Introduction
Federal Bank, a leading Indian Private sector bank, was grappling with a problem found in most financial institutions – too much data, too little time to search for relevant information. Every time any associate wanted some facts or figures, they had to go through several documents – some of them containing only legal terms, which made the task arduous. When Federal Bank approached Cogniwide, they wanted–
1. A simple interactive tool where their agents could put forth their queries and one that could process natural language
2. A platform that could 'learn' to 'understand' human speech and 'find' relevant data without any mix-ups
3. an intelligent solution that could traverse thousands of documents in the blink of an eye and identify the appropriate document
The Story of Federal Bank
A bank with more than 1200 branches across India would have too many routine tasks that would take up the bulk of their staff's time, and that's the problem Federal Bank faced when they approached Cogniwide. Federal Bank understood that to ramp up their service and to ensure that their employees did not lose motivation, they would have to resolve this issue at the root. The best solution would be an intelligent virtual assistant that could take over the mundane parts of their job – such as answering queries or searching inside documents.These were time-consuming and mind-numbing jobs on which their associates would spend precious hours which could have otherwise been used in productive ventures. Additionally, a solution that could provide answers to questions in minutes or even seconds, instead of manually searching answers for hours together, would reflect the bank's approach to business – a pioneer in digital banking solutions as this was the first bank in India to digitalize all its branches in the country. Therefore, Federal Bank decided to install CogniAssist– a virtual assistant that would serve as a window for natural language conversations along with CogniDiscovery – an intelligent document search platform that would support CogniAssist in its quest for answers.
The challenge before CogniAssist
Taking up a considerable chunk of the associates' time and requiring a substantial investment in terms of money and efforts in training the associates to fetch relevant data according to the situation, this mundane yet necessary task was becoming an inconvenience for their staff. Yet, this was an integral part of their smooth and efficient operations. The established rules and laws form the organization's backbone as they help resolve any ethical quandaries or disputes that may arise as a part of everyday operations. Hence, all the information had to be at their associates' fingertips as only that would help in servicing their customers faster while ensuring that their employees did not lose their interest in the job due to the mundanities.
Why Federal Bank Chose CogniAssist and CogniDiscovery
Most solutions by major organizations only offered cloud-hosted solutions, while Federal Bank wanted an on-premise installation to have more control. As very few players in the market could take up this challenge, and Cogniwide offered one of the best solutions, Federal Bank decided to go ahead with CogniAssist.
Additionally, being a large organization, Federal Bank had several unstructured documents that needed to be traversed by the AI platform. Learning to 'understand' unstructured documents is no mean feat for an AI solution, yet our product could do so without any glitch. Hence, Federal Bank decided to go ahead with
• CogniAssist – an interface for fetching information according to the situation using natural language
• CogniDiscovery – for intelligent document search that could 'learn' to interpret unstructured documents
• CogniNaas – as the NLP backbone
Additionally, they also incorporated a workflow builder tool from Cogniwide to build banking workflows by themselves. This is a user-friendly platform that requires very little training and one over which they had complete control as it does not require third-party intervention.
To summarize, Federal Bank chose solutions from Cogniwide due to the following reasons –
• On-premise installation
• A user-friendly platform that does not require third-party intervention
• Enhance employee productivity by simplifying mundane yet time-consuming tasks
• Ability to traverse unstructured documents by understanding the structure of any document
• Increased data security
• Ease of integration with other applications
How Federal Bank used CogniAssist to take the drain out of routine tasks
Banks operate on laws, and while it is commendable that experienced staff could fetch relevant data from the corners of their mind within seconds, it is not so easy for new hires. Furthermore, if a situation demanded a scarcely used law or rule, then quite possibly any staff would take hours to fetch the relevant information. However, in today's scenario, if a customer is kept waiting, chances are they might lose patience. Furthermore, one small negative review could catapult into a giant outburst on social media.
Large documents that contain vital information, such as interest rates or laws, need to be referenced often in banks to avoid inaccuracies or ethical conflicts. Hence, as a part of the training, their workforce had to be taught to reference documents and retrieve information. While this might seem simple, it is impossible to avoid inaccuracies due to human limitations. Furthermore, on busy days, searching for information or referencing documents would be time-consuming and further restrict the valuable time that could be used to service all their customers. Hence, when Federal Bank approached Cogniwide, their main intention was to install a platform that required minimal intervention from their staff or third-party vendors and was easy to use.
CogniAssist checked all the right boxes. It could talk in natural language to their assistants and could be taught to understand the situation and fetch relevant data. The initial proof of concept was highly successful, after which Federal Bank wanted to integrate the solution with a document discovery platform through which their associates could search inside any documents and get the relevant data as paragraphs on their interface. Hence, CogniAssist was combined with CogniDiscovery, which further quickened the results.
After the on-premise installation, all the documents were uploaded to the platform. Once the platform was 'trained' to understand documents with a few sample files, it had its own 'knowledge base.' After this, when any query was put to the semantic search platform, it brought up the most accurate hits. The agent could then pick the appropriate document pertaining to their situation. This was because the question is first put to the chatbot knowledge repository, and if any matching results were available, it would pop up on the screen as paragraphs. The user could click on the relevant text to view the complete document. If there were no matching results, the query would be rerouted to the CogniDiscovery platform to fetch the top five matching results.
Additionally, several agents could know the names of the documents they were searching for. Hence, the solution was tweaked to support deterministic searches. What used to take several minutes or even hours before had been brought down to a few seconds.
With more usage, newer scenarios to enhance the solution were identified. For instance, document tagging was enabled, which are keywords that could further improve the swiftness of the searches. Once a match is found, the relevant information and critical components such as titles, subtitles, and tables are highlighted to help agents skim the documents before zeroing in on the appropriate segment. Keyword extractions and tagging according to concepts were also added as part of further enhancements. Additionally, CogniNaas, the brain behind the NLP, could 'teach' CogniDiscovery to 'learn' from existing documents.
As more and more situations for automation were recognized, Federal Bank wanted a workflow builder that could teach the intelligent virtual assistant to respond to newer scenarios identified for automation. This was a low-code/no-code solution that required no prior coding knowledge, and their staff could learn to use it with minimal training and create workflows by themselves with no intervention from Cogniwide.
Highlights:
• Intelligent searching
• User-friendly tooling that requires zero coding skills to configure
• Easy integration with chatbots
• Understands unstructured documents
• Delivers passages as answers to natural language queries
• NLP feature that can teach CogniDiscovery to learn and understand documents
The Results
The response has been utterly positive as the bank staff has comfortably adapted to the platform. Furthermore, the significant reduction in document search time and the ease of querying for information has increased employee engagement.
The Way Ahead
The ease of usage of the intelligent virtual assistant and document discovery platform has encouraged the organization to purchase licenses from Cogniwide for building their own chatbot. Our team is enthusiastic about supporting them on this transition as we would like nothing more than watching more people enjoy the benefits of Artificial-Intelligence powered platforms to do their routine tasks and thereby get more time for other productive pursuits.
