• Inability to promptly identify weaknesses in sales scripts or objection handling.
• Lack of data for training new employees and improving the skills of existing ones.
• Loss of potential customers due to unsatisfactory service quality.
• Absence of a unified tool for analyzing communication effectiveness.
• The need to process not only standard Ukrainian but also Surzhyk (mixed Ukrainian-Russian speech).

Results

1. A fully automated analytics system. Instant access to data on service quality, sales performance, and customer needs.
2. Process acceleration: The time required to analyze a single call has been reduced from tens of minutes to just a few seconds. This allows analyzing 100% of calls instead of only a selected portion.
3. Increased productivity: Department managers receive instant reports, enabling them to make faster decisions and adjust team performance.
4. Keyword and customer query analysis helps quickly adapt offers, leading to a 10–15% increase in sales.
5. Reliability and cost savings: The AWS-based solution ensures fault tolerance and scalability, handling large call volumes without failures. Automatic analysis of 100% of calls has reduced manual monitoring costs by up to 30%.
6. Efficiency: While the exact savings percentage may vary, process automation significantly reduces labor costs and increases overall efficiency.

Description

“AutoNova-D” is a national distributor that owns the largest warehouse of spare parts in Ukraine. The company specializes in wholesale trade of original and non-original spare parts for cars, primarily of the German group. It provides a full range of goods necessary for vehicle maintenance and repair by forming a unique assortment.

Challenge:

Lack of control over the quality and efficiency of managers’ phone conversations. Before implementing the solution, “AutoNova-D” faced difficulties in assessing the performance of its sales managers in terms of telephone communication with customers and partners. Hundreds of phone calls each day contained valuable information, but listening to and analyzing them manually was impossible. This led to the following issues:

• Inability to promptly identify weaknesses in sales scripts or objection handling.
• Lack of data for training new employees and upskilling existing ones.
• Loss of potential clients due to unsatisfactory service quality.
• No unified tool for analyzing communication effectiveness.
• The need to process not only literary Ukrainian but also Surzhyk (mixed Ukrainian-Russian speech).

Solution:

To address these challenges, Softprom, as an Amazon Web Services Advanced Partner, developed and implemented a comprehensive solution for automatic analysis of phone conversations. The solution is based on AWS artificial intelligence and machine learning services, as well as integration with the client’s existing infrastructure.

Used AWS Services

• AWS: Amazon Transcribe: For automatic conversion of audio recordings into text. The service successfully handled speech specifics (Surzhyk, technical terminology).
• AWS Bedrock (with the NovaPro model): For semantic analysis of call transcripts, detection of key topics, emotions, and other valuable insights.
• AWS S3, Lambda, SQS, EventBridge, SSM Parameter Store: These services were used to build a reliable, scalable, and fully automated architecture that allows: downloading call recordings from Binotel via API, processing them sequentially, storing data and analysis results, and managing processing through a configuration file.
• AWS CDK: For automating the deployment of the entire solution.
• AWS Route 53, Certificate Manager, Cognito, Application Load Balancer, ECS Serverless, ECR: For deploying a convenient web application where “AutoNova-D” specialists can independently test and adjust prompts for analysis.

Solution Architecture

The diagram below shows the architecture of the solution.

1. Call recordings from Binotel are automatically uploaded to Amazon S3.
2. S3 events trigger AWS Lambda, which processes the files.
3. A call transcript is created from the audio recording using Amazon Transcribe.
4. Text analysis is performed using AWS Bedrock AI service and the AWS Nova Pro model with prompts configured by the client.
5. The analysis results are sent to Microsoft Power BI for visualization and dashboard creation.

Softprom’s Role

• Conducting presentations and consultations on AWS service selection.
• Developing and implementing a full “turnkey” solution.
• Integrating with the Binotel telephony system and Microsoft Power BI analytics system.
• Providing an AWS test account.
• Training the client’s employees to independently manage and improve the solution.

Project Stages

The “Call Analytics” solution by Softprom was developed step by step.

Initially, AutoNova-D was offered to test phone call analysis based on GenAI using AWS Bedrock and the NovaPro foundation model. A scheme for storing and sequentially processing call recordings was implemented using AWS S3 and S3EventNotification. First, a conversation script with speakers was generated using AWS Transcribe, then converted from JSON into a plain text file, after which the script and corresponding prompt were sent to AWS Bedrock for analysis in a format suitable for further BI system processing.

AutoNova-D was impressed with the results. As a next step, at the client's request, Softprom implemented automatic call loading from Binotel via API and full automation of the call processing workflow. AWS services such as AWS Lambda, AWS SQS, SSM Parameter Store, and AWS EventBridge Rule were used for this purpose.

Later, at AutoNova-D’s request, Softprom developed a configuration file-based control scheme for call loading from Binotel. This enabled the selection of internal lines, time intervals, and call loading frequency.

AutoNova-D’s analysts configure call analytics in Power BI independently.

Additionally, to enable the client to test their own call analysis prompts, Softprom developed a web application and deployed it in the client’s AWS account. The following AWS services were used for the application infrastructure: AWS Route 53, Certificate Manager, Cognito, Application Load Balancer, ECS Serverless, ECR, and S3. The site was developed using Python, the Streamlit framework, and AWS SDK.

Results

Before / After:

Before: Manual and time-consuming listening of calls. No systematic approach to analysis. Inability to quickly respond to communication issues.

After: A fully automated analytics system. Instant access to data on service quality, sales efficiency, and customer needs. Managers can now focus on data analysis instead of data collection.

Key Metrics:

• Process Acceleration: Call analysis time decreased from tens of minutes to a few seconds. This enables analysis of 100% of calls instead of selective ones.
• Productivity Growth: Department heads receive instant reports, allowing faster decision-making and team adjustments.
• Keyword and Customer Request Analysis helps quickly adapt offers, leading to a 10–15% increase in sales.
• Reliability: The AWS-based solution is fault-tolerant and scalable, easily handling large call volumes without data loss or failures.
• Cost Efficiency: Although the percentage of savings may vary, process automation significantly reduces labor costs and increases efficiency.
• Expense Reduction: Automatic analysis of 100% of calls eliminated the need to hire additional quality control staff, reducing operating expenses by up to 30% in manual monitoring processes.

Client Feedback

Thanks to the solution provided by Softprom, we have gained full transparency in our communications with customers and partners. Now we know exactly what our managers say, how they handle objections, and what concerns our customers and partners the most. This has given us a powerful tool to improve our service and boost sales.

 

Softprom did more than just provide a service. The company acted as a strategic partner that:

• Listened to and understood the unique needs of “AutoNova-D”.
• Provided access to cutting-edge AWS technologies.
• Developed a comprehensive, customized solution tailored to the client’s business specifics.
• Integrated the system with the client’s existing infrastructure (Binotel, Power BI).
• Conducted training and provided tools for independent solution development.