Case Study Banner

CASE STUDY

Enhanced asset visibility and improved decision making through AI enabled conversational bot for enterprise digital asset management

industry-iconCLIENT :Confidential
industry-iconINDUSTRY :Pharmaceutical
industry-iconDURATION :4 months
CLIENT :
Confidential
INDUSTRY :
Pharmaceutical
DURATION :
4 months

Business case

A global bio-pharma digital operations team achieved improved decision-making, increased cost savings, and streamlined software licenses management through the implementation of multiple genAI enabling technologies while leveraging legacy CMDB and other bespoke asset management systems. Additional business benefits include improved transparency of operations and increased employee productivity.

Historically, the digital team relied on complex data extraction and labor-intensive manual processes to retrieve critical technical information related to application data, server details, ticket histories, support contacts and other relevant data needed to address time sensitive incidents. These incidents ranged from cyberattacks, unscheduled production outages, and even planned events. The legacy environment evolved across multiple technical eras, and the repository evolved with limitations to address datacenters, client server, cloud, and externally hosted environments.  

The CrowdStrike incident provoked the organization to transform how they engage with their source systems. Digital operations moved away from slow and complex data queries requiring manual SQL queries into the vast technology landscape. With over 30,000 applications in the organization, including more than 1,000 systems in Research & Development, teams were overwhelmed by repeated queries. 

Our solution

After a thorough needs analysis, i2e solution experts decided on building a chatbot based on the GPT 4.0 model.

This chatbot could translate natural language questions into SQL queries by fetching data from the internal application database created on Oracle. The chat bot was trained to act as a sequel assistant, so it was capable of converting natural language into SQL queries.

The output of the chatbot is expected to be in tabular form. To achieve this, the team created a knowledge based that acted as additional understanding to the LLM model to give the output in the desired format. This knowledge base also acted as a reference repository for the teams. The team achieved a 60 to 70 % accuracy while providing the option to keep a human oversight to the chatbot’s answers.

The users were familiar with the type of queries to be given, however, they were not sure of the prompts which can work for the chatbot, so additional training was also given which helped them acclimatize with the prompts. 

Challenges overcome

  • Developing a separate knowledge base or reference repository for application data for the chatbot to understand and answer queries. 

Benefits

  • Automated conversion of natural language to SQL queries
  • Detailed answers in the desired format for users with display name, reference name of columns
  • Saved time in writing SQL queries to fetch information
  • Easy information retrieval during crucial cyber attacks 

Results

Digital Asset Management Solution
RobotAsk your question
Please select an option to continue