How one CPG company’s example of a chatbot business application transformed their information search.
Business applications for Natural Language Processing are today’s hot topics. Examples of Natural Language Processing (NLP) include natural language generation (i.e. converting data into human-readable sentences and paragraphs), speech processing, and semantic search (answering written human-language questions, as opposed to SQL queries or something similar). But how can this be used in a business setting?
Well, few businesspeople will dispute the need for data-driven decision-making. And few large organizations have a shortage of information. But uncovering the insights within the data? That can be tricky. Delivering them on time can also be difficult.
One of our clients faced this very problem. The hero of their particular story turned out to be an NLP-powered chatbot that allowed users to speak or type questions in their normal, everyday language and get an instant answer.
The Problem: Disorganized Data and Long Waits
First, a little background. The client, a leading CPG manufacturer based in the Middle East, had an enormous amount of data: historical data, third-party reports, research data, marketing data, and so on. The initial problem was that this data was literally scattered throughout the company. But once the data was centralized and organized, a key question remained: How could users get actionable insights directly, without going through some other employee or department?
It was important to the client that any employee – not just those with tech skills – be able to access the information in their data library.
Clearly, the chosen solution would have to have a very simple interface – asking employees to follow strict search rules (or worse, learn a querying language like SQL) was out. And the solution would also need to be available in a mobile version. For these reasons, we decided to implement two of today’s buzziest technologies: NLP and chatbots.
And with that decision, this became more than just another project. It became a case study for NLP and chatbots in the business world.
An Example of Chatbots and NLP in Business Applications
Chatbots are definitely trending right now. Interestingly, the 2018 State of Chatbots report revealed that 69% of respondents felt that chatbots excel at giving quick answers to simple questions, and 38% felt chatbots provided quality answers to complex questions. This was exactly the functionality we wanted.
Further, employees needed to be able to simply type or speak common questions, such as:
- What is the profit margin on Product X?
- How did Product Y perform in 2018?
- Who is the largest consumer of Product Z?
One popular example of Natural Language Processing is breaking down users’ written or verbal questions and figuring out what the user wants to know. To do this in our client’s application, the AI back-end had to instantly search every piece of information in the data library, find the most recent and relevant item, and send it back to the users. The answer is supplied in two forms: a summarized reply (e.g. “Product X’s profit margin in 2018 Q4 was …”) and a shareable, presentation-ready version of the source.
If a specific report (or passage from a report) is requested, the chatbot can return the relevant data on a very detailed level – a definite convenience when you’re dealing with 400-page-plus documents! And the Android and iOS apps included in the project made this NLP search bot available wherever needed – at a sales call, during an offsite meeting, or anywhere the user happened to be.
Of course, no system performs perfectly from the start: anyone who’s used predictive text keyboards knows that! So, we built self-learning capabilities into the system. Users can give the chatbot direct feedback on how well it performed on their question, and they can also modify their question to get a more exact result. These instances are continuously fed into the underlying AI components, helping the system become more accurate as it matures.
Natural Language Processing Applications and The Future of Data-Driven Decisions
Thanks to this application of NLP, our client’s business has become more agile. Its employees spend much less time manually searching for information, and insights are being delivered far more promptly.
While several different technologies were used to accomplish this, NLP was the game-changer. Its ability to process plain-language queries enables every employee to find data-backed answers. This reduces potential bottlenecks and lessens the demand on the client’s research teams. It also makes finding insights almost as personal and instantaneous as opening an email.
We confidently expect Natural Language Processing’s applications in the business and marketing world to grow. As this case shows, it has the potential to transform how and when data is delivered – and that has the potential to transform how businesses perform in today’s markets.
To read how this solution was developed and technologically implemented for the client, please read this case study by IBM.