The newsletter, called Coming to Order, costs $69 per year; EyeOn's address is TK. Also, the Lotus manuals that come with the program are surprisingly good.

Agenda will also increase the temptation to spend more money for a faster machine. Although Agenda will run on any IBM-style computer (and, with DOS-compatibility software like NameTK, on a Macintosh too), its performance improves dramatically on fast machines with a lot of RAM. When you run a word- processing program on a slow computer, only certain functions, like checking the spelling, are noticeably sluggish. Everything about Agenda slows down, including the crucial value of being able to switch quickly from view to view. If you don't have a 386-based computer, with at least 2 MB of RAM (or a 286 that can use expanded memory), it would be better not to try Agenda until you do.

There is one other potential complication with Agenda, which I hesitate to mention because it sounds so much worse than it usually turns out to be, but which it would be irresponsible not to mention at all. For certain users, with certain machines, Agenda can be disastrously crash-prone.

As computers have become more complex and powerful, software-designers have tried to do things that would have been impossible just a few years ago. This is especially true with 386-based computers, which are continually swapping programs in and out of certain blocks of memory. Users now take it for granted that they can half have a dozen programs, a disk cache, a mouse- driver, and other utilities all running at the same time.

Most of the time, amazingly, the juggling works, but inevitably it breaks down from time to time. When that happens and some piece of data gets misplaced, usually the worst possible consequence is that the computer freezes and must be turned off and restarted. If you're using a word-processor, you may lose your last few minutes or even hours of work. But if you're using Agenda, a small data error can, under certain nightmarish circumstances, start a chain of cascading errors that corrupt your entire Agenda data file, eliminating months of work. The early release of Agenda 2.0
contained a flaw that led to
self-induced file damage. Lotus claims that all internal flaws have been eliminated, that the vast majority of Agenda users never encounter crashes at all, and that the minority of problems that remain come from conflicts with other programs. In my own case, I had maddening daily Agenda crashes for several months, until I made adjustments in a popular (and otherwise well-behaved) disk-caching program called Super PC Kwik. From that time on, Agenda has not crashed once for me. The Agenda technical staff has become well-experienced in diagnosing the cause of crashes, which seems to vary from machine to machine. You can contact them by phone or Compuserve if problems arise. Lotus offers a no-questions-back refund if you're unsatisfied with Agenda within 90 days.

C:\WP51\WORKING\AG_ATLAN.DOC