Tuesday, 26 February 2008

e-mail sucks

I can remember when I first got e-mail. It was long before any of my readers (I dare say) got e-mail, somewhere around 1971. E-mail (it wasn’t called that then) had a profound effect on my staff. Instead of walking in and chatting when they had an issue, they sent e-mail. But, in 1971 sending e-mail wasn’t quite as simple as it is today. It took a lot of time and only worked every now and then. They would sit for hours trying to make an e-mail go through instead of simply knocking on my door.

In the early days, you only got e-mail from people you knew.
Life was simple. Only computer scientists and military people had access to the ARPANET, which conveyed e-mail, so it was a small world. Computer scientists arranged lunch and sent each other papers. Life was simple. E-mail was grand.

Then suddenly, things started to change. First it was people who had access to all the e-mail addresses that existed at the time who felt the obligation to tell everyone what they thought about everything. Then it was people looking for a place to live or a bicycle to sell that just had to let everyone know about it.

Now everyone has e-mail. Spammers, your employees, your neighbors, business folks you want to talk to, and many people who you don’t want to talk to. For people who work in large enterprises, e-mail is out of hand. There are some important messages. It is just hard to find them when you get a few hundred a day. And then it is hard to find the ones you saved, on another day when you need to find them again. There are just so many of them. Storing them by sender doesn’t work when the same sender sends you zillions and subject headers are usually useless.

What to do?

Stop using e-mail.

Please, give me a few more paragraphs, before you stop reading because you think you are dealing with the ravings of a madman.

Amazingly enough, businesses ran quite effectively before there was e-mail. They also ran before there were computers. Nowadays both thoughts seem impossible. A lot of business simply has to be run by computer. Purchasing for example. Business sell that way, they buy that way, and nothing will stop this electronic replacement for the old method. On the other hand, business deals, and agreements although they may be mediated by e-mails and documents produced on a computer, are still pretty much face to face (or telephone) affairs. Why is this? E-bay exists after all. Why can’t business deals be handled by e-bay?

The reason is simple enough to understand. We need to trust the people we are dealing with and this means we need to know them. E-bay won’t do except for simple purchases.

Actually computers don’t work that well for anything that is very important. Really important folks in a business have secretaries to screen their calls and to handle their appointments. A good secretary knows whom her boss wants to see or talk to and whom he doesn’t. And the secretary knows the priorities of who ranks over whom in significance for her boss.

So, why can’t a computer do the secretary’s job? Why do secretaries still exist? E-mail makes it seem as if they should no longer exist. Far less people have them these days in part because there is e-mail and everyone has a computer. Nevertheless, we all would rather have a good secretary. Why is that?

A key element of a good secretary’s job is coordination. A secretary who simply arranges appointments for her boss without understanding the priorities of requests; or which meetings should have multiple attendees and who they should be; or how to limit wasting time with travel that is unimportant or badly arranged; or which are the urgent issues in the business are at a given time; cannot be very effective.

And, e-mail understands none of this.

So, you get hundreds of messages because your e-mail isn’t as bright as even the dullest secretary.

What is needed is a replacement for e-mail. And, it must be as good as a good secretary.

The good news is this can be done.

Of course the question is how it can be done. The obvious answer is wrong.

The obvious answer, one that AI people have talked about for decades, is building a computer that simulates and therefore is as smart as, your secretary. This will not happen any time soon. Your secretary knows a lot of stuff. She probably knows what you like to eat and how things are with your wife and kids, and who in the company really doesn’t like you and what people are saying behind your back. So forget that. There is no reason to, nor any real hope for, simulating all the human, everyday, knowledge that a secretary has.

Fortunately, we do not have to simulate a secretary to fix the e-mail problem in large enterprises.

Why not?

Because, in a large enterprise people do not typically act as free agents. Real people, when they are not playing a role, simply do what they want to do. So, they could be writing an e-mail to their friends about anything, about their mood, or about what they plan to do on Saturday night, or about a random thought they just had. Not so in a large enterprise. In business, people are careful to do what their job specifies that they should be doing. They might send a personal message as well, but that is not my point. A lot of what they are writing about is what they should be writing about, what they are obligated by the nature of their job to be writing about. Furthermore they are writing to precisely those people who need to know about the issues the writer is communicating, often because there is some action the recipients will need to take as a result of having read the e-mail. Or at least that is what should be going on. Seen that way, e-mail is not a free form exercise but part of a very routine set of processes that fit together in very specific way within the enterprise.

This is good news.

It is good news because some e-mails are more important than others. A computer can gauge an e-mail’s importance correctly if that computer has a model of what is important in an enterprise. Your e-mail program does not know that an e-mail about bad weather or crewing problems is urgent for someone in the shipping business. Nor does it know who is tracking what and who needs to know what. E-mail relies on the idea that the sender knows all that and that urgency can be marked by simply saying something is urgent. But if everything is marked urgent or if people who need to know are left out of a mailing, there is a problem. Further if an e-mail should have caused its recipient to put an action on his to do list but it didn’t because he spilled his coffee while reading it, something may not get done later on down the line. None of this has to happen if the message system that an enterprise employs is based on a detailed process model of how information flows, to whom and why, and what goals that information is meant to relate to in the enterprise. In other words, a complete model of the enterprise must be the basis of all message traffic. That model is what we rely upon the secretary to have. Of course, the secretary may not actually have that model or may have a defective or out of date model.

Business communication in a complex enterprise needs a detailed model of who does what in the enterprise and each person’s particular role in every critical situation. This is not only possible -- it is mission critical.

How can this be done?

By understanding and detailing everyone’s role and the tasks associated with that role, all tied to the goals that those tasks are meant to satisfy. Add in a goal calculus that makes clear what goals supercede what other goals in given circumstances and you have a model of the enterprise. Then, an e-mail sent by a person is not sent by that person but by that role and the subject is not whatever the sender said the subject was, the subject is how what the sender is about to communicate will affect a given process in the enterprise in the context of the satisfaction of a given goal.

Do you know what everyone in your company is doing and why they are doing it? Then stop allowing them to write whatever they want to whomever they want and start making sure that e-mails are part of the design of how things work, and how coordination happens. Don’t let e-mail do what it was never meant to do, namely make your enterprise function well. E-mail was simply not invented for that purpose.

Tuesday, 19 February 2008

Knowledge Management For Real People

Knowledge Management (KM) means different things to different people.

Based on its experience and on the existing large body of knowledge management literature, the team decided that knowledge management software should provide virtual “places” where users can organize information, services, and tools to support their particular needs, while at the same time maintaining and updating information in a more general context.

Or, in other words, Lotus thinks KM is the creation of a communal library. This notion of KM as the construction of a communal library is made clearer by a recent IBM publication touting something Satyam developed for the petroleum industry called Spandan, which is “a complete knowledge management (KM) portal built from several IBM Lotus and IBM WebSphere software products.”

Spandan consolidates all the knowledge in the R&D organization into a central knowledge repository, organized in a taxonomy that makes sense to everyone," says Agrawal. "It captures lessons learned and best practices as they're created or refined. It enables users to quickly find both explicit knowledge, such as knowledge in documents, and tacit knowledge, such as knowledge in discussion threads or in an expert's mind through a single unified search. And it combines all these functions with collaboration features, such as instant messaging and teaming, that help people work together in context with knowledge.

Of course, there is a problem with communal libraries, ones that are created by the users of a system or the knowledge workers in a company. Not everyone wants to contribute to them. In fact, not everyone wants to even bother using them. Companies often have to resort to forcing people to use them and to contribute to them.

Satyam proudly announced its solution for this:

Satyam configured Spandan to automatically award knowledge-points, or "K-points," to scientists as they contributed to the system, and to tally and publish K-points for each scientist at the end of every quarter. Maybe more important, Satyam preloaded the system with knowledge and data that was most important to the scientists. "We populated the repository with the previous five years of project reports, technical data sheets, tour reports and all documents types they identified as the ones they consulted most," says Agrawal. "This made them start using it, and once they saw how useful it was to be able to find important information quickly, they were motivated to add more."

Their message is clear. If you make people use a collaborative library they will. (And otherwise they aren’t likely to.) Personally I haven’t been in a library, virtual or real, in twenty years. Guess I wasn’t offered enough K-points.

(Of course some people do contribute to communal encyclopedias like Wikipedia. But it is a really small proportion of the community does so.)

What is the real issue in KM? This is easy to answer.

Libraries are about your attempts to find information outside your normal activities. Well-indexed libraries make information easier (though often not easy) to find.

In life, the bulk of information finds you in the course of your every day activities, when you interact with other people in the enterprise, or when a colleague needs your help on a problem. Or, most importantly … when you are reminded of something you have already experienced. There are two points to be made here.

1. People do not usually seek information – it seeks them.
2. When people do seek information, their greatest source of information is themselves and they find what they need without actually having to try.

Any KM system that hopes to work in a real world where people act like people and not like members of a community where they have to play their part in order to get K-points needs to be built around the following principles:

1. Information finds the user by knowing what the user is doing and therefore knowing what information the user might need.
2. Users do not ever try to add information into the KMS. Their work is naturally part of the KMS and is added automatically.
3. The KMS knows what it knows and is getting smarter all the time.

For people who work in the real world, the key issues must find them in their natural environment without expect them to start searching for them.

It takes special circumstances to get people to contribute to the library. Here it is, again from the cited IBM article:

But like most expert communities, the oil company's R&D scientists were reluctant at first, and saw knowledge sharing as a potential threat. "The scientists were worried that contributing and sharing knowledge would lessen their importance in the organization," Agrawal says. "For example, if one scientist was the exclusive expert in a certain area, and a question in that area came from the Parliament or the Ministry of Petroleum, that person got his or her chance to shine and to feel important. Our challenge was to convince them that they became more important by contributing, rather than hoarding."

Scientists in any industry are not even remotely like people who work on real life everyday jobs. Scientists publish papers about original ideas and they worry that those ideas will be stolen. People with more normal jobs worry about doing their jobs properly and receive information and work within the bounds of e-mail and an enterprise software system. They do not expect to do research in a library like a scientist would. And, they may not even know when they need more information, while it might be very important for the enterprise if they did. These are very dissimilar worlds.

Any KMS for the normal world of work must understand and have a detailed model of the world in which that work is taking place. If the knowledge to be managed is shipping knowledge, then the KM system must know about shipping in detail so that it doesn’t even for a second think that a bridge procedure refers to a card game, and it cares about weather information if and only if a ship it cares about might encounter that weather. A KM for shipping needs to know about issues about particular ships, and ships like those ships, and what to do about similar issues when they happen in new circumstances. In short a KM for shipping must serve as a corporate memory that knows more than any one individual might know. It must track what is going on in daily events and relate what is going on at the moment to what has gone on in the past to see if it can help manage goal conflicts and hence manage risk.

In order to use such a KMS, the system must use indices that find knowledge that are intuitive. But, what is intuitive in one line of work is not intuitive in another. For shipping the system must speak shipping language and use shipping concepts to index shipping information. Any system that did not d o this would be about as useful as hiring a librarian who specialized in English literature to catalogue you companies knowledge.

In a KMS that uses roles and tasks to describe actions there is an implicit assumption that given an action, it is reasonable to expect another particular action to follow.

The organizational principles of a KMS for real life tasks must contain predictions and expectations about the normal flow of events in standard situations. Whenever an expectation derived from that structure fails, its failure must be marked. Problems must be stored with respect to the action sequence in which they took place. Real world corporate processes are not like science. These processes are predictable and understood. The question is not how to manage document flow but how to manage processes for opportunity and for risk. KM for industry is not KM for scientists.

The problem for KM in the corporate world therefore, is to make use of the organizing principles in a work situation and utilize them in a KMS that organizes the information in that domain. These organizing principles will certainly be about roles and tasks and the goals associated with them. The ways in which goals can be satisfied or blocked and the plans that are used to execute those goals must be the cornerstone of any KMS.

[1] The Lotus Knowledge Discovery System: Tools and experiences
W. Pohs G. Pinder C. Dougherty M. White, IBM SYSTEMS JOURNAL, VOL 40, NO 4, 2001

[2] Satyam’s IBM Lotus KM Solution helps petroleum company fight change with knowledge, (IBM publication on the web)

Wednesday, 13 February 2008

Knowledge Management = Document Management?

Why do people believe this?

All knowledge is not contained in documents, that much seems obvious.

But let’s assume that this were the case. Let’s assume that all the knowledge of the world was written down in identifiable documents. How would we find what we wanted?

Not by labeling the document well. Why not? Because documents have more than one idea in them typically.

In 1952, The Encyclopedia Britannica published Their Great Books Series. They decided which books in history were the most important and published a 54-volume set. The editor was Mortimer Adler. There were very few books of recent vintage in this set. When I asked Mr. Adler about this, he told me that nothing of interest had really been written in the 20th century, but that is another story.

Here is the reason I am mentioning this. Encyclopedia Britannica also published in 1952, something called the Syntopicon, which was touted as an “Index to the Great Ideas.” Note that it is not documents that were indexed by Adler but ideas. So under Education, for example, you have six pages of subtopics such as the disadvantages of being educated which refer to pages in texts written by Plato, Aristotle, Descartes, Locke and others.

Creating this index was a massive undertaking, but an important one. These books would be useless as references without the Syntopicon. Note that it is ideas that are referenced, not documents.

This leads to some obvious conclusions:

1. Indexing knowledge is what knowledge management is all about.
2. It is ideas, not documents, that need to be indexed.
3. If knowledge in a given field is to be made useful in a computer-based knowledge management system, that knowledge had better first be indexed in ways that make sense to the practitioners in that field, and can be found easily by a computer program that is meant to help them
4. In the enterprise as compared to a research environment, ideas need to be indexed by the occasions in which they will need to be used and software needs to deliver them to people at that time.