By using website you agree to our use of cookies as described in our cookie policy. Learn More

TDWI Upside - Where Data Means Business

Alexa, Will Chatbots Be the Future of BI?

Chatbots are all the rage because of their easy-to-use interface for end users. Several tech giants have opened platforms that allow companies across all industries to take advantage of this technology as a front end for their business intelligence systems.

From Apple's Siri to Amazon's Echo to Google's Google Home, major technology companies continue to release products that replace the traditional user interface with a voice-activated, natural language interface. The reason is simple; this interface is easy for users to operate and allows these products to be seamlessly integrated into everyday conversation. If you need an answer, it's as simple as asking, and the results are returned in the voice of a nice digital assistant.

The next step is the democratization of this technology so it can be used to build all types of interfaces. Both Google and Facebook have led efforts to do this. With Facebook's release of and Google's acquisition and public release of, developers across multiple industries now can overlay their apps with a natural language answering interface. Apps built with these natural language interfaces are becoming commonly known as chatbots.

When you look at the most obvious use cases for natural language interfaces, business intelligence jumps right to the top. These interfaces are optimized for answering questions and business intelligence is fundamentally about providing information to answer questions.

Currently much of the information being returned by these products is general in nature, such as the weather forecast or the current news, but imagine the possibilities if these same assistants were able to return key pieces of information about your business and the data that is flowing through its systems. Now all your investment in data analysis can pay off with a new way to deliver crucial bits of wisdom to key executives.

Putting Together a Natural Language Interface

Building a good conversational interface goes beyond merely creating a voice input into your business intelligence system. It requires that the interface convert the audio presented by the user into a set of what are known as intents and entities.

An intent is the target action of the string of words provided to the interface. This is what the user wants to know or do. Entities are the objects and attributes about the action that provide additional context. These are what provide an engine to filter down the universe of data to a subset on which the intent will act.

Both and are available for developers to start integrating with key chat technologies that users interact with every day. By simply wiring up these interfaces to your business intelligence system, you can create a natural language interface on top of your company's data and expose it to key decision makers.

There are four key steps that you need to take: train the chatbot, refine its understanding of entities and intents, connect to the data, and connect to the user interface.


The first step with and is to develop a set of intents. These are the actions that the chatbot needs to anticipate and can handle. These would be activities such as get sales or get inventory levels.

An easy method to identify these intents and entities is to compile a list of the questions that users commonly ask your business intelligence team. With this list, you can start to pull together groups of activities that the users want to accomplish and the tasks that the chatbot will do.

The interface for each platform allows you to enter these questions directly in and it will try to associate the question with the intent and entities that you have defined. This provides the initial set of training for the artificially intelligent parsing engine.


Once you have entered the questions and defined the entities and intents, the next step is to help the engine understand different synonyms for the entities. This helps the engine recognize patterns in the questions and create an abstraction model that can respond to variations of these questions that users will ask. The engine does not have to be an expert in all industries because it works on this abstraction model you refine for your specific needs.

The goal of refinement is to try many ways a user would ask the question and make sure that the identified intents and entities parsed from that question are correct. This will increase the potential for the engine to recognize previously unseen questions.

Connect to Data

As a cloud-based product, these chatbots have no access to your data until you make it available. Their function is not to query your databases directly but to parse the question to a single intent and associated entities. From here, you develop a webhook interface into your systems that receives these parameters.

The processing remains completely within your environment and leverages your understanding of the data and the problems your business faces. Once the processing is complete, an answer is returned to the chatbot, which in turn sends it back to the interface.

Connect to Interface

Finally, you will want to open up this interface so it can receive input and provide output to your users. and have multiple integrations, such as SMS, Facebook Messenger, Slack, or Amazon Alexa, that are ready to be connected to your accounts. From here, a user initiates a conversation with your bot through these common interfaces.

Working through these four steps will allow you to layer a natural language on top of your business intelligence systems and provide users a whole new way to get answers to their questions.

By making a relatively easy path from user to data and back again, these platforms bring the power of natural language answering to companies across multiple industries. No longer is it only the largest tech companies with armies of data scientists that can enable their users with this technology. You can add it on top of your existing business intelligence solution to provide quick and easy answers to key decision makers.

About the Author

Troy Hiltbrand is the senior vice president of digital product management and analytics at where he is responsible for its enterprise analytics and digital product strategy. You can reach the author via email.

TDWI Membership

Accelerate Your Projects,
and Your Career

TDWI Members have access to exclusive research reports, publications, communities and training.

Individual, Student, and Team memberships available.