Modzy Smart Call Center Solution

Dashboard - where user can view overall data and start recording the call
Smart Embed - allow user to easy embed MSCCS features into their existing application
Call Center Management - allow user to start the call recording, manage call data and view call data analysis.
Keyword Management - allow admin to manage keyword for audio keyword spotting to better understand customer interest.
Customer Management - allow user to manage customer data and bind audio data to customer record to view real time data analysis.
Report & Analysis - allow user to view overall sentiment score, satisfaction rate, keyword spotting and some data statistic chart.
Smart Recording - system will record the audio and process the audio data in background asynchronously
Mockup Modzy App - this page is to demonstrate the deployment plan of MSCCS.

Inspiration

This calls will recorded for training and quality purposes

I believe many people have heard this sentence before. But what happen to the recorded call data ?

Based on my previous work experience and market research in large and medium-sized companies, I found out that most large and medium-sized companies are not utilize the recorded call data and many companies are still manually organize and process the call records.

There are some existing smart call center software on the market, but most of them only focus on pre-call and post-call processing. In addition, it is not easy to process audio data for multilingual calls like Malaysia.

So I think there may be a solution to automate the post-call process and utilize all these call data to improve the company's business processes and performance, better understand customer interests and satisfaction, and make better business decisions.

Therefore I created MSCCS - Modzy Smart Call Center Solution.

What it does

MSCCS is a smart call center software that can process and utilize post-call data to help the company grow. Since pre-call and pre-call processing are quite common in the market. So in this project I will emphasize how this post-call processing can help business growth and how Modzy AI can bring benefits to this post-call data processing.

The following is a list of MSCCS features:

Smart recording : Allow users to start recording, then system will stores it in the cloud and performs post-call data AI processing after the call session end.
Upload Audio : Allow users to upload audio file to peforms post-call data AI process.
Quick Recording : Floating button at the bottom right to allow users quickly start the recording from any page on the dashboard
Quick Embed : Allow users to embed MSCCS features into their existing application
Keyword Register : Allow user to register their keyword to better understand customer interest
Customer Real Time Analysis : Allow user to view customer data analysis
Audio Data Analysis : Allow user to view audio data analysis
Overall Report and Data Analysis : Allow user to view overall data analysis
Post-call data AI processing : Utilize the call data and automate the process and generate analysis data. ( Refer miro board for more clear image )

Innovativeness

Post-call data processing is like a niche topic in the call center software market. There are sentiment analysis for call center solutions on the market, but they actually do not make full use of audio data and highlight the post-call analysis function.

Therefore, I believe that post-call data processing is unique and innovative in the market, and it will help companies grow through the use of call data.

Business Impact

Automate business process and data organization
Enhance customer service experience
Better understading on customer interest, keyword and satisfaction level.
High level business analysis & decision making

How I built it

Project Background

The project is built using the Laravel framework and hosted on the Alibaba Cloud ECS server. Since the recording feature needs to be processed using HTTPs, so I purchased a domain and register SSL certificate with Let's Encrypt.

Overall project planning is using Miro to illustrate the input, process flow and output.

All AI queries are using Modzy AI Model, except audio transcription uses the Google Speech-to-Text API due to the limitation of the basic account.

UI & UX

Ensuring a good user experience is always my top priority. The system is design with modern and Minimalist style, and its referring to Modzy color theme. I also did some UX analysis and applied UX design to the system, such as quick recording, UI theme consistency, overall eye-catching design, etc.

To ensure the project is running smooth and in high performance, the project resources is compiled with Laravel Mix and project is hosted in HTTPs environment.

Project Security

I have implemented some security mechanisms, such as role verification, company verification, SSL protection, ID encryption, IP jumphost, domain whitelist for embedded solutions, SQL injection protection, etc.

Async Job Process Handling

Since processing the AI queries may take some time, if using the synchronize method to process AI jobs, user will have to wait a long time, and the user experience is very poor. Therefore, I use asynchronous job processing to process AI queries, and use Laravel queues and Ubuntu Supervisor with 4 processors to enable it to run 4 asynchronous tasks at the same time.

Because there are too many AI queries that need to be processed, timeout issues may occur even running asynchronous in the background. To avoid timeout issues, AI queries are divided into different job using dynamic programming techniques - refer to Miro board for more details.

Challenges I ran into

The main challenge I encountered was handling multiple Modzy AI queries. Because the query is not responded in real time, different job requests need to be called to check the response. It is easy to process a single AI query, but it is troublesome to process multiple queries in linear ordering.

At first I used the synchronous method to process the audio data, but then I realized that processing it would take a long time and the user experience would be very bad, so I switched to the Laravel queue - asynchronous method to process the query.

The asynchronous method is completely different from the synchronous method, because it runs in a background process and needs to understand job processing, server processor, timeout processing technique, job error handling, and job response handling.

Accomplishments that I're proud of

The main accomplishments that I're proud of is that I learned asynchronous processing techniques and completed the system within 2 week.

What I learned

I have learned a lot of Modzy AI model, sentiment analysis and algorithms, asynchronous job processing and browser audio processing.

What's next for Modzy Smart Call Center Solution

The current MSCCS system only focuses on the post-call data processing , because pre-call and peri-call processing are quite common in the market. But the future plan of MSCCS is to provide an All-in-one call center solution, including pre, peri and post-call processing. Utilize the call data and work with company data to better enhance customer experience and help company grow.

What's next for MSCCS :

pre-call data AI processing
peri-call data AI processing
API development and documentation
PWA development
Gamification
Modyz AI composer package for PHP development ( I do plan to create a composer package but I don’t have enough time to optimize the code )
More input source : from phone call, social communication and work with existing call center solution.

The current MSCCS system improvement:

Code optimization
Security improvement
Responsive design
Enhance job processing algorithm
Real time error handling
Smart filter
Large data handling

Deployment Plan

Here's what I thought for deployment plan. Since there are so many existing App gallery in the technical documentation, why does Modzy not release these applications and turn them into ready-to-use application.

alt text

So that users can try out the application, apply the solution directly to their business, and embed or integrate the solution into their existing application. It will help users better understand how these AI models work in actual scenarios and improve their user experience. At the same time Modzy can also turn these App Gallery into sales.