Czech financial savings financial institution Česká spořitelna, a division of Austria’s Erste Group, just lately collaborated with AI answer builder DataSentics to discover using GenAI in name facilities. Česká needed to enhance high quality management and optimize prices of their inbound name middle operations, which obtain round 2 million calls per yr. They selected the Databricks Knowledge Intelligence Platform to experiment with each inside and exterior AI fashions to evaluate the effectiveness of name middle brokers.
Exploring a High quality Management System for Buyer Assist
The decision middle staff at Česká spořitelna needed to check a top quality management system powered by GenAI that may make sure that brokers adhere to scripted pointers throughout buyer interactions. A vital problem for Ceska was making certain constant agent communication for routine buyer inquiries. When clients name about account balances, brokers must direct them to on-line banking options, a key enterprise requirement that drives digital adoption and operational effectivity. The help staff wanted a scalable technique to confirm agent compliance and preserve communication requirements throughout hundreds of buyer interactions. To realize this, the staff started through the use of Whisper, a speech-to-text mannequin from OpenAI, to transcribe conversations precisely. The problem was to provide human-readable textual content that precisely represented spoken phrases utilized by name middle brokers with out distorting their that means. The transcriptions wanted to make logical sense and replicate the intent of the dialog precisely for additional evaluation.
Following the transcription, the staff explored integrating each inside GPT fashions and open supply fashions reminiscent of Mixtral to judge their effectiveness. GenAI fashions have been examined in a simulated QA function, the place they have been tasked with answering particular questions reminiscent of “Did the agent redirect the client to on-line banking?”. The objective of this train was to evaluate how properly these fashions might mimic human understanding and decision-making when verifying compliance with established pointers. By evaluating the efficiency of each the interior GPT mannequin and the open supply fashions, the staff aimed to search out the best answer for bettering customer support by way of automated AI-driven high quality management.
Advantages of the Databricks Knowledge Intelligence Platform for GenAI
The DataSentics staff evaluated a number of choices for this answer, and in the end selected to deploy the Databricks Knowledge Intelligence Platform and Mosaic AI instruments at Česká spořitelna for a number of causes:
- Knowledge Administration and Governance Advantages: Unity Catalog makes knowledge simply accessible for various fashions whereas retaining delicate knowledge beneath restricted entry.
- Complete Knowledge Processing Capabilities: the Databricks Platform helps your entire workflow of preprocessing of name middle knowledge, from transcription to high quality management. This allows us to provide intermediate outcomes that may be leveraged for different fashions and tasks, reminiscent of advertising, threat evaluation, regulatory compliance, and fraud detection.
- Mannequin Coaching and Assist: Databricks gives strong help and experience for GenAI, together with mannequin structure and coaching capabilities. This made it a great platform for testing and deploying open supply fashions rapidly, enabling us to experiment and iterate effectively.
- Ease of Cluster Creation: With Databricks, it’s simple to create clusters and deploy open-source fashions. This streamlines the experimentation course of and permits us to focus extra on mannequin efficiency and fewer on infrastructure administration.
Insights and Outcomes
All through the venture, we experimented with numerous segmentation methods and gathered a number of worthwhile insights:
- High quality of Enter Knowledge is Essential: The standard of the audio recordings diversified from shopper to shopper, with some talking quietly or from a distance, which may later have an effect on the accuracy of the transcription. Whisper or related programs may also help resolve the issue.
- Class Definition is a Should: We discovered that if classes can’t be simply outlined for people, it’s equally difficult for LLMs to grasp them. This strengthened the necessity for clear and exact class definitions to coach the fashions successfully.
- Open-Supply Fashions Ship Outcomes: Open-source fashions demonstrated that they might compete successfully with proprietary fashions like ChatGPT. This discovering is critical for companies trying to optimize prices whereas nonetheless reaching high-quality outcomes.
What’s Subsequent
With GenAI instruments powered by Databricks Mosaic AI, Česká spořitelna staff are actually capable of acquire entry to solutions present in a spread of paperwork by way of “sensible search” performance. For instance, the buying staff could must seek the advice of lots of of pages of course of documentation on how one can management and approve funds to completely different nations. Earlier than leveraging Databricks, it could take staff hours to search out the proper info they want. Now, RAG-powered search offers staff solutions inside seconds, together with citations and hyperlinks to the supply doc.
Wanting forward, there are many alternatives to discover extra GenAI workloads at Česká spořitelna. We intention to create a sturdy integration between Databricks and Česká spořitelna’s inside database name middle recordings. This can unlock new use instances reminiscent of churn detection, sentiment evaluation, and gross sales sign detection since Databricks is the go-to platform for streaming knowledge. These day by day experiences will permit Česká spořitelna to react to modifications in actual time whereas reaching price reductions with improved high quality assurance of their name facilities.
This weblog put up was collectively authored by Petra Starmanova (Česká spořitelna), Tereza Mokrenova (DataSentics), Dalibor Karásek (DataSentics) and Joannis Paul Schweres (Databricks).