The question is what sort of segmentation scheme to use. Intel has created one based primarily on behaviours and attitudes towards technology. Its categories are as follows:

• Functionalists - Primarily manufacturing workers (but including some office workers) who use IT occasionally but don't rely heavily on "office IT" to perform their job functions.
• Cube captains - Spend the majority of their time in the office, are very mainstream in their office IT needs and are overall very happy with the tool sets they have.
• Nomads - Heavy users of remote access and mobile IT, whether while travelling or working in remote offices.
• Global collaborators - Interface often with people around the world; they resemble nomads but work across time zones and need access to collaboration tools, anywhere, anytime.
• Tech individualists - They want and adopt early the latest IT tools and are willing to take risks with them.

These probably wouldn't be the right categories for all organisations, but I view it as a great step forward for Intel to create and address them. Intel's next step is to put them in the context of business process and business unit needs.

My own hypothesis is that the best way to segment knowledge workers would be by the roles they perform within the organisation. I would guess that determining whether you're a "field sales analyst" or a "midlevel marketing manager" would drive the type of work you do and how it could be done more productively and effectively. Of course, that will be difficult and perhaps expensive. Most organisations don't even know how many roles they have. I suspect the only role-based segments that might make sense are those in which there are many workers in a single segment, or in which better productivity or performance is mission-critical.

A great example of such a role-based approach is at the global telecom company BT, which has focused considerable effort on its 15,000 customer contact workers. The focus for these workers was less on increased productivity (typically measured in call-handling times) and more on improved customer service through better availability of relevant information and knowledge. BT implemented a new role-specific portal, BT AdvisorSpace, within its customer contact centres. BT's goal is to make available all needed information and knowledge in real time while the customer is on the phone - and eventually to bring the relevant information to the screen automatically based on the current customer transaction. Already the new system has led to a several percentage-point increase in customers feeling that their adviser was helpful and knowledgeable (it's at 97 per cent now). The advisers' confidence in the information they use is up by 23 per cent. To me, that's a great example of what an organisation can accomplish when it focuses its efforts and information resources on a particular role.

Of course, there are important ways to improve productivity that don't involve IT. One is to ensure that there are measures of productivity and effectiveness in place (easy in call centres, harder with more autonomous knowledge workers). Another is to develop business process standards such as the Capability Maturity Model I mentioned in my last column. A third is to treat every intervention into knowledge work as an experiment - with measures, a control group, clear hypotheses about the result and so forth.

I'm more confident than ever about the importance - and the difficulty - of addressing the topic of knowledge worker productivity. Just remember: It's the Next Big Thing, and you heard it here first.

Tom Davenport is director of the Accenture Institute for Strategic Change and professor of IT and management at Babson College