We have built a solution for agencies which provide the caretaker services for parents who are in search of babysitters for their child. When users call the agency after business hours or when agents are not available for assistance, we are routing them to leave a voicemail with their babysitter requirement and contact number. With this solution, agents can focus on more complex tasks rather than manually retrieving voicemails, analysing them and coming up with a resolution. When the caller dials the agency phone number during office closed hours or peak hours when agents are not available to serve them, we route the caller to the voicemail menu where we ask them to leave a voicemail with babysitting requirements and their contact details, etc. Once the voicemail is available, we extract it and convert this speech to text using OpenAI’s whisper API which gives us the voicemail transcription. After that, we meticulously perform the prompt engineering for ChatGPT API to provide us all the required information from voicemail like intent, sentiment, babysitting date and time, etc in JSON format. Using this information, we query the EmployeeSchedule table which is in the H2 database. Once we have the information about availability of babysitters, we query RedisJSON to get the employee profile information like employee name, contact details, date of birth, languages spoken, image, etc. We then build a PDF document using itext library. This PDF containing available babysitter information will be sent on the caller’s WhatsApp. After this, we send an SMS to the agency as an alert notification about the customer enquiry and ask them to get in touch with the customer. Github link - https://github.com/technocouple/technocouple-caretaker-assistant Video link - https://drive.google.com/drive/folders/1NBew2U0Xgtm04ubQszjLvZV92fowR6-D?usp=sharing Presentation - https://drive.google.com/file/d/1TBMSU5Ohyn1v2P2u_RqbZOpuCvWv1Crq/view?usp=share_link DEMO is at the end of the video.

Today, we are presenting an idea which can help in routing the Chat/Voice interaction in the contact center to the right skilled agent. As we all know Contact center plays a bigger role in any organization when providing the best customer service to the users. Ideally, in any digital interaction in a contact center ecosystem, like voice, chat or email involves a virtual agent which is the first layer to interact with the customer. VA tries to welcome the user and understand the Intent. Based on the intent, query the VA tried to have the conversation. If the Intent is complex we might need human assistant to look upon, we have a routing strategies traditionally build. Based on the parameters collected by virtual agent, routing strategy executes and determines a skilled agent but might not be appropriate everytime. So today, we thought this can be resolved by AI or provide AI prediction as an additional parameter in determining the right Queue and right agent. In this demo, We have defined 5 Queue and given a Name as shown here. Q1_GB_CHAT: Agents belonging to this queue, can help with general banking queries like account information, online bank, branch address, contact number only. And this agent can speak english, spanish. Q1_GB_CHAT_ESP: Agents belonging to this queue, can help with general banking queries like account information, And this agent can speak only spanish. Q2_HL_CHAT_ENG: Agents belonging to this queue can help with mortgage loans, home loans, home loan eligibility. The agent belonging to this queue can only speak english. Q3_CC_CHAT: Agent belonging to this queue can help anything related to credit card. Q4_FR_CHAT: Agent belonging to this queue helps with fraudulent activity on credit card. So we have given simple description or defining this Queue. We have created a Prompt template, with this description & chat summary injected dynamically into Prompt template. This prompt is sent to Text-bison model, which responds with the Queue ID.