AI based Natural language Enterprise Search for Asset Management Companies

Machine learning brings us a step closer to humanizing responses for user queries. The search interface available on traditional platforms, websites does not meet the speedy, accurate and personalized results. Moreover, the search becomes a lot more cumbersome when it comes to enterprise. The various data sources, structure and legacy applications hamper internal searches.

Current challenges with Asset Management Companies (AMC):

The asset management industry is at a critical juncture in its history. After a decade in which fee compression and the growth of index investing have driven sweeping change, these shifting economics are now converging with the relentless advance of regulation and disruptive technological breakthroughs. In response, asset managers are investing in innovation and reinvigorating their products and processes. Industry pundits point to a set of six challenges asset managers need to address:

1. The shift from active to passive investing
2. Demographic changes
3. Shifts in demand
4. Changing regulations

In big house AMC’s, most enterprise data are unstructured. such data manifests as a huge variety and volume of documents making it difficult to search and extract such information needed by different departments such as marketing, sales, compliance & auditing and finance. Most of this unstructured data is spread across organizational silos, to search among them by traditional methods with keyword-based search systems that require business users  figuring out the right combination of keywords, and usually return many hits, most of them irrelevant to our query. Huge time spent on information search negatively impacts business productivity, ability to quickly take decisions and growth.

Asset Management Company Use Case

The use case pertains to searching specific business information from myriad mutual fund offerings and terms documents available in the public domain of a leading AMC. An ability to quickly search important information pertaining to various type of fund specifications and investment terms can help improve marketing agility and improve customer service. When we search for sections of questions in these documents using traditional search, it was time consuming, and the answers were available after tedious research. The objective was therefore to enable an easy and faster way to search the MF documents with precise answers by consolidating document data in a central location.

Increasing data Findability & Transparency

With large AMC catalogues and pricing specific to the mutual fund business accounts, many AMCs find it challenging to offer the right products to the right person at the right time while meeting growing buyer expectations of transparency, immediacy, and convenience. On top of these customers, AMC advisors from competitors armed with the right information keeps your sales team out of the loop until you realize it late in the customer journey.

AI based enterprise information search removes manual labour and automatically optimizes the search experience though human centric interactions so AMC teams can focus on key business problems and effectively serve customers who really need them.

Enabling Natural language based, Enterprise AI search

We used Amazon Kendra to address the use case requirements that can bring paradigm shift in the way AMCs can search data in real time to make smarter decisions.

Amazon Kendra is an enterprise AI search service that enables users to intuitively search unstructured data using natural language. It is highly available and scalable, tightly integrated with other AWS services such as S3 that can store massive amounts of unstructured data and offers an enterprise-grade security.

Amazon Kendra uses deep learning models to understand natural and complex language queries and document content and structures.

Key features and benefits of Kendra based AI search solution includes the following,

• Accuracy – Unlike traditional search services that use keyword searches where results are based on basic keyword matching and ranking, Amazon Kendra attempts to understand the content, the user context, and the question and returns the most relevant word, snippet, or document for your query.
• Simplicity – Provides a console and API for managing the documents that user want to search. It easy to integrate Amazon Kendra into your client applications, such as websites or mobile applications.
• Connectivity – Connect to third-party data sources to provide search across documents managed in different environments.
• Faster search – Faster indexing and improved accuracy in search
• Access Control List – Amazon Kendra also helps to ensure that search results adhere to existing document access policies by scanning permissions on documents, so that search results only contain documents for which the user has permission to access.

Setting up Amazon Kendra included the following activities

1. Creating an index
2. Add unstructured data sources
3. Test search in Natural language and deploy

Creating an Index

On the Amazon Kendra console, we created an index with the Amazon Resource Name (ARN) of an IAM role that has permissions to Amazon S3 bucket.

Add data sources

Amazon Kendra provides a way to add data sources from multiple sources including from Microsoft SharePoint and Salesforce. It provides a flexible schedule which syncs data periodically or on demand.  We loaded all the mutual fund brochures and offering documents and Mutual fund FAQs prepared with answers were also updated as a Comma separated file in Amazon S3 as the unstructured data for search.

Natural language-based search testing and Deployment

After the data was indexed in the Kendra service, we were able to test and tune the search experience directly in the Kendra search console. For example, in the silos of brochures of mutual funds and term documents, we searched for the fund manager of a specific mutual fund (Sun Life Frontline Equity). The manual search process was quite complex and time consuming. Meanwhile, Amazon Kendra returns the results of the search as shown below.

Further, Amazon Kendra provides a high-availability service for production workloads.
that helps adding more capacity for query-based or storage-based requirements.

The Kendra based natural language use case example provides a glimpse of how enterprises can leverage AI/ML based search that provides faster and accurate and context relevant results apart from the ease of use for any business users.

References:  https://aws.amazon.com/kendra/

Written by :   Dhivakar Sathya  Umashankar N  & Sankar Nagarajan