Introduction

Organisational decisions increasingly include social, environmental and economic concerns, and they have become much more complex and interconnected than in the past. More effective ways must be found to support the vast array of knowledge content (Holsapple & Joshi, 2001) that will be required in the highly interconnected, wicked decision-making situations of the future (Courtney, 2001). Moreover, many organisations are increasingly adopting collaborative work practices as a means to boost creativity, innovation and productivity. The tasks of team workers are frequently non-routine and knowledge-intensive (Sense, 2007). Collaborators in teams should not only apportion the work based on individual expertise, but also achieve a seamless flow of, both formal and informal (Ramesh & Tiwana, 1999), contextual (Kwan & Balasubramanian, 2003; Ahn et al., 2005) knowledge among the team members.

Knowledge management (KM) has received considerable attention as an enabler for decision making (Holsapple, 2001; Bolloju et al., 2002) mainly because, on the one hand, decision making is a knowledge-intensive activity with knowledge as its raw materials and products and, on the other, the aim of KM is to provide timely and contextual knowledge to decision makers. KM systems have been used to support decision making in knowledge-intensive environments such as constructions (Chen Tan et al., 2006), law enforcement (Chen et al., 2002), nuclear plant operations (Spangler & Peters, 2001), marketing/customer relationship management (Shaw et al., 2001) and health care administration (Pedersena & Larsen, 2001).

Although knowledge content is important for organisational decision making and knowledge flow is important in collaborative work practices, the methods, tools and actual KM implementations have largely failed to integrate the two main approaches for KM, which are commonly referred to as codification and socialisation, respectively (Hansen et al., 1999; Kankanhalli et al., 2003; Kühn & Abecker, 1999):

• The codification approach focuses on knowledge content – its creation, storage and reuse in computer-based organisational memories (Lehner, 2000). The main role of information technology in this approach is to support the organisation and retrieval of articulated, documented knowledge. It is also referred to as the content-centred approach.
• The socialisation approach focuses on knowledge flow and mainly regards KM as a social communication process. In this approach, knowledge is closely tied to the individual who developed it and is shared mainly through person-to-person contact. The main role of information technology in this approach is to help people communicate knowledge, rather than store it. It is also referred to as the personalisation approach.

The Know-Net KM solution (Mentzas et al., 2002) is built around the concept of knowledge assets. It aims at a harmonisation of the codification and socialisation approaches. This paper gives an overview of the Know-Net software system and describes how the system utilises comprehensive metadata as part of so-called Knowledge Objects (KO) that are used for harmonising the two approaches. Moreover, it highlights, using examples from a real KM implementation, how the Know-Net method complements the system in ensuring that the two approaches are truly integrated.

System-based integration of the codification and socialisation KM approaches

Motivation and basics of the Know-Net system

If a comprehensive IT infrastructure for KM is to harmonise the codification and socialisation approaches, it should provide seamless access and linkages to the whole wealth of explicit and tacit knowledge residing in an organisation. This must be done in a situation facing manifold aspects of heterogeneity, to mention but a few of the most important: (1) there may exist different versions and multiple storage formats of the same content; (2) in a large organisation, information about the same topic may be organisationally dispersed due to the division of labour in different departments that may maintain their local organisational memory systems with a unique focus and perspective; (3) similarly, geographical dispersion may occur, especially in international businesses; (4) there may exist different kinds of information about the same topic (a product, a market, a technology, etc.), namely new ideas, business tenders, discussion threads, best practices, lessons learned, project reports, technical documentation, etc., all about the same topic; such different kinds of information are typically spread over different tools and applications.

In order to deal with the aforementioned manifold dimensions of heterogeneity, the Know-Net system has been developed based upon the idea of defining and managing KOs instead of directly managing 'pure' content. A KO contains content plus a set of metadata. From the content point of view, a KO can comprise, inter alia, a lesson learned document, a best practice template, a documented new product idea, the answer to a frequently asked question or a contribution to an online discussion (see Mentzas et al. (2002) for further detail on KOs). KOs act as the 'glue' between the two approaches of codification (they do represent content) and socialisation (they are created, discussed, shared, consolidated and applied in KM processes).

Arguably, there is no definitive answer as to which metadata should be provided for content annotation. Standardisation efforts are emanating from the Digital Library community (e.g., the Dublin Core initiative), and there are specific answers tailored for the KM domain (see Abecker et al., 1998; Kingston & Macintosh, 2000). The Know-Net system is configurable in the sense that the complete metadata schema can be defined in the Ontology Editor described below. There are two relevant dimensions. (1) It makes sense to have some contextual metadata (i.e., in which process or context has content been created?). Within the Know-Net system, such contextual information is partially derived from the application that is used to create content (e.g., Lotus Domino, any text editor, or a specific KM application). (2) Further, there should be metadata about the content of the information artefact (i.e., the topic(s) it deals with). We build graphical representations of concept maps reflecting multiple views and facets of the actual application domain. These graphical maps can be used: (a) at storage time, for conceptual indexing of content; and (b) at search time, for graphically specifying search conditions when looking up KOs. We call these maps metadata schema (defining the relevant metadata attributes) and indexing ontologies (defining domain knowledge structures for metadata-attribute values) because they represent important ontological structures underlying the users' work. As an example to illustrate the idea, we refer to O'Leary (1998), which identifies the ontological structures used by multinational consulting firms to organise their lessons learned documents. For instance, some of the most important dimensions there are the industry of a customer (automotive, chemicals, etc.), the process to be improved (customer relationship management, new product development, etc.), and the tools and methods applied in delivering services to clients.

KO annotation and retrieval approach

Figure 1 gives a high-level schematic overview of our KO annotation and retrieval approach. The whole approach is based upon a freely configurable metadata schema defined with the Ontology Editor. Here we can specify a hierarchical schema of metadata to be attached to each KO. When specifying the ranges of allowed values for a metadata attribute, we not only provide for simple data types like strings, dates or enumeration types but also give the opportunity to describe the set of possible attribute values by a graphical map of some domain knowledge structure.

Figure 1.

KO annotation and retrieval approach.

Full figure and legend (125K)

The metadata schema is used for three purposes. First, it is the database schema defining allowed instances of KOs. The database of the Know-Net system contains an entry for each KO, which specifies its metadata values plus a link to the original information artefact that may reside elsewhere (in a file server, for example). Second, the metadata schema is used to configure the user input forms for KOs via the Document Indexing Interface. The input interface is created dynamically, depending on the current metadata schema and indexing ontologies. Third, the metadata schema is used in the dynamic generation of the Advanced Search Component, which allows users to browse the metadata schema and specify their information needs by articulating search constraints (i.e., metadata-attribute values required for a relevant KO) in a graphical and step-wise manner, as explained below.

While users specify their knowledge requirements as they proceed through the metadata schema, each interface activity is directly passed to the underlying search engine, which immediately updates the current result set corresponding to the actual knowledge need formulation. This step-wise or incremental update of retrieval results allows for

1. intertwined searching and direct interaction with the respective KOs (which can be opened for reading/editing in separate browser windows while the search goes on in another window), and
2. further specifying or refining the knowledge requirement (in the case that metadata or source inspection of currently retrieved KOs show that this formulation must be refined).

The three main interface elements of the metadata-based annotation and retrieval engine (Document Indexing Interface, Advanced Search Interface and Ontology Editor) are discussed in more detail in the following sections.

The document indexing interface

Because the main aim of our system is seamlessly to integrate the codification and socialisation approaches, metadata annotation is directly integrated into the input forms of any KM application, including those that support the socialisation approach. This means, for example, that each input form of a discussion or a library database (as shown in the background screenshot of Figure 2) contains input fields for the metadata attributes. For each such attribute that has been associated with an indexing ontology when defining the metadata schema, a link called 'Choose' is added to the right of the input field. Clicking on this link opens the appropriate concept map as another clickable image, as shown in the foreground screenshot of Figure 2.

Figure 2.

Indexing of KOs.

Full figure and legend (634K)

Left-clicking on a concept in the window automatically inserts the respective index concept into the attribute field; de-selecting a concept by clicking again removes it from the list of index terms. In the example, we illustrate the indexing dimensions used for describing a research project: namely, (1) the tangible 'Products' to be delivered by the project; (2) the 'Document Status' (draft, final, approved); (3) the 'Content Type' (idea, discussion entry, project deliverable, etc.); (4) the 'Work-Package' to which a document relates; (4) 'Administrative' issues to which the KO is possibly related (such as Project Contract, Cost Statements, Intellectual Property Rights Agreement, etc.); (5) 'Meetings' relevant to a document or presentation; and (6) finally, 'External Inputs' to the project a KO could come from (such as other research groups, relevant related work, basic technologies, etc.) We see that a KO can often contain indexes with respect to several of these dimensions, which enables flexible usage of the tool and powerful search facilities.

The advanced search interface (KASI)

When the user enters the KASI via the 'Search' button, s/he will reach a screen as shown in Figure 3. The screen is divided into three main areas:

• The upper part (region 1 of Figure 3) containing control elements
• The middle part (regions 2 and 3 of Figure 3) allowing the user graphically to specify his/her information needs
• The lower part (region 4 of Figure 3) displaying query results and other output streams if wished

Figure 3.

Advanced metadata search.

Full figure and legend (192K)

In order to ease potential problems arising through complex, under-specified or not (initially) fully specified searches in heterogeneous information spaces, three main ideas underpin the KASI search application:

1. The metadata schema itself is browsable.
2. Attribute value models can be denoted for indexing and for querying in a graphical manner, which exploits an individual's ability to remember and deal with visual structures.
3. Each user action triggers immediate feedback in the result panel, thus supporting a kind of 'iterative refinement' search.

These ideas are reflected in the KASI as follows. First, in the interface region (region 2 of Figure 3), the metadata schema is presented in a manner that can be folded and unfolded to locate the metadata attribute the user wishes to employ in expressing his/her information need. If there is a large, difficult-to-oversee metadata schema that is uncomfortable to navigate, or if the user does not know the exact part of the metadata s/he has to search in detail, the three buttons at the left of the navigation bar (region 1 of Figure 3) can be used to unfold or fold together, respectively, the whole folder tree level-by-level, or unfold it completely with one click.

Another aid to navigation in the metadata schema is the text search field included in the control and navigation bar (region 1 of Figure 3). Here, the user can type in an arbitrary string and, if there is a metadata attribute or any possible value of some attribute containing this string, the metadata-schema tree is opened up to this attribute and the range of values of this attribute is shown on the right-hand side (region 3 of Figure 3). Having found the appropriate metadata attribute in the schema, it can be marked either by clicking to select, clicking again to deselect or by using buttons in the control bar.

The interface region on the right-hand side (region 3 of Figure 3) displays the possible values allowed for the attribute selected on the left-hand side. Having selected an attribute in the process of characterising the actual information needed, the specific search condition can be formulated. Three types of metadata attributes are available: (i) standards attributes (see region 1 of Figure 4), such as date, that can be expressed via pull-down menus; (ii) index maps (see region 2 of Figure 4 and the section 'The index ontology editor' below) – some attributes can be graphically selected (green colour) or de-selected (red colour) if they have been defined as Index Ontologies during ontology building; and (iii) value lists – attributes, the range of which has been declared as a value list, are presented as lists with click-boxes (see the 'Author' attribute in region 3 of Figure 4).

Figure 4.

Use of the advanced search interface (KASI).

Full figure and legend (215K)

The actual answer set resulting from a current query is always displayed in region 3 of Figure 3, in the lower part of the system interface. The other functionalities of region 3 concern tool inspection and debugging and are somewhat peripheral for the context of this paper.

A further KASI service, offered in region 4 of Figure 3, is to define personal push agents that periodically (e.g., every hour, once a day, once a week) execute complex queries and actively deliver the search results to the user via e-mail. To this end, the user enters his/her 'Personal Agents Space', authenticates himself/herself, specifies a search condition as discussed above and then saves this query with a name, a timing condition and a delivery mode. Because specifying the 'right query' for a complex problem might be difficult, but of interest to more than one person in a group, queries can be marked 'public'. Public queries are visible to all users registered in the system who are allowed to subscribe and are then continuously provided with the respective search results. Public queries can also be used for implementing 'mandatory readings' that must be considered by all group members.

The index ontology editor

The Index Ontology Editor allows the system administrator to design and change the metadata schema. The Editor supports a multi-faceted attribute hierarchy (i.e., Project X can be part of the Industry, Process and Tools index ontologies). Relations in the index ontologies are restricted to taxonomical (is-a) ones, as these have been found sufficient in our application examples for metadata annotation and indexing purposes. In particular, we did not need specific reasoning that would benefit from other types of semantic relations (like part-of or domain-specific relations). The tool output is used to determine the profile of the search interface: that is, the metadata attributes to be presented. Moreover, the tool automatically produces the concept maps – the graphical representations of indexing ontologies.

Case study

The main purpose of this section is to illustrate, using examples from a real KM implementation, how the Know-Net method complements the Know-Net system in harmonising the codification and socialisation approaches. The Know-Net method helps to (i) strategically plan and operationally design a holistic KM infrastructure that is aligned with business strategy; (ii) facilitate in planning the organisational changes required for KM to succeed; and (iii) provide ways of evaluating the impact of the KM initiative on the overall performance of the organisation. A detailed description of the methodological modules cited in this paper is presented in Apostolou et al. (2007).

Like the Know-Net system, the Know-Net method exploits the theoretical approach of using KOs as the unifying elements of the two approaches. The core methodology includes a set of 'audit-leverage' module pairs, which assist the KM consultant in analysing and leveraging, using the Know-Net system, the core knowledge assets of an organisation. For example, there are module pairs that focus on the analysis and leverage of knowledge in business processes and in knowledge networking structures such as Communities of Practice. Furthermore, there exists one cornerstone module (named 'Develop the Knowledge Asset Schema') that focuses on the development of the metadata schema and indexing ontologies to be used for annotating KOs. This module acts as a unifier of the Know-Net method, being constructed with input from the audit modules, while supporting the consistent execution of the leverage modules.

Inter alia, all audit modules aim to identify in detail the knowledge assets utilised in the 'as-is' situation. Leverage modules aim to specify the 'to-be' situation, in which knowledge assets should be created, shared and used. The so-called 'Analyse Business Processes' module mainly examines knowledge as content (codification approach). It produces process maps depicting the key knowledge assets currently being used or created in the business processes under examination. The module 'Develop the Knowledge Asset Schema' collects this information, along with similar information about the same knowledge assets generated by other audit modules, arranges possible overlappings, logically groups information and creates the merged metadata schema, indexing ontologies and KO representations for the knowledge assets under examination. Subsequently, the module 'Leverage Knowledge Networks' designs and organises knowledge networks and proposes the metadata schema specified in module 'Develop the Knowledge Asset Schema' for the future annotation of KOs. The 'Leverage Knowledge Networks' module focuses mainly on knowledge flows within collaborating teams (socialisation approach). Figure 5 illustrates this method-based harmonisation of the codification and socialisation approaches. It should be noted that merging ontologies is a non-trivial task and a topic of ongoing research work (see, for example, de Bruin et al., 2004; Dou et al., 2004).

Figure 5.

Homogeneous representation and integration of business process-related knowledge with knowledge network-based knowledge.

Full figure and legend (160K)

In the following, we present the results of the application of the Know-Net method and system in the multi-national management consultancy, MC-Co. Note: MC-Co is a fictitious name for a real management consulting firm. The pilot addressed two major goals: (1) One of the three focal areas of the KM initiative in MC-Co (derived after the application of the strategic planning component of the method – not presented herein) concerns the 'deliver services' business process, which is a primary operational business activity of any consulting firm. Through the introduction of KM-related processes, the aim was to avoid 're-inventing the wheel' or otherwise duplicate tasks during the planning and execution phase of a consulting assignment. (2) The second focal area refers to the development of Thematic Area Networks, which have been identified as the company's core knowledge networks. Thematic Area Networks are communities of practice developed around the core knowledge areas (i.e., thematic or subject matter areas) that comprise the cornerstone of MC-Co's service offerings.

'Deliver Services' business process

Initially, the 'Deliver Services' business process was carried out in two basic steps:

1. Initiate project – Project manager assigned. The project manager identified the scope of the consulting engagement, formed the project team, identified the objectives of the assignment and developed an overall work plan for the project.

2. Execute project – This was the actual implementation phase, in which the project team worked to deliver the system/study to the client.

In this process, several knowledge-related problems were identified. First of all, the phenomenon of 're-inventing the wheel' arose, primarily through ignorance. Consultants were duplicating work, simply because they did not know that it had already been done for a previous client and that the knowledge therefore existed. Moreover, knowledge sharing was limited to informal communities of experts, a scenario that was closely related to the lack of a systematic approach for capturing, organising and sharing knowledge.

To support the 'deliver services' function, a new process was designed in two cycles. During the first cycle, three specific actions were taken: (i) the introduction of two additional process steps; (ii) the introduction of KM processes; and (iii) the development of a customised KM application managing the so-called 'knowledge input–output–review tables'. The two new process steps, in addition to the existing steps 'initiate project' and 'execute projects', were 'plan project' and 'evaluate project' as explained below.

Plan project – This step encompassed all the planning activities, starting with the identification of the scope of the consulting engagement and the formation of the project team, and ending with the detailed project plan, including timeline, individual responsibilities, specific tasks, software development/systems integration details and financial budgeting. In this step, non-financial project targets were set (e.g., development of new know-how, client relationship improvement). Criteria for quality assurance were set and a Quality Director appointed.

Evaluate project – This was the closing phase of the project. The team evaluated the results, assessed their impact and collected lessons learned and client feedback.

As far as the newly introduced KM processes were concerned, these were as follows:

• Filling out the 'Knowledge Identity Report' (i.e., a report containing information about the practices/service/thematic areas, knowledge input, knowledge output, knowledge reviews to be used in the project) during the initiation project phase
• Generating Knowledge Review Reports during the course of the project to enhance, refine or correct the content of the Knowledge Identity Reports
• Formulating the Knowledge Final Report at the end of the project
• Actively notifying the corporate Knowledge Office about new knowledge (e.g., methodology, best practice) developed within the project, outdated past knowledge, and new knowledge needs arising from the project.

Although it was a significant step forward, the newly introduced activities did not immediately produce the expected results. The main cause of the problem was attributed to the strict implementation of this 'system-based' approach to all projects, without regard for their importance or duration. There was much work to be done, a number of forms to be filled out and some resistance from the people responsible for carrying out those tasks. A second cycle of action was therefore put forward, characterised by a loosening of the implementation of the four-step process. The full process was followed strictly only in the case of selected, major projects. For less important ones, only the Knowledge Identity Report was mandatory. Moreover, an After Action Review step was introduced, in which the Knowledge Final Report for each project was produced (Figure 6).

Figure 6.

Knowledge leveraging within the 'Deliver Service' process.

Full figure and legend (72K)

Figure 7 shows screenshots of an application of the Know-Net system, customised to support the collection and storage of KOs that have been identified during the analysis of the 'Deliver Services' business process. The screenshot depicts the input forms for the 'Knowledge Input', 'Knowledge Output' and 'Knowledge Review' KOs. Figure 8 shows all the KOs identified, their attributes and their relations to other KOs and to Indexing Ontologies.

Figure 7.

Custom KM application in support of the 'Deliver Service' process.

Full figure and legend (245K)

Figure 8.

Knowledge assets and main related KOs and attributes linked to the 'Deliver Services' process. 'IO' refers to indexing ontology and 'A' to attribute. For instance, 'Subject Matter' can take its value from a three-level indexing ontology of approximately 500 terms.

Full figure and legend (95K)

Development of thematic area networks

The main knowledge assets required for service delivery are subject knowledge, industry knowledge and knowledge of management methodologies. Following the 'Develop Knowledge Networks' module of the Know-Net method, one Thematic Area Manager was appointed for almost every network. Thematic Area Managers were active consultants: subject matter experts who were responsible for collecting, storing, updating and advancing knowledge in their specific area of expertise. A groupware application of the Know-Net system with discussion databases, real-time collaboration and other group collaboration facilities was customised in order to support Thematic Area Networks. Figure 9 shows the KOs identified, their attributes and their relations to other KOs and to Indexing Ontologies.

Figure 9.

Knowledge assets and main related KOs and attributes linked to the thematic area networks. 'IO' refers to indexing ontology and 'A' to attribute.

Full figure and legend (60K)

To leverage Thematic Area Networks, effort was put into creating informal settings for member interaction. Because the emphasis was on open dialogue with no pressure to arrive at a resolution, the Thematic Area Manager was continually canvassing to identify new areas of interest or challenge for the community. The mechanism instituted for the recognition of participation was for Thematic Area Managers to inform senior management of successes. This information was accompanied with a request for a personal note of appreciation from senior management to individuals, commending their work and acknowledging how their contribution has affected the bottom line.

The problems encountered during this implementation were mainly related to the time required for Thematic Area Managers actively and effectively to fulfil their new role. It was regarded by consultants as 'peripheral' and not as a main activity. In a move to resolve this problem, senior executives adopted a more formal approach to Thematic Area management in order to promote it as a major business activity. As a first step, time was explicitly allocated for carrying out this task, while the performance of Thematic Area Managers was linked to the company's performance evaluation system.

Evaluation results

In order to evaluate the KM initiative in MC-Co, we followed the steps outlined in the Know-Net measurement method. We started by identifying the firm's vision, strategy, critical success factors (CSFs), key knowledge assets associated with the CSFs and related performance measurements. The three CSFs for MC-Co were: (1) to improve consultants' competence development and informal knowledge sharing through active participation in Thematic Areas; (2) to reuse and capture new knowledge resulting from projects; and (3) to generate revenues from new services. Table 1 shows the measures developed for each CSF and related results after 1 year of company-wide system usage. From a quantitative perspective, one can see that the only areas where the pilot fell behind targets were consultants' participation in Thematic Area Networks (Measures 1 and 3) and idea contribution (Measure 5). These two areas require deep integration of KM processes and systems in the day-to-day working life of consultants, something that was difficult to accomplish fully within the limited period or the pilot.

Table 1 - MC-Co's measurements.

Full table

In the following section, we discuss how the use of the Know-Net system affected the main processes in the knowledge cycle as well as decision making within MC-Co. The discussion is enriched with qualitative data from interviews with key stakeholders.

Knowledge generation

Brainstorming was a typical knowledge-generation technique in MC-Co. Consultants would collaboratively brainstorm to develop a strategy to win a client account, to plan an assignment or to collect lessons learned. With the introduction of Know-Net, new ideas (Measure 5 in Table 1) were captured electronically. Any consultant could freely enter a new idea in the system. The system was capable of supporting a review and evaluation workflow for ideas. This feature was, however, not activated in order to avoid discouraging consultants from entering their ideas. Although this approach was just one small step in holistically managing new ideas, it helped in taking better advantage of ideas than before by providing persistence and visibility to ideas, throughout the company.

Apart from internal brainstorming, a significant percentage of knowledge in MC-Co was developed in the context of client assignments as explained in the section 'Deliver Services business process' and captured as learnings/after-action reviews (Measure 6) and best-practices/best-knowledge (Measure 7). Furthermore, a series of successful collaborations with strategic partners (typically specialised and international consultancies) had given MC-Co the opportunity to acquire valuable experience and know-how. There was an effort to capture this experience and know-how, customise it for the local markets and combine it with related know-how of MC-Co in order to develop new service offerings. The end result was called 'Service Knowledge Packs' (Measure 8) and included methodologies, analytical tools and other material required for service delivery. In addition, MC-Co assessed the sales impact of its new service offerings with Measure 9 'Percentage-share fees from new services'. A partner recalls: 'When I joined MC-Co, a team was working on an IT strategy development project for a large public sector organisation. For MC-Co, it was the first project of its kind, and very little in-house knowledge was available. We were collaborating with a leading international consulting company. Based on our experience during that project and its final outcome, we developed our own methodology for similar IT strategy assignments that was captured in the related "Service Knowledge Pack", in Know-Net. The methodology was later used in many other similar projects. Any refinements or extension to the methodology were also captured in Know-Net. Now, it is one of the methodologies that is well-tested and extensively used, although it was not developed from scratch within MC-Co.'

Knowledge organisation

The physical library of MC-Co was considered limited and outdated, containing primarily deliverables and proposals. The information that was collected and bought by consultants during the course of projects remained, for the most part, with team members and was eventually lost. The electronic library section of Know-Net (Measure 4) provided an easy mechanism to upload and annotate the electronic content each team had collected. Moreover, the consultants' skills and competencies section of Know-Net (Measure 2) was used to codify some of the knowledge links that had been logged in the minds of top management and founding members. Examples of such tacit knowledge include who might know what, who has worked on which assignments in the past, what reports have been bought, and who might have contacts in a particular area.

Knowledge sharing

Before the introduction of Know-Net, knowledge sharing within MC-Co occurred at the individual and team level, stimulated by specific projects. While some informal networks existed, there was no formal, organisation-wide system to share knowledge and learning. As a result, at a particular point in time, someone or other in the organisation was 'reinventing the wheel'. It often happened that two teams were struggling with the same issues in complete isolation from one another.

A Project Manager from Public Services recalls: 'We were working for a large public insurance company on an IT strategy development assignment. During project initiation, we intuitively thought of following a specific IT methodology that had been developed by MC-Co and used for many such projects. The approach that we would follow called for a very detailed analysis, whereas the brief for the assignment did not require us to go into minute details. My participation in the IT strategy Thematic Area Network helped me locate another team that was finishing on a similar assignment and using a different, higher level approach that I thought would have been more appropriate for our project as well. Luckily, this approach was documented in the Knowledge Identity Report of the other team's project and was annotated with appropriate metadata that helped us locate it. This helped us save a lot of time collecting some meaningless data and researching some useless details.'

A partner from MC-Co explained how the introduction of the system helped align bids to actual assignments: 'The people who prepare the bids are mostly the ones who just have theoretical knowledge and no real life experience of such projects. Since consultants are rarely consulted at the bid preparation stage, it is no surprise that they go overboard with promises to the client and, when the project team begins to work on the assignment, they find it unrealistic to deliver everything promised within the specified time. Using Know-Net, people preparing the bids are able to browse through best practices and Learnings/After Action Reviews captured by active consultants. This gives them a better understanding of what is feasible and what is not.'

Decision making

Decision making in MC-Co was a highly unstructured activity – non-repetitive and non-routine. A decision in MC-Co was typically taken in a unique context and might never again be repeated. The issues pertinent to producing a decision were typically not well understood. The alternatives from which an informed choice should have made were vague, difficult to compare and contrast and could not easily be evaluated with respect to the company's purposes and goals. There was great difficulty in even identifying the alternatives. A partner in MC-Co explained: 'The biggest challenge in taking decisions in MC-Co was the scarcity or even unavailability of relevant knowledge. Besides having relevant knowledge more readily available, the KM system helped us document decision-making cases so that we could take advantage of past learnings in future decision making situations.'

Decision making in MC-Co typically took place in two contexts. First, operational decision making was exercised by project managers seeking to ensure that specific project-related activities were carried out effectively and efficiently by the project team. Knowledge collected, organised and made available through Know-Net helped to leverage lessons learned, best practices and expertise in everyday decision making. Second, in strategic decision making, the company was seeking to determine overall company objectives and to plan organisational processes and resources in the light of a changing business environment. Strategic decision making required ample availability of market knowledge as well as business intelligence. Through the introduction of Thematic Area Networks focusing on different market sectors, the company was able to collect valuable market knowledge (Measure 1), contributed by active consultants specialising in specific industrial sectors, thus broadening the participation of consultants in the strategic decision-making process.

Summary, related work and outlook

The Know-Net KM system outlined in this paper builds upon an ontology-based information annotation and retrieval approach that provides (i) seamless access to explicit knowledge and (ii) linkages to tacit knowledge residing in an organisation. We have illustrated how the Know-Net method complements the system in harmonising the codification and socialisation KM approaches. A plethora of frameworks and methods have been developed for KM (e.g., Rubenstein-Montano et al., 2001; Mentzas et al., 2002). Research into the direction of a system-method combination for KM is somewhat scarce; however, there are examples:

• The GPO-WM method by Fraunhofer IPK, which is based on a comprehensive enterprise and process modelling (Mertins et al., 2000).
• The KODA method by IMS GmbH, which is based on a communication analysis (Abecker et al., 2002).
• The KM extension of the ARIS business-process modelling method (Scheer, 2000).
• The CommonKADS enterprise, business process and knowledge analysis method (Schreiber et al., 2000), which comes with UML modelling tools and working sheets.
• The DECOR (Abecker et al., 2001) and PROMOTE (Woitsch et al., 2004) approaches that pursue a method-guided introduction of KM activities into business-processes, as well as the introduction of process-oriented content archives.

All these efforts have produced some method-driven system support for KM-oriented analysis and modelling. They all tend to start from business processes rather than from a dedicated KM strategy as Know-Net does. In addition, most have a strong focus on business processes and knowledge as content, but tend to neglect project work and process-independent knowledge networks in the organisation (i.e., the socialisation KM perspective). The method outlined in this article provides practical guidelines for methodologically implementing a KM initiative that fuses the socialisation and codification approaches.

An innovative feature of the Know-Net software system is its use of KOs as the central unifying element. For the moment, we have integrated ontology-based annotation and retrieval tools with IBM/Lotus Domino and, as a proof of concept, we have shown that our tools can be seamlessly integrated into Domino applications. There exist many other groupware systems like BSCW (www.bscw.de), which provide similar functionalities. Integrating the Know-Net system into other platforms is a straightforward task, as only the insertion of a hyperlink into the content-creation form is required.

The case of MC-Co demonstrates how both the socialisation approach (i.e., dealing with Thematic Area Networks) and the codification approach (i.e., dealing with knowledge as an object that is consumed or produced during consulting projects) can interoperate in practice. For instance, one can see how knowledge developed during a consulting project by a specific team can be codified so that it is accessible, not only by another consultant in the same or another project team, but also by a subject matter expert, or a member of a Thematic Area Network. Similarly, a best practice or even a simple contribution in the discussion database of a Thematic Area Network can be retrieved and utilised by a team engaged on a relevant project.

Another innovative characteristic of the Know-Net system refers to the ontology-based indexing and retrieval approach, which promises power for complex search as well as extensibility for more intelligent search algorithms that use, for example, query expansion techniques (see McGuinness, 1998; Guarino et al., 1999; Paralic & Kostial, 2003). Practical questions related to such a search approach concern such issues as the effort for manual indexing, or the roles and responsibilities for ontology creation and maintenance. There is much ongoing research on such issues, for example regarding:

• ontology creation driven by business-process analysis (see Jansweijer et al., 2000; Abecker et al., 2001);
• ontology creation supported by text analysis (see, e.g., Staab et al., 2000; Maedche, 2002); and
• automated classification of text documents in order to widen the metadata annotation bottleneck (see Mladenic, 1999; Aas & Eikvil, 2001; Paralic & Bednar, 2003).

Today's well-developed, so-called KM tools are usually web-enabled content management systems dealing with manifold types of document storage formats. They are often coupled with some agent technology for personalised push services and automated Internet search; usually, they have sophisticated algorithms for extracting semantic content from text documents in order to allow for high-precision information retrieval (see Maier, 2004). These products offer excellent services for what they promise, such as portals, content management or text mining. Our approach and system offers value-added services in the sense of complementary, easy-to-integrate knowledge organisation and retrieval facilities. These facilities grant decision makers seamless access to the knowledge residing in various IT platforms and tools used either for collaboration or data storage purposes. As such, they foster the harmonisation of the codification and socialisation KM approaches. Because the business processes of most organisations are already embedded in existing information systems utilising technologies such as ERPs, knowledge-based decision making should be tightly coupled with existing systems (Tauber & Schwartz, 2006). Our approach maps out a way of easily and methodologically integrating existing systems into KM processes. It thus allows for the timely and contextual provision of knowledge to decision makers, while combining the different types of data and knowledge that reside in various forms and systems in an organisation.

References

1. Aas K and Eikvil L (2001) Text categorisation: a survey technical report 941. Norwegian Computing Center.
2. Abecker A, Bernardi A, Hinkelmann K, Kühn O and Sintek M (1998) Towards a technology for organizational memories. IEEE Intelligent Systems 13(3), 40–48. | Article |
3. Abecker A, Bernardi A, Ntioudis S, Herterich R, Houy C, Legal M, Mentzas G and Müller S (2001) The DECOR toolbox for workflow-embedded organizational memory access. In Proceedings of ICEIS 2001 Third International Conference on Enterprise Information Systems, Setúbal, Portugal, July 7–10, 2001, pp 225–232.
4. Abecker A, Hinkelmann K, Maus H and Müller HJ (2002) Geschäftsprozessorientiertes Wissensmanagement. Springer-Verlag, Berlin.
5. Ahn HJ, Lee HJ, Cho K and Park SJ (2005) Utilizing knowledge context in virtual collaborative work. Decision Support Systems 39(4), 563–582. | Article |
6. Apostolou D, Papailiou N and Mentzas GN (2007) Developing knowledge networks: a practical methodology and experiences from cases. International Journal of Knowledge Management Studies 3–4(1), 330–355.
7. Bolloju N, Khalifa M and Turban E (2002) Integrating knowledge management into enterprise environments for the next generation decision support. Decision Support Systems 33, 163–176. | Article |
8. Chen H, Schroeder J, Hauck RV, Ridgeway L, Atabakhsh H, Gupta H, Boarman C, Rasmussen K and Clements AW (2002) COPLINK connect: information and knowledge management for law enforcement. Decision Support Systems 34, 271–285. | Article |
9. Chen Tan H, Carrillo P, Anumba C, Kamara JM, Bouchlaghem D and Udeaja C (2006) Live capture and reuse of project knowledge in construction organisations. Knowledge Management Research & Practice 4, 149–161. | Article |
10. Courtney JF (2001) Decision making and knowledge management in inquiring organizations: toward a new decision-making paradigm for DSS. Decision Support Systems 31, 17–38. | Article |
11. de Bruin J, Martin-Recuerda M, Manov D and Ehrig M (2004) State-of-the-art survey on ontology merging and aligning. SEKT Deliverable D4.2.1, http://www.sekt-project.org/rd/deliverables/wp04/sekt-d-4-2-1-Mediation-survey-final.pdf.
12. Dou D, McDermott D and Qi P (2004) Ontology translation on the semantic web. LNCS Journal of Data Semantics, Volume II, Lecture Notes in Computer Science, Vol. 3360 pp 35–57, Springer, Berlin.
13. Guarino N, Masolo C and Vetere G (1999) Ontoseek, content-based access to the web. IEEE Intelligent Systems 14(3), 70–80. | Article |
14. Hansen MT, Nohria N and Tierney T (1999) What's your strategy for managing knowledge? Harvard Business Review 77(2), 107–116. | PubMed |
15. Holsapple CW (2001) Knowledge management support of decision making. Decision Support Systems 31, 1–3. | Article |
16. Holsapple CW and Joshi KD (2001) Organizational knowledge resources. Decision Support Systems 31, 39–54. | Article |
17. Jansweijer W, van de Stadt E, van Lieshout J and Breuker J (2000) Knowledgeable information brokering. In E-business: Key Issues, Applications and Technologies (STANFORD-SMITH B and KIDD PT, Eds), pp 402–408, IOS Press, Amsterdam.
18. Kankanhalli A, Tanudidjaja F, Sutanto J and Tan BCY (2003) The role of IT in successful knowledge management initiatives. Communications of the ACM 46(9), 69–73. | Article |
19. Kingston J and Macintosh A (2000) Knowledge management through multi-perspective modelling: representing and distributing organisational memory. Knowledge-Based Systems. 13(1–3), 121–131. | Article |
20. Kühn O and Abecker A (1999) Corporate memories for knowledge management in industrial practice: prospects and challenges. Journal of Universal Computer Science 3(8), 929–954.
21. Kwan MM and Balasubramanian P (2003) KnowledgeScope: managing knowledge in context. Decision Support Systems 35, 467–486. | Article |
22. Lehner F (2000) Organisational Memory. Hanser Verlag, Munich.
23. Maedche A (2002) Ontology Learning for the Semantic Web. Kluwer Academic Publishers, Dordrecht.
24. Maier R (2004) Knowledge Management Systems. Springer-Verlag, Berlin.
25. McGuinness DL (1998) Ontological Issues for Knowledge-Enhanced Search. Frontiers in Artificial Intelligence and Applications. IOS Press, Washington.
26. Mentzas GN, Apostolou D, Abecker A and Young R (2002) Knowledge Asset Networking: A Holistic Approach for Leveraging Corporate Knowledge. Springer-Verlag, Berlin.
27. Mertins K, Heisig P and Vorbeck J (2000) Knowledge Management: Best Practices in Europe. Springer-Verlag, Berlin.
28. Mladenic D (1999) Text-learning and related intelligent agents: a survey. IEEE Intelligent Systems 14(4), 44–54. | Article |
29. O'Leary D (1998) Using AI in knowledge management: knowledge bases and ontologies. IEEE Intelligent Systems 13(2), 34–39. | Article |
30. Paralic J and Bednar P (2003) A text mining for document annotation and ontology support. In Intelligent Systems at the Service of Mankind (Elmenreich W, Hachado JA and Rudas IJ, Eds), pp 237–248, Ubooks, Germany.
31. Paralic J and Kostial I (2003) Ontology-based information retrieval. In Proceedings of the 14th International Conference on Information and Intelligent Systems, IIS 2003, Varazdin Croatia, 23–28.
32. Pedersena MK and Larsen MH (2001) Distributed knowledge management based on product state models – the case of decision support in health care administration. Decision Support Systems 31, 139–158. | Article |
33. Ramesh B and Tiwana A (1999) Supporting collaborative process knowledge management in new product development teams. Decision Support Systems 27, 213–235. | Article |
34. Rubenstein-Montano B, Liebowitz J, Buchwalter J and McCaw D (2001) A systems thinking framework for knowledge management. Decision Support Systems 31, 5–16. | Article |
35. Scheer AW (2000) ARIS – Business Process Modeling, 3rd edn, Springer-Verlag, Berlin.
36. Schreiber AT, Akkermans H, Anjewierden AA, de Hoog R, Shadbolt NR, Van de Velde W and Wielinga BJ (2000) Knowledge Engineering and Management: The CommonKADS Methodology. MIT Press, Cambridge.
37. Sense AJ (2007) Stimulating situated learning within projects: personalizing the flow of knowledge. Knowledge Manage Research & Practice 5, 13–21. | Article |
38. Shaw MJ, Subramaniam C, Tan GW and Welge ME (2001) Knowledge management and data mining for marketing. Decision Support Systems 31, 127–137. | Article | ISI |
39. Spangler WE and Peters JM (2001) A model of distributed knowledge and action in complex systems. Decision Support Systems 31, 103–125. | Article |
40. Staab S, Maedche A, Nedellec C and Wiemer-Hastings P (2000) Report on the ECAI'2000 Workshop on Ontology Learning, Berlin, Germany.
41. Tauber D and Schwartz DG (2006) Integrating knowledge management with the systems analysis process. In Encyclopedia of Knowledge Management (SCHWARTZ DG, Ed), pp 259–266, Idea Group Reference, Hershey.
42. Woitsch R, Karagiannis D, Höfferer P and Teuchmann B (2004) Process oriented knowledge management: a service oriented approach. In Proceedings of the DEXA Workshops, (GALINDO F, TAKIZAWA H and TRAUNMUELLER R, Eds), pp 471–475, Springer, Zaragoza, Spain.

About the authors

Dimitris Apostolou is a Lecturer in the Informatics Department of the University of Piraeus. His research focuses on group support systems, knowledge-based decision support systems and knowledge management. He holds a Ph.D. from the National Technical University of Athens in Greece and an M.Sc. in Information Technology from University College London, England.

Andreas Abecker heads the 'Knowledge Management' team at Karlsruhe (Germany) University's technology transfer centre for ICT. His research interests include ontology-based systems, business-process-oriented knowledge management and semantic technologies. He holds a Ph.D. in Applied Computer Science from the University of Karlsruhe.

Gregoris Mentzas is a Professor of Information Management at the School of Electrical and Computer Engineering of the National Technical University of Athens (NTUA) in Greece and Director of the Information Management Unit (IMU) at the Institute of Communication and Computer Systems, Athens, Greece.