Solution

Quantiphi initially provided advisory services to this real estate company which enabled them to identify the issues pertaining to the virtual agent. We shared the results of the evaluation and areas of improvement for the company’s existing Virtual Agent.

Quantiphi recommended the next steps of implementation (prioritized high, medium, and low based on business impact) and concluded to resolve high and medium priority issues.

Implemented the voice bot using Dialogflow ES as per best practices recommended in Phase 1.

Adding 8 -10 variations of training phrases, the right use of entities and respective synonyms, handling custom fallback messages for each negative scenario & optimal use of context and its respective lifespan helped our client to reduce live agent transfers and improved user experience.

Correct use of nested flows during positive/ negative utterances & use of context-based flows helped the solution to scale for future use case enhancements.

Reduced execution time and memory with optimal file structuring and use of helper functions to execute them in different files.

Implemented Google-managed container base images which enabled updating for patches of common vulnerabilities. Using Secret Manager for fetching environment variables on Cloud Run helped us to securely store API keys and passwords.

Developed a virtual agent analytics dashboard to track the performance and improve the efficiency of the bot without compromising the customer experience.

Results

Automated 24*7 voice assistance by improving the accuracy and user experience of the virtual agent and reduced high live agent transfers after implementing high and medium priority issues.
Reduced live agent handovers and accurate reporting of data using the right entities with crisp and optimal synonyms. Scaled the solution for future use case enhancements with the correct use of nested flows during positive/ negative utterances and optimal use of context-based flows.
Scaled the solution for future use case enhancements with the correct use of nested flows during positive/ negative utterances and optimal use of context-based flows.
Supported users to complete the conversation and improved user experience with the optimal use of output context and its lifespan.
Improved security by implementing Google-managed container base images and Secret Manager for fetching environment variables.