About the Client
Our client is a leading software publisher, and their products are used by enterprises globally. Their product lines span software development, games, consumer electronics and operate many subsidiaries in different geographies around the world.
Client Challenge
With the advent of GenAI and how it accelerates and brings efficiencies and accuracy when delivering support, the client wanted to adopt a GenAI and gig support model to help customers with their issues. They wanted to see how they can customize their copilot tool with GenAI so that support becomes much faster, accurate and responsive. They were also looking to see how they can use a combination of GenAI and copilot fueled by the gig model to address scalability when delivering support.
Movate Solution
Movate’s team studied the client need and arrived at the following scope of work:
- Set up Gen AI content curation engine
- Implementation of the OnDemand platform and onboarding of Azure SMEs (Gig Experts) for Human validation of content curation
- Utilize Gig For Good and peer experts to handle queries within the system, ensuring secure access.
- Peer-reviewed content to train the Copilot libraries
Setting up the Gen AI content curation engine
The first step we did was to build a content curation engine. The base content was sourced from the various documentation available, from customer interactions, feedback, knowledge bases and other data sources of the client. We fed this content to an LLM to train the LLM to provide human readable and reasonable answers to the query prompts it receives. The model is powered by a type of machine learning method called GANs or Generative Adversarial Networks. GANs use powerful deep learning methods to uncover and understand patterns and make inferences around them. Using this knowledge they can produce a required output to an input question. Another process constantly checks for inconsistencies, incorrect output and makes constant changes to the model. This process continues till the outputs generated don’t have any blemish. We integrated the Gen AI engine with the DFM and other knowledge bases using API calls for end-to-end connectivity.
The output of the curation engine would be used by the copilot which would help agents deliver flawless support, increase their efficiencies and value.
Deploying the OnDemand platform
While the content delivery was getting polished and ready, our team identified and harnessed the expertise of Directly Ondemand and brought in a team of Azure experts to humanly validate the content generated by the GenAI platform. This check was important as these experts would be using this content to deliver support to returning customers and it is important to ensure consistency and reliability of the answers provided.
The content generated and fed in needs constant scrutiny from time to time, so we assembled a team of peer experts to ensure robust content verification. The peer reviewed content is used to train the copilot libraries. Over time, the system becomes more efficient as it can synthesize, understand intent and craft natural, human readable output that can help support agents become the best. The advantage using GEN AI and copilots is that copilots are now trained to take and handle basic interactions, queries and can accelerate support queues because it is able to connect information faster. This results in faster case closures, heightened agent response and better agent allocations.
Business Benefits
- Enhanced accuracy and error reduction in Copilot solutions
- Elevated response quality
- Seamless scalability to manage growing workloads and adapt to changes