Irrespective of the frequency to which it is referred, the concept of "know
ledge work" remains surprisingly ill-defined and generally not well underst
ood. In this paper we look at a case study of a legal team presented with a
system to automated routine aspects of documentary evidence analysis. The
difficulties experienced reveal important lessons for the way analysts addr
ess the design of systems to support knowledge work. Initiatives popular in
the 1980s and 1990s, which encourage process perspectives, fail to highlig
ht the human elements adequately, preferring to concentrate on synchronicit
y and effectiveness of the process. In contrast we argue a need for a more
holistic and interpretivist approach. In particular, the needs to avoid con
ceptualisation through simple task analysis; to gain an understanding of co
re activities; and to identify informing activities and implicit links betw
een tasks. Knowledge workers are informed by their work and capturing the "
wrong" tasks forces them to bypass the IS. (C) 2000 Elsevier Science Ltd. A
ll rights reserved.