How Artificial Intelligence Can Increase the Acceptance of Business Intelligence and Analytics
Acceptance of Business Intelligence and Analytics
Without a doubt, acceptance is one of the key success factors for business intelligence and analytics (BIA). Besides evidence coming from many studies , it is obvious that BIA is not an end to itself, but exists to support decision making in businesses. Consequently, it is crucial that casual business users are aware of BIA and use it in their daily work.
Figure 1. Technology Acceptance Model 
A common and well-proven approach to acceptance of technology is the Technology Acceptance Model (TAM)  shown in Figure 1. Concisely said, it states that the “Perceived Ease of Use” and the “Perceived Usefulness” determine the actual use of a system, respectively its acceptance. In the field of BIA, ease of use has always been a painful point, since casual users often depend on the IT department to get information or have to use unresponsive and complicated systems. Additionally, the perceived usefulness often suffers when received information from a system is inconsistent, incomplete or in wrong granularity.
Self-Service BIA as a Silver Bullet
Self-service BIA promises a remedy for these problems by empowering casual business users to create their own reports, analytics and dashboards. In fact, after a rusty start with complicated tools and halfhearted business initiatives, self-service BIA comes more and more in the focus of companies thanks to new technologies and a greater awareness for data in the business.
Explained by the means of the TAM model, self-service improves the ease of use with simple tools that enable casual users to find the information they need - when they need it. Thereby, it also increases the perceived usefulness and eventually the acceptance of a system.
In spite of all the advantages, self-service also introduces several new challenges. For instance, it increases the risk of conflicting reports, data inconsistencies and data silos caused by non-trained users selecting, transforming and visualizing data according to their means and wishes. Consequently, a self-service approach should go hand in hand with a thought-out architectural and governance concept, as well as a proper training .
Another major challenge in the self-service world is to find one’s way in the large data sets of today’s business. With more and more data coming from various data sources, it becomes more difficult to select the right data, transform it in a reasonable way and gain meaningful insights. Admittedly, this applies to casual users as well as to BIA professionals. However, over-strained business users and wrongly processed data sets can put the newly acquired BIA acceptance at risk.
Artificial Intelligence on the rise
Artificial Intelligence (AI) has gained a lot of attention due to its popularity in the startup sector. Beyond the hype, there are impressive technological advancements and new platforms  that make AI accessible for a broader audience and easily adaptable to many new use cases.
The goal of artificial intelligence is to enrich machines with cognitive capabilities to create an intelligent agent that perceives its environment and takes actions that maximize the chance of success at some goal . For this, AI encompasses various concepts from machine learning, natural language parsing and many other fields.
The BIA community discusses AI mainly in the context of big data, data discovery and other innovative analytical scenarios, e.g. machine learning for predictive analytics. However, with a broader understanding, AI can support in many other areas as well, e.g. increasing BIA acceptance by making self-service systems more “intelligent”.
Utilize AI to improve Acceptance of Self-Service BIA Systems
This section discusses how AI can positively affect the ease of use and the perceived usefulness of a system in order to increase its acceptance in the business. Particularly, it illustrates how an automatic recommendation engine can do that.
Recommendation engines that predict the preference that a customer would give to an item are well known from online shopping and many other digital services. There are two common approaches to calculate individual recommendations. The first one is called collaborative filtering and uses an individual’s (and similar individuals’) behavior in the past to derive preferences, e.g. other people who watched this movie also liked. Another idea is content-based filtering that uses characteristics of items to recommend items with similar characteristics, e.g. you liked a romantic movie so you are likely to like other romantic movies. Implementations in practice usually combine these approaches to get recommendations that are more relevant.
The goal of a BIA recommendation engine is to select the right information at the right time and granularity for an individual user. For example, if a user from marketing opens his self-service BIA suite, there will automatically be queries, dashboards and diagrams that are relevant for his current tasks. In contrast, if a user from another department opens the same suite, the dashboard may be completely different depending on his context and business role. So rather than a BI developer having to gather requirements and create tailored dashboard environments, AI is doing this automatically.
Such a recommendation concept obviously improves the ease of use of a system because it pre-selects relevant information and thereby accelerates the analytical process and helps to avoid an information overflow. At the same time, it increases the perceived usefulness of a system because, if it is working correctly, the system gives useful information in the right form and possibly even more, e.g. information that a user would not have searched for .
Figure 2. Exemplary scope of an intelligent information recommendation engine
Figure 2 sketches dimensions that a BIA recommendation engine should consider to give meaningful recommendations. An obvious starting point is the individual profile of a user that on the one hand includes his business role, current projects and other personal information. On the other hand, this also encompasses past behavior (e.g. what did he or she search for? How did he or she filter or visualize data?). According to the collaborative filtering approach, recommendations become better by also considering the actions of similar individuals. Consequently, the professional environment is another relevant dimension (e.g. what did co-worker search for?). Furthermore, external data may also improve the quality of recommendations. For illustration, think of logistics, where depending on the current weather condition, different reports may become more or less relevant. Lastly, metadata is essential to create relations between data items, queries and dashboards and thereby is the actual heart of a working recommendation engine.
Admittedly, the implementation of such a concept requires an underlying BIA architecture that allows the integration of data throughout an entire company, adequate governance structures and most importantly a sophisticated metadata management. Nonetheless, if implemented right, such a recommendation engine can change the way casual users interact with BIA systems and make analytics an ease for them.
However, an often-discussed risk of automatic recommendations is that users may end up in an information bubble and miss certain information or forget to question provided data. This is somehow contrary to the self-service idea and the reason why such a recommendation concept should first give a holistic view of information and, second, provide a way for users to dive into data and understand its lineage.
This article briefly illustrated how self-service and AI can improve the perceived ease and usefulness of systems in order to increase their acceptance in the business. For that, it introduced the idea of a BIA recommendation engine that automatically selects, transforms and visualizes the right information for every individual business user and thereby makes BIA systems more accessible and raises their acceptance.
The idea of intelligent recommendations in BIA is not completely new. Many BIA systems already encompass components that suggest certain visualization types or algorithms depending on data sets. Furthermore, there are business-wide data catalog systems or metadata search engines that try to help users find the right data in a search engine manner .
Another interesting approach are natural language generation (NLG) tools that use AI to automatically analyze data and provide an English language translation of what is significant and meaningful in the data . This additionally improves the ease of use of BIA and is another step towards the ultimate goal, where virtual AI-based business assistants automatically provide all necessary information for business users at the right time and the right form with a minimum of user interaction.
 Hawking, P., Sellitto, C. (2010): Business Intelligence (BI) critical success factors.
 Davis, F. D. (1986): A technology acceptance model for empirically testing new end-user information systems: Theory and results (Doctoral dissertation, Massachusetts Institute of Technology).
 Eckerson, W., Delvin, B. (2016): A Reference Architecture for Self-Service Analytics.
 Russell, S. J., & Norvig, P. (2010). Artificial Intelligence (A Modern Approach).
 To name a few: https://www.tensorflow.org/, https://azure.microsoft.com/services/cognitive-services/, www.ibm.com/watson/
[6.1] Marti, Y-V. (1996): A Typology of Information Needs, in: Gilad, B, Herring, J. P: The Art and Science of Business Intelligence Analysis.
[6.2] Taylor, R. S. (1962): The process of asking questions.
 Eckerson, H. H (2017): NLG Tools Automate Analysis