An Onerous Integration Task
But companies face both technical and managerial challenges in making CRM/KM integration work. In fact, the integration is often daunting. The reality is that these systems were not designed to interoperate, so moving information from one to the other can be difficult. The underlying data is complex, and custom application programming interfaces (APIs) are used to connect each data source. Usually, the process requires a significant amount of custom integration work.
How much work is needed depends on customer expectations, which can run the gamut, and the design of the KM system. The more grandiose the business' goals are, the more work is required. The number of information sources is a key consideration. How many external data sources will feed into the information system? Here, the more information collected, the bigger the challenge in sifting through the data. "With the popularity of products like Microsoft's SharePoint, large businesses find themselves with tens of thousands of departmental knowledge management systems," notes David Lloyd, CEO at IntelliResponse. Sorting, consolidating, and maintaining that information on an enterprise level represents a gargantuan undertaking.
Another difference revolves around the timeliness of the data. Do companies want their knowledge information to be made available in real time or in batch mode? Executives' knee-jerk reaction is to demand the former, but the latter is a more realistic option.
Knowledge Management Systems Ain't So Smart
Most KM systems rely on some type of search technology to pull relevant data from vast pools of information. But search sometimes is not an optimal approach. It presents agents or customers with a list of choices that they have to sift through rather than with the answer to their question. "An agent does not want to have to look through a ten-page PDF to find the one paragraph that tells [him] how to reset a browser," says Richard Simmons, CEO of Creative Virtual USA.
The KM system needs to put the corporate information into context and present the right data to the user. "Corporations are drowning in information, but there is a real thirst for knowledge," explains Diane Berry, senior vice president of marketing and communication at Coveo.
To quench that thirst, vendors have been developing various artificial intelligence engines that sift through data and pull out the relevant items. But these features present enterprises with more challenges. The technology is new, so it is far from foolproof. In some cases, businesses unknowingly become guinea pigs as vendors tweak the algorithms driving their solutions. In addition, the artificial intelligence features require a great deal of customization, because no two companies have the same knowledge bases or the same knowledge needs. Unlike traditional applications, such as a general ledger system, the data (knowledge) constantly changes. And unlike most application development cycles, the customization requirements continue as the KM solution moves forward into production rather than being a one-time task done early in the development phase. In sum, corporations have to allocate staff to tweaking the system throughout its lifetime.
Centralize or Decentralize?
The business also needs to decide how to house its data. Some corporations want to put all of the customer knowledge into one central file. "Companies are increasingly focusing on building a central repository mainly for customer service and marketing collaterals which can be accessed by sales teams," Sheth says.
But the one-size-fits-all approach can be vexing. Firms grapple with the question of "Who owns the data?" If one department is in charge, this group may tweak the data so it does not work with another department's applications. If no one is in charge, the data structure may lack direction.
Another option is to leave the information in various databases and rely on connectors to pull it out when needed. This approach can be simpler to deploy in the short term because all of the information does not need to be completely homogenized. However, it can be difficult to maintain as the solution moves forward. The enterprise may end up with various iterations of the data, running in different locations, which can lead to confusion.