Two very interesting and seemingly contradictory trends are taking place in the information landscape. On the one hand, technologies are evolving and changing at a blistering pace, with venture capital pouring into startups at unprecedented rates. On the other hand, organizations are still struggling to solve prosaic yet seemingly intractable problems around information access and management. How can these two phenomena exist side by side?
In the more mundane world, the same information management problems that organizations have been dealing with for years continue to plague every department. Accounting does not get its reports on time and can’t closeout its books by the deadline. Marketing does not get timely feedback on promotions and customer acquisition programs. And engineering workflows are limited by bottlenecks in data flows. E-commerce managers lack control of large product catalogs, and routine search and information retrieval still present day-to-day challenges.
In some cases, bottlenecks are due to manual processes and incompatible data formats. In others, the different systems are owned by different groups with no overall ownership for data quality or supporting processes. An upstream process may be causing a problem by not correctly managing data input for example, but those impacted downstream do not have the authority for making changes to improve that process. All of these issues can affect routine processes, and would also interfere with more sophisticated applications that a company might want to develop.
We see the shining light on the horizon of what could be and are still dealing with the day-to-day problems of finding reports or the latest plan, getting the answers to customer questions or finding the latest proposal to reuse for a new project.
How do we go from what is practical to what is possible?
I was speaking with a customer who reached out to us after reviewing a consultant’s report that outlined multiple Artificial Intelligence (AI) projects to address the company’s information management challenges. The report was quite comprehensive and well researched, well written, and spot-on with regard to approach and recommendations. But the report neglected to take into consideration the foundational requirements around data quality and structure that would be required for such an approach to be successfully deployed.
One recommendation from the report was that the company begin using chatbots. A chatbot can be an excellent interface to content and data to enable self-service and reduce support costs, but that information needs to be available in a format that can be accessed with the bot. The customer who had contracted for the consultant’s report was beginning to realize that although the guidance was good, the means to get there was lacking.
AI can be a terrific adjunct to existing technology infrastructure and customer support processes – but only if the underlying systems and processes are working. A chatbot is a channel to product data, marketing content, support knowledge and other corporate assets. However, if the data is of poor quality, missing or incorrect, machine learning and AI technologies will be limited in terms of what they can do.
AI can help make sense of large amounts of information such as that collected from customer data in call centers, or technical reference information, and add structure to unstructured content. However, it but does require a reference architecture and in some cases an ontology.
An ontology is a knowledge representation for the enterprise and consists of multiple taxonomies and the relationships between them.
In fact, organizations are increasingly leveraging AI with traditional information architectures to enable emerging technologies such as voice interfaces for ecommerce. Starbucks has an ordering app that uses a sophisticated content and data model to recognize the thousands of possible combinations that comprise its selection of beverages. The app works well, but the voice interface required that the underlying data be correctly structured so that the system recognizes a user’s request.
Voice search, conversational commerce and other “VUI” (voice user interfaces) are increasingly being deployed to leverage Alexa, Siri, and the incredible progress in speaker independent speech recognition that is in the palms of our hands with every smartphone. Voice interfaces are different from chat or traditional text interfaces. They require new ways of thinking about the user experience. However, they still leverage the data models needed to solve all types of findability challenges.
Just because your company has not yet solved the simpler problems of information management does not mean it can’t benefit from the innovative technologies of the future. But it does mean that some foundational work in organizing the information will be needed before either type of problem can be solved.