The solution

Understand the current business needs and design a set of business use case, including a proof of concept and designs for the prototype. Assess technology options to support the solution and select/implement an advanced analytics solution in the Cloud for call transcription, sentiment analysis and LLM summarisations. Supported the team on using the Cloud platform, analytics solution and implementing custom-built models for analysing the call records.

Major Pain Points

• FTE count for auditing continues to grow, as customer numbers grow, increasing operating costs.

• Customer requests are becoming more complex and, on average, taking longer to process.

• Customer complaints are increasing.

• The cost of compliance continues to rise and decision-making support is all manual.

Outcome

LLMs predicted customer complaints below a human operator, but were able to determine vulnerable customers to an equivalent level as a human operator. LLMs were able to summarise large volumes of data much faster, and to a similar level of quality and accuracy as a human operator. Based on the state of GenAI, the approach taken proved not economically viable due to the overall operating costs of the platform and running the LLMs. Whilst there are many technology options available for supporting advanced analytics and integrating with LLMs, the few evaluated were comparable across providers.