GPT-4o and Beyond

Today, OpenAI hosted a Spring Update session where they introduced an exciting new model capable of processing audio, vision, and text in real-time. This model is twice as fast as GPT-4-Turbo and costs half as much. We’ve already upgraded the CMA to harness this powerful new model.

Current Use

We’re thrilled to keep enhancing our CRM platform with the latest advancements in Generative AI and Machine Learning. Our customers are already using OpenAI’s Whisper Large-v3 model for multi-lingual translation of call recordings. These transcriptions can be analyzed by GPT-4o and future models to generate concise, high-quality conversational summaries. These summaries help counselors and customer service representatives quickly grasp client interactions, including detailed housing counseling session summaries required for audits and overviews of enrollment and customer service calls. Additionally, OpenAI’s models are being used to evaluate conversations and efficiently handle complex ISO queries, significantly reducing the time required for analysis. A one-hour conversation can now be summarized, analyzed, and reported on within seconds.

AI Task Analysis Engine

Recently, we introduced generative AI capabilities into our Task workflow, integrating AI throughout the business process. This integration allows the analysis of client information across various communication channels like email, text, and phone, helping to securely identify the best-fit programs using financial data. This enables the creation of insights and reports that were previously unattainable with traditional statistical methods. The AI Task Engine also provides real-time recommendations to assist clients effectively.

GPT Agents

We have also introduced GPT Agents outside of the CRM to help build SQL queries for generating custom reports within the CMA. These agents can access the CRM’s database schema and are a step toward natural language reporting. Our GPT Agents can answer questions about system functionality and even process screenshots to provide step-by-step guidance on specific tasks. While these agents are still in the prototype stage and can occasionally produce suboptimal solutions, each advancement in reasoning and memory processing in new models enhances their performance. We are optimistic that these improvements will make interacting with the system easier and faster than ever.

Realtime Audio Processing

We’re also excited about the real-time audio capabilities demonstrated today, which will be available in the coming weeks. We’ve developed Voice Agents to assist with enrollment, qualification, counseling, and customer service. However, these agents were not ready for prime-time as they previously faced delays and pauses as they processed information. The new audio agents are a significant improvement, offering tone and emotional detection, nuanced communication styles, multi-language support, and smooth handling of real-time interruptions. As we offer a comprehensive solution, including phone system hosting, we look forward to integrating these tools to better support your counselors and agents in delivering high-quality assistance to more people.


While OpenAI and others work on enhancing these models and their technical infrastructure, we’re committed to exploring how these tools can best support you and your customers. This presents an opportunity to reassess all job roles and discover how these tools can enhance, assist, and redefine our workflow to maximize satisfaction for both our team and those we serve.

We’re excited to unlock the potential of GPT-4o and future models, whether it’s GPT-5 or something entirely new.

New Feature: AI Task Analysis Engine

Last year marked a transformative phase for our technology, as we introduced a feature that allowed custom analysis of Call Log transcriptions. This breakthrough enabled our CRM users to unlock new potentials, pushing the boundaries of what we imagined was possible. From evaluating the quality of customer service calls against rigorous ISO standards in mere seconds to generating insightful funding notes for Housing Counseling and enhanced summary notes for Customer Service interactions, the impact was profound. Users leveraged our technology to not only analyze the text but to delve deeper, extracting meta-data to spotlight outliers and streamline business processes. The capability to pose complex, multi-layered questions and receive high-quality insights from the GPT-4 model was both surprising and incredibly empowering.

Building on this momentum, we’re thrilled to introduce a groundbreaking enhancement: the deep integration of Large Language Models directly into the workflow process through Tasks. This new development is designed to not just augment, but revolutionize how tasks are analyzed, automated, and acted upon within your organization. Imagine having the ability to dynamically generate AI-driven insights and actions directly from routine tasks, all tailored to enhance efficiency and decision-making.

We are eager to see the innovative applications you’ll dream up with this new capability. Your creativity and feedback are what drive us forward, and we’re committed to evolving this technology together.

New Feature Overview: AI Task Analysis Engine

The AI Task Analysis Engine is designed to seamlessly integrate AI capabilities into organizational workflows by allowing tasks to trigger dynamic AI prompts. These prompts, defined in DocumentTemplates, are then evaluated by a Large Language Model. The results of this evaluation are stored in the ClientMessages table, providing a rich dataset for analysis and further action. This feature includes several key components:

  • AI_PROMPT_TASK DocumentTemplate: A new type of document template that allows the creation of dynamic AI prompts.
  • Dynamic AI Prompt Evaluation: Upon task creation or completion, an AI_PROMPT_TASK document is sent to a Large Language Model for evaluation.
  • Enhanced ClientMessage Table: Modifications to the ClientMessage table enable the storage of AI Task Analysis results, adding a new dimension to client communications.
  • Expanded Smarty Variables: Inclusion of ClientMessage and CallLog records in Smarty variables, enabling the creation of AI prompts based on client interactions.
  • ClientMessageType Field: A new field in the ClientMessage table to distinguish between different types of messages, including AI-generated messages.

Benefits and Applications

The AI Task Analysis Engine is a versatile tool that can be leveraged in various ways to benefit an organization. Here are some examples:

Enhanced Reporting

Organizations can use AI to analyze client interactions and communications, generating comprehensive reports that highlight trends, preferences, and areas for improvement. For instance, AI can identify common issues raised by clients in secure messages or emails, allowing businesses to address these concerns proactively.

Automated Decision Making

By evaluating AI_PROMPT_TASK documents, the AI Task Analysis Engine can assist in automating decision-making processes. For example, it could analyze financial data to recommend Debt Management Plan strategies or assess client satisfaction levels to suggest customer service improvements. This automation speeds up decision-making and ensures it is backed by comprehensive data analysis.

Custom Client Communications

With the integration of ClientMessage and CallLog records into Smarty variables, organizations can automate the creation of customized client emails based on the results of prior analysis results. These emails can provide focused recommendations based on an individual’s financial situation and geographic location, enhancing the personalization of client communications. For example, this feature could highlight available housing assistance programs in the county of the client by performing realtime web searches to identify programs that best fit the financial circumstances of the client.

Streamlined Operations

The ability to store AI Task Analysis results in the ClientMessage table ensures that all client-related information is easily accessible. This centralization of data streamlines operations and improves the efficiency of client communication management.


The AI Task Analysis Engine represents a significant step forward in integrating AI into business processes. By enabling dynamic AI prompt evaluation and the automated creation of personalized client communications, this feature not only streamlines operations but also enhances the quality of client interaction and decision-making. As organizations continue to seek ways to leverage technology for competitive advantage, the AI Task Analysis Engine stands out as a tool that can transform data into actionable insights and tailored communications, paving the way for a more efficient and responsive business model. As AI models continue to advance and we move beyond multi modal generative models, we will continue to update the available tools so that you can leverage cutting edge technologies. All these changes are motivated by the intention to help you expand your capability to help our community.

Changes – 2024 Q1

Welcome to our quarterly blog post where we outline the significant enhancements, updates, and changes introduced in VERSION 24Q1. Our continuous effort to improve our platform has led to numerous additions and improvements that aim to enhance functionality, increase efficiency, and improve the overall user experience. Here’s a comprehensive summary of what we’ve achieved in the past quarter:

Major Additions

  • Client Forms: A significant leap forward in custom data management, allowing the storage of custom data using JSON. This addition is aimed at providing a flexible solution for capturing and storing client-specific information in a structured format.
  • Appointment Tasks with User Availability Schedule: We’ve introduced a sophisticated system to manage appointment tasks, which now incorporates user availability schedules. This feature ensures that appointments are scheduled efficiently, taking into account the availability of the involved parties.


We’ve made several updates to our technology stack and integrated tools to ensure our platform runs smoothly and securely. Notable updates include:

  • Updates to PHP, including the new PHP 8.3.4, ensuring our platform is running on the latest, most secure PHP versions.
  • jQuery updated to version 3.7.1.
  • Fullcalendar updated to version 6.1.11, enhancing our calendar functionalities.
  • ACE editor updated to version 1.32.8, providing an improved code editing experience.
  • Chart.js updated to version 4.4.1, ensuring better and more versatile charting capabilities.
  • Updates to PEAR packages, including Mail 2.0.0 and Mail_Mime 1.10.12, enhancing email functionalities.
  • Updates to front-end libraries including @tabler/core and @tabler/icons-webfont, bringing new features and icons to the user interface.
  • Other updates include htmldoc 1.9.18, qpdf 11.9.0, and Smarty 4.5.1, each contributing to improved document processing and template management.


Changes this quarter span across various aspects of our platform, from error handling and user interface improvements to backend logic enhancements. Highlights include:

  • Enhanced error handling capabilities, particularly for Peregrin errors, improving the platform’s stability and reliability.
  • A series of user experience improvements, such as adding dynamic headers to Manager Reports, and refining the paperwork process.
  • Significant backend updates, including migrating from PHP 8.2.x to PHP 8.3.4, to ensure compatibility and take advantage of the latest PHP features.
  • Changes aimed at improving user interface and interaction, such as adding logic to register Smarty classes and functions, fixing SQL errors, and improving the layout and functionality of reports and forms.
  • Numerous fixes and enhancements targeting specific functionalities, such as Client Budget, Client Doc management, and Appointment Scheduling.

This quarter’s updates underscore our commitment to providing a robust, efficient, and user-friendly platform. We’re excited about the positive impact these changes will have on our users’ experience and look forward to continuing to evolve and improve in the coming quarters.

Stay tuned for upcoming changes in 2024 Q2 as we add further integration between Task workflows and AI analysis/feedback.

Changes – 2023 Q4

Welcome to our 23Q4 software update summary! This quarter, we’ve introduced several major additions and updates, alongside various changes that aim to enhance user experience and system efficiency. Let’s dive into the highlights.

Major Additions

Enhanced Document and Extranet Package Management

We’ve overhauled our Document and Extranet Package Management system for improved usability and flexibility. This refactor provides a more intuitive interface for managing documents and extranet content, ensuring that users can easily navigate and utilize these features.

Deluxe Check Disbursement Batching

A significant addition this quarter is the support for disbursement batching into a single Deluxe Check. This feature streamlines the payment process, allowing for more efficient fund distribution and reduced processing times.

Refactored Client Budget Support

The Client Budget system has been refactored to include comprehensive support for income, deductions, and CoApplicant details. This update ensures a more detailed and accurate representation of client financial situations, enabling better budget planning and management.

Peregrin API Integration

We’ve introduced support for the Peregrin API accountStatus refresh, enhancing our system’s ability to remain synced with current account information. This integration ensures that account statuses are up to date, providing accurate data for decision-making processes.


This quarter, we’ve updated several core components and libraries, including FreeBSD 14.0, Apache 2.4.58, PHP 8.2.14, MySQL 8.0.35, and many more. Noteworthy updates include the ARMv6.0.2 SDK, the GPT4-Turbo 128K model, and various frontend libraries like Bootstrap 5.3.2 and Chart.js 4.4.0. These updates ensure that our platform remains on the cutting edge, offering robustness, security, and the best user experience.


We’ve made a series of changes aimed at refining existing features and addressing user feedback. Some of the notable changes include:

  • qpdf and PHP Updates: The qpdf library was updated to 11.7.0, and PHP to 8.2.14, ensuring compatibility and security.
  • Enhancements to Client and Creditor Views: New columns and filters were added to improve data visibility and navigation.
  • Document Template Refinement: DocumentTemplate types were renamed, and the unused documentTemplatePath field was removed, streamlining the document management process.
  • Initial Support for Peregrin API: Beginning 2023.12.18, we started supporting Peregrin API accountStatus refreshes, marking a significant step towards improved data accuracy.
  • Interface and Usability Improvements: Updates were made to the user interface, including migration to Bootstrap 5.3.x and enhancements to report views.

Stay tuned for more updates as we continue to evolve and enhance our platform to meet and exceed the expectations of our users. Thank you for your ongoing support and feedback, which play a crucial role in shaping our product’s future.

Peregrin API AccountStatus Refresh

We are happy to announce an enhancement to our Client Support Software. Our CRM now supports the integration of Peregrin’s API AccountStatus refresh, a feature that promises to streamline your financial data management and improve the accuracy of your account status information.

How to Activate This Feature

To take advantage of this new capability, you will need Peregrin API keys. We understand the importance of a seamless setup process. Therefore, we encourage you to reach out to us. Our team will handle the request for these keys on behalf of your organization, ensuring a smooth and efficient integration. There is no additional cost for utilizing this functionality either from Client Support Software or Peregrin.

Utilizing the AccountStatus Refresh

Once this feature is activated, you can easily update Creditor records that are linked through Peregrin. You simply need to input the Peregrin Creditor ID or enter ‘1’ to enable the functionality.

Upon activation, a new Peregrin Refresh link will automatically appear next to eligible ClientCredit records within the DMA/Creditors tab of our CRM. This link is your gateway to up-to-date financial data.

The Peregrin Refresh Link – Keeping Your Data Current

By clicking on the “Peregrin Refresh” link, you can initiate a balance update for your ClientCredit records. This feature is particularly valuable because if the information provided by Peregrin is more current than the data already in the system, the ClientCredit record is automatically updated with the latest balance information.

We are committed to continuously improving our services and offerings to meet your evolving needs. We look forward to seeing how this update will positively impact your organization’s data management processes.

For any questions or assistance in activating this feature, please do not hesitate to contact us.

Integrating GPT-4 Turbo 128k

The integration of OpenAI’s GPT-4 Turbo 128k model into Client Support Software’s CRM system marks a technical advancement in handling customer interactions. This model allows for the processing of longer call transcriptions, which is particularly useful for organizations that require detailed records for compliance and monitoring purposes.

For ISO call recording monitoring, the CRM’s ability to process and analyze lengthy call transcriptions in full helps businesses ensure that their customer service complies with ISO standards without having to manually review recordings.

In the realm of financial counseling, the CRM now aids in generating NFCC detailed notes from calls, which can be used for maintaining compliance and for reference in subsequent counseling sessions.

Calls can be analyzed to provide detailed and helpful feedback to counselors after they have completed a counseling call.

Housing counseling services, which often involve complex conversations, can benefit from the CRM’s summarization capabilities. This tool condenses long calls into concise summaries, thus streamlining the review process for counselors.

The GPT-4 Turbo model’s key attributes include a larger memory span, which gives it the ability to process much larger text summaries, up to around 1/2 million characters. This unlocks the ability to create a DocumentTemplate in the CMA with full customer history and utilize GPT to perform multiple kinds of previously unavailable analysis. The turbo version is much faster than the previous version of GPT-4, reducing wait times for analysis to complete and increasing productivity.

We are excited to see how you will use these new capabilities as we continue to enhance the tools available for you in order to provide outstanding financial counseling to consumers.

Changes – 2023 Q3

Excited to announce the latest round of updates and enhancements, which brings new features, important fixes, and other changes to ensure smooth and efficient functionality. Below is a comprehensive breakdown of the recent modifications.

Feature Additions:

  • New Reports and Columns:
    • Added Spinwheel Usage Report.
    • Added links to view Sent and Returned RPPS DMP records within the Client Credit page.
    • New Creditor column: Date Last ClientCredit Created added.
    • Added Last Clear Payment column to Client Balances report.
    • Added ‘Date Opened’ column to the LMA/Creditors tab.
  • New File and Format Support:
    • Support for new file types in ClientDocs (heic, heif) added.
    • Added ACH Return CSV file format for Seacoast Bank.
  • Additional Features:
    • New feature to show multiple matches message during the ClientCredit Automatch process.
    • Improved billerID Automatch tool for better handling with Chase and Bank of America.
    • Added CoApp emails to Client/Email Activity section.
    • Enhanced Call Queue with new dispositions and filters.
    • Added ‘Show num ClientTask Links’ filter to Task Management report.
    • Added CoApp and CoID columns in various views and reports.


  • Error Resolutions:
    • Fixed Manager Report error related to client’s timezone determination.
    • Fixed error in Custom Reports and Automatch BillerIDs.
    • Fixed issue with recomputation of $Totals variable.
    • Fixed errors related to ClientTask and CMAForm formatting.
  • Logic Corrections:
    • Closed Spinwheel accounts with a balance are not automatically marked as Charged Off.
    • Corrections in Spinwheel class, Creditor name formatting, and Spinwheel account status determination.
  • User Interface Adjustments:
    • Checkbox in ACH Batch is hidden when Routing Number is invalid.
    • Resolved issue with display of Balance information in Client/Creditors tab.


  • Software Updates:
    • PHP updated to 8.2.11
    • htmldoc updated to 1.9.17
    • netpbm updated to 11.03.05
    • MySQL updated to 8.0.33
    • Smarty updated to 4.3.2
  • Logic Updates:
    • Updated Spinwheel and Equifax Credit Report import logic.
    • Updated Call Queue and Client/Creditors display and functionality.
    • Enhanced Manager Report layout and columns.
  • Other Updates:
    • Updated ClientSpinwheelID and Spinwheel API.
    • Updated CSP Stats section and various other fields and sections for improved performance.

Other Changes:

  • Functionality Enhancements:
    • Removal of certain system tasks and requirements for smoother operation.
    • Adjustment in Client/Creditors and Payment views.
    • Enhancements in Task Management and Manager Reports.
    • New logic for handling Closed Leads and Spinwheel leads.
  • Synchronization:
    • Sync with TotalCommon for various updates.
    • Util::getDurationIntervals() updated to accept more values.
  • Miscellaneous:
    • Added description to Client/Include in Payment Batch field.
    • Adjusted SMS message sending times according to the client’s timezone.
    • Enhanced support for SIP Endpoints and WebRTC phones.

Thank you for your continued support and feel free to reach out with any questions or concerns about the latest updates!

Changes – 2023 Q2

Many changes and enhancements have been made to the CMA during Q2. Details are available in the internal CMA Changelog. 

Feature Additions:

  • Added Spinwheel User Profile and Liability Refresh support.
  • Updated design of Extranet.
  • Introduced support for custom endpoint for Whisper call recording transcriptions.
  • Introduced support for using oobabooga/text-generation-webui as an API endpoint for CallLog sentiment analysis.
  • Implemented Google Workspace Gmail messages display within the Client record.
  • Introduced Round Robin Lead Assignment.
  • Added DocumentTemplate Send To/Send From options.
  • Implemented formatting of call transcriptions for inbound voicemails.
  • Added Settlement Amount field to ClientCredit record when Hardship Indicator A or C is selected.
  • Enabled Task: Job Creation SQL to associate tasks with clientCreditIDs.
  • Introduced new System Setting: DMP_MIN_ACCOUNT_AGE_MONTHS that specifies the minimum account age for participation in a DMP.
  • Added new userPhone field to be used as the destination of DocumentTemplates sent via SMS.
  • Added round robin lead assignment functionality within Auto Import To Lead.
  • Added ability to view CallLog details through the EditCallLog form.
  • Added ability to store $Contract variables on the Extranet Contracts.


  • Fixed linking of CallLog records to Clients when more than one phone number matches.
  • Fixed computation of Program APR in Creditor Hardships when the hardship indicator has a null APR and specifies an APR Term.
  • Fixed Unread Message Count in the Client Portal.
  • Corrected the use of 0.00% APR in Creditor Hardships.
  • Fixed Client Balance Report: Payment Filter frequency when Monthly is selected.
  • Corrected processing of email attachments in jobs/ParseEmail.php.
  • Fixed parsing of email attachments emailed to CMA.
  • Fixed refererURLs when viewing the ClientTask Log.
  • Fixed Manager Report LMA CallLog Duration (minutes) for current Leads.
  • Corrected permissions for Global Report Filters.
  • Fixed formatting of names with special characters in Navigation search.
  • Fixed Phase TimeLine report filtering.
  • Fixed CallLog detail Ext Name filter when filtering by Agency.
  • Fixed saving of Report Filters as Global.


  • Added icons to actions and tabs throughout the system.
  • Enhanced the import process of Call Log records by cleaning WebRTC shadow extensions.
  • Enhanced LMANavigation to include agencyID (aID) column and advanced search functionality.
  • Updated description for Fax Transmission Status in the ClientCredit form.
  • Updated Find Inbound Caller to handle webrtc shadow extension prefixes.
  • Updated call icons in the Client record.
  • Updated Call Queue Report to display current dispositions.
  • Updated Payroll report to filter the Agency by the Client and not the User.
  • Updated DataRetentionCleanup.sql queries to include Clients without Payment or ClientLog records.
  • Updated the Deluxe specifications to truncate payee names in excess of 50 characters.
  • Updated CallLog report to allow filtering by any date range.
  • Migrated Extranet steps to use Bootstrap 5.3.

Other Changes:

  • Enhanced CallLog report and CallLog detail with additional filters, columns, and percentage breakdowns.
  • Improved the readability and functionality of various reports by adding new columns and filters.
  • Enhanced ClientCredit details with a more detailed explanation for “Date to Queue” field.
  • Improved functionality of Agency fields and calculations in the budgeting process.
  • Enhanced the transcription process with Whisper models to exclude consecutive lines of similar text.

New Feature: Custom AI API Endpoints

We are excited to announce a new feature for our CRM customers: Custom AI API Endpoints. This feature allows you to use your own AI models or open source alternatives to the ones provided by OpenAI, depending on your needs and preferences.

Custom Endpoints can help you save money on AI costs while enhancing your data privacy. You can run your own AI models on your own hardware, without sending any data to third-party services. For example, you can use the open source Whisper model to transcribe call recordings using the same API as OpenAI. Whisper is a large language model that can run on a consumer GPU with 11GB of RAM.

Large language models are powerful AI tools that can generate natural language texts for various purposes. They are trained on massive amounts of data and learn to capture the patterns and nuances of human language. As the size of these models increases, they can exhibit emergent capabilities that go beyond their original training objectives. For example, some large language models can answer questions, write summaries, generate code, and even create music or art.

However, running large language models requires a lot of computational resources, especially video RAM (VRAM). VRAM is the memory that is used by the GPU to process graphics and other tasks. The more VRAM a GPU has, the larger and more complex models it can run. For example, a 30 billion parameter model requires 24GB of VRAM, which is available in some high-end consumer GPUs. As of June 2023, recent models of this size are comparable to GPT-3.5, one of the most advanced models from OpenAI. By combining multiple GPUs, it is possible to run even larger models with more emergent capabilities.

Open source large language models are advancing at a rapid rate, thanks to the efforts of researchers and developers around the world. As our customer, you have the option of using OpenAI’s models or hosting your own open source models. As hardware capability inevitably improves and prices come down, you will be able to run more powerful and privately hosted models that can easily integrate with your existing customer data.

The future is bright and full of possibilities as we leverage these technologies in service of society. If you’re interested in exploring this option, please contact us for more details. We hope you enjoy this new feature and we look forward to hearing your feedback.

New Feature: Gmail integration with Client record

We are excited to announce a new feature for our Client Support Software: Gmail integration within the Client record. This feature allows you to view all the emails related to a client from your Google Workspace account without leaving the CMA. You can also filter out irrelevant or sensitive email addresses by using exclusion patterns. Here’s how it works:

  • To enable this feature, you need to have a Google Workspace Service level account with Domain Wide delegation. This allows our software to access your emails on behalf of your users. Please contact us so that we can set it up on your behalf.
  • After enabling the feature, you can go to any Client record in the CMA and the relevant Gmail messages associated with the Client email will be displayed in the Info tab, under the Email Activity section. You can click on any email summary to see the full text message in the CMA.
  • The Email Activity section is updated asynchronously, so there is no performance penalty when loading the Client record.

We hope this feature will help you improve your communication with your clients and streamline your workflow. If you have any questions or feedback, please contact us at

New Feature: AI Call Log Transcription Analysis

Following up on the addition of Call Log Sentiment Analysis, we’ve added the ability perform transcription analysis within Client Support Software’s CRM and OpenAI’s language models. This feature allows you to create and use AI prompts for Call Log Transcription Analysis, which can help you gain insights into your customer conversations and optimize your service delivery.

How can AI help you with Call Log transcription analysis?

Call Log transcription analysis is a process of extracting insights from the transcripts of phone calls. It can help you understand the needs, preferences, and feedback of your customers, as well as identify areas for improvement in your products or services. AI can help you with Call Log transcription analysis by automating the transcription of audio files, applying natural language processing techniques to analyze the content and sentiment of the conversations, and generating reports that summarize the key findings and trends. AI can also help you with Call Log transcription analysis by providing recommendations and suggestions based on the data, such as how to improve customer satisfaction, retention, and loyalty, or how to optimize your sales and marketing strategies.

AI prompts are questions or commands that you can use to interact with a large language model (LLM), such as OpenAI’s GPT and get relevant and useful responses. For example, you can use an AI prompt to ask an LLM to summarize the main points of a Call Log transcription, or to suggest actions or solutions based on the customer’s feedback.

By using Document Templates, you can create AI prompts that are tailored to your specific Call Log transcriptions and your desired outcomes. You can also use variables and placeholders to make your AI prompts dynamic and adaptable to different situations. For instance, you can use a variable to insert the customer’s name or previous communication history into your AI prompt, or you can use a placeholder to indicate where you want the LLM to fill in some information.

What are some examples of AI prompts for Call Log transcription analysis?

Here are some examples of AI prompts that you can use or modify for your Call Log transcription analysis:

  • In one sentence, describe what the customer wanted and how the agent resolved the issue.
  • Write a brief summary of the call, including the customer’s problem, the agent’s solution, and the customer’s satisfaction level.
  • How would you summarize this transcription for a manager who wants to know the main outcome of the call?
  • Analyze the sentiment of the customer and the agent during the call:
  • Using a scale from 1 (very negative) to 5 (very positive), rate the sentiment of the customer and the agent at different stages of the call.
  • Identify the emotions that the customer and the agent expressed or implied during the call, and explain how they affected the communication.
  • What was the overall tone of the call? Was it friendly, professional, angry, frustrated, etc.? How did it change over time?
  • Suggest improvements or best practices for the agent based on the Call Log transcription:
  • Based on this transcription, what are some things that the agent did well and what are some areas that they could improve on?
  • Give three specific feedback points for the agent, along with examples from the call and suggestions on how to implement them.
  • How could the agent have handled this call more efficiently or effectively? Provide some concrete tips or recommendations.

How can diarization be performed based on the conversational context?

Diarization is the process of identifying who is speaking when in a multi-speaker audio recording. Diarization can help you segment your Call Log transcriptions into speaker turns and assign labels or names to each speaker. This can make your transcriptions more readable and easier to analyze.

One way to perform diarization based on the conversational context is to use an LLM that can recognize speech patterns, cues, and features that indicate speaker changes. For example, an LLM can use contextual clues such as pronouns, names, roles, topics, etc., to identify speakers based on their content. A prompt can be used to ask the LLM to perform the diarization of the transcription. If there are more than two speakers, a prompt like “Perform diarization on the following phone conversation that may include speakers such as [IVR, Voicemail, Agent, Client]”.

Adding new Prompts

Under System Management/Document Templates, add a New Record, select Type = AI_PROMPT_CALLLOG

Enter the prompt using plain text and optional dynamic Smarty variables available within the CMA

Analyzing Call Log Transcriptions with the Prompts

A Call Log record can be viewed from within the Client record or from within the Call Log Detail report. In both cases, when viewing the individual transcription, the predefined prompts are available for selection. They will populate the prompt text area which can be further customized for a personalized analysis.

Once the desired prompt is written, submit the record, and analysis is performed. The resulting. analysis is now available to view either as a popup when hovering over the Call Log notes or within the Call Log record. The prompt that is used, is stored with the response so that the response can be understood within its context. For example:

We hope that this help your organization provide constructive feedback to assist counselors who in turn assist clients. Please contact us for implementation questions.

Changes – 2023 Q1

Many minor changes and enhancements have been made to the CMA during Q1. Details are available in the internal CMA Changelog. 

The following is a highlight of changes:

🔧 Improved check handling and money-matching magic, making your financial tasks a breeze!

🌐 PHP 8.2 is here to stay, future-proofing for the next 2 years.

🧹 Data cleanup just got smarter, keeping your information fresh and tidy.

🔍 New dynamic filters for SQL Reports make your documents more versatile than ever.

⏱️ Track your breaks alongside your sessions with enhanced Session Log Calendars.

🔎 Dig deeper into client balances with new filter options.

🎨 Out with the old, in with the new Bootstrap Icons for a sleeker look!

📅 Task views now come with a handy Calendar view for a fresh perspective.

📈 Visualize your Task workflow like never before with our directed graph layout.

⏳ Customize Imported Lead waiting times with ease.

💰 Take control of extra funds allocation with our new customizable settings.

  • – add Setting::DISBURSEMENT_EXTRA_FUNDS_ALLOCATION_LIMIT to allow customization of the amount of extra funds that are automatically allocated when generating Disbursements
  • – add Setting::DISBURSEMENT_EXTRA_FUNDS_ALLOCATION_ALGORITHM to allow customization of the order in which extra funds are allocated if an account is not explicity set.
    Avaiable options are:
    • 1=Lowest APR,
    • 2=Highest APR,
    • 3=Lowest Original Balance,
    • 4=Highest Original Balance,
    • 5=Lowest Payment,
    • 6=Highest Payment.

📞 Transcribe Call Log recordings using OpenAI’s Whisper API.

😀 Call Log records with Transcriptions now display the sentiment of the conversation with top 3 emotional tones to save you time when analyzing calls!

New Feature: Call Log Sentiment Analysis

We are happy to announce a new feature for Client Support Software’s CMA: Call Log Sentiment Analysis. This feature builds on yesterday’s addition of Call Log Transcription to provide sentiment analysis using OpenAI’s text-davinci-003 model. In this post, we will explain what sentiment analysis is and how it is useful to have in a CRM.

Sentiment analysis is the process of identifying and extracting the emotional tone and attitude of a speaker or writer from their words. It can help you understand how your customers feel about your products, services, or interactions with your agents. Sentiment analysis can also help measure customer satisfaction, loyalty, and retention.

With Call Log Sentiment Analysis, you can now automatically analyze the sentiment of your call transcripts using OpenAI’s text-davinci-003 model. This model is one of the most advanced natural language processing models available today. It can handle complex and nuanced language expressions and generate accurate and consistent sentiment scores.

Call Log Sentiment Analysis works by assigning a score between 1 and 5 to each call transcript. A score closer to 1 indicates a very negative sentiment, while a score closer to 5 indicates a very positive sentiment. In addition, the top three emotional tones are identified next to the sentiment score.

You can access Call Log Sentiment Analysis from the CMA Call Log Detail Report. You will see an emoji-coded display of your call transcript with the sentiment scores for the entire conversation. You can also filter your call logs by sentiment score using the “Sentiment” filter in your call log list.

Call Log Sentiment Analysis can help you improve your customer service and sales performance in many ways. Here are some examples:

  • You can identify unhappy or dissatisfied customers and take proactive actions to resolve their issues or offer them incentives.
  • You can identify happy or satisfied customers and ask them for referrals, testimonials, or reviews.
  • You can monitor the quality and effectiveness of your agents’ communication skills and provide them with feedback or training.
  • You can discover insights into your customers’ needs, preferences, pain points, or objections and use them to improve your products or services.
  • You can track trends and patterns in customer sentiment over time and across different segments or regions.

We hope you enjoy this new feature and find it useful for your business. We are always working on improving the CMA to provide you with the best tools for customer support and sales. If you have any questions or feedback about Call Log Sentiment Analysis or any other feature, please contact us.

New Feature: Call Recording Transcriptions

Client Support Software has integrated OpenAI’s text-to-speech API known as Whisper in order to optionally transcribe call recordings for customers that use our PBX hosting product. This feature will increase productivity and benefit CMA users by allowing them to quickly review the transcription of a conversation and extract key insights.

Whisper is a state-of-the-art speech-to-text model that can transcribe audio files in any of the supported languages into text . It can also translate and transcribe audio files into English. Whisper has received immense praise from the developer community for its accuracy and speed.

With Whisper integration, CMA users can easily access the transcription of any call recording using the Call Log reports. They can search for keywords, filter by date, duration, caller ID, or extension, and export the transcripts as CSV files. They can also listen to the original audio file while reading the transcript.

This feature will help CMA users to:

  • Improve customer service by identifying pain points, feedback, and satisfaction levels from call recordings.
  • Enhance training and coaching by reviewing call transcripts and providing feedback to agents.
  • Save time and resources by avoiding manual transcription.

In order to enable this feature, your company should open an account with OpenAI to obtain an API key. Existing customers may contact us to enable this functionality. We hope you enjoy this new feature and we look forward to hearing your feedback.

Using Spyderbat for our SecOps

How Client Support Software uses Spyderbat to secure its cloud native environment

Client Support Software (CSS) is a leading provider of enterprise relationship management solutions for the debt and housing counseling industries. We help our clients streamline their workflows, improve their customer service, and grow their businesses.

As a cloud native company, we rely on Linux VMs and FreeBSD Jails to run our applications and services. Our hosted PBX solution integrates with our CRM software. It runs on Linux and we practice locking access down to reduce the attack surface as much as possible. However, we also face various security challenges in this dynamic and complex environment, such as:

  • Application drift: As we deploy new versions of our software, we need to ensure that they behave as expected and do not introduce any vulnerabilities or performance issues.
  • Supply chain attacks: We use third-party components and libraries in our software development lifecycle (SDLC), which could be compromised by malicious actors or contain hidden backdoors.
  • Zero-day exploits: We need to protect our systems from unknown threats that target unpatched vulnerabilities or exploit novel techniques.

To address these challenges, we partnered with Spyderbat, a cloud native runtime security platform that uses eBPF (extended Berkeley Packet Filter) technology to provide unparalleled visibility and protection for Linux VMs. With over two decades of experience hosting and building Customer Relationship software, we became a Spyderbat Design Partner. This allowed us to provide feedback on use cases that would help tune the product for our usage needs and help those with similar workflows.

Spyderbat is a game changer for us because it gives us the ability to harden our Linux cloud runtime environments and keep our applications rolling. Here are some of the benefits we get from using Spyderbat:

Flashback: Time travel for troubleshooting

Spyderbat’s Flashback feature is like having a continuous runtime digital recorder. It allows us to go back in time and see every step that led to an event of interest, such as a service interruption or operational changes made as a result of a software update.

With Flashback, we can eliminate the pain of scanning logs or reproducing errors. We can instantly pinpoint the root cause of any issue by viewing OS kernel traces that reveal every process, file, network, user, syscall, signal, etc. involved in the causal chain.

Flashback also provides early warning signs of troubling traces by alerting us when it detects anomalies or suspicious behaviors. This helps us proactively prevent problems before they escalate or impact our customers.

Guardian: Application drift detection

Spyderbat’s Guardian feature enables us to reduce interruptions by automatically comparing running applications against prior versions. It alerts us when it detects application drift, which could indicate bugs, performance degradation, or security breaches.

With Guardian, we can end application drift by having the insight to instantly course correct and get our application back on rack.

Interceptor: Signatureless attack prevention

Spyderbat’s Interceptor feature provides automated runtime attack eradication that stops attackers in their tracks using kernel-level eBPF data. It can instantly detect and surgically block problematic traces as they begin to unfold, without relying on signatures or rules.

With Interceptor, we can block attacks targeting known or even unknown vulnerabilities, including:

  • Supply-chain attacks
  • Data exfiltration
  • Malware, Ransomware, and Cryptojacking
  • Zero-Day attacks

Spydertraces: Visual overview of Organization

Sometimes, an image is worth a thousand words. In this case, the causal graphs provided by Spyderbat are worth more than a thousand seconds. Being able to visually see connections and relationships between Linux VMs, processes, and network connections, we are able to quickly get a conceptual grasp of what activity is transpiring throughout our server deployments. This allows us to ask better questions about how we can further secure our infrastructure and saves valuable time by answering existing security questions.

Together, these features allow us to stay ahead of security threats, and focus on providing excellent service to our customers.

Changes – 2022 Q4

Many minor changes and enhancements have been made to the CMA during Q4. Details are available in the internal CMA Changelog. 

The following is a highlight of changes:

    🌐 PHP 8.1 compatibility achieved, setting new standards for the CMA!

    📊 HUD reference data now updated to May 2022 for even better accuracy.

    🚩 Client address changes are now clearly marked with a CHANGED tag.

    👋 Farewell ZipWhip integration, you’ll be missed!

    🔶 HUD validation leveled up with orange highlights for required 9902 fields.

    📱 Twilio SMS reporting gets a boost with added error codes.

    💳 Client-focused payment records now match their preferred Payment Type by default.

    🗓️ Keep track of ClientCredit due dates with new Original Due Day field.

    📞 Call Log Detail Report now includes Campaign fields for better Lead Source insights.

    💡 Dynamic Monthly Payment calculations made easy with Lead-specified details.

    🔍 Credit Report pulls now offer even more valuable information.

    📝 Automatic Budget item population from Credit Report pulls—saving you time!

    Changes – 2022 Q3

    Many minor changes and enhancements have been made to the CMA during Q3. Details are available in the internal CMA Changelog. 

    The following is a list of relevant changes:

    🤖 Daily Task creation made easy with our new automated Task Job and SQL magic!

    📏 HUD module updated to meet the fancy ARM6 V15.2.6.1 spec.

    💼 Payroll PTO rule entry streamlined for your convenience.

    📃 Document Template Management views now reveal connections to other docs—no more guesswork!

    🔍 AuditLog expands its horizons with additional tables.

    🏢 Even more HSCP goodness added to the HUD module.

    ✅ HUD 9902 Report submission now comes with extra validation logic for accuracy.

    🎡 Creditor-level FairShare logic is now simpler and more efficient. Enjoy!

    Changes – 2022 Q2

    Many minor changes and enhancements have been made to the CMA during Q2. Details are available in the internal CMA Changelog.

    The following is a list of relevant changes:

    📍 US zip code database now up-to-date with the 2022.03 release, keeping you on track!

    📦 Extranet Package Management got a makeover, making your job way easier.

    📄 LMA Paperwork Report now buddies up with Task Status report for extra insights.

    🏠 Housing Mean Data leveled up to 2022 values for accurate reporting.

    🔒 Remote logins to CMA will now auto-update PBX firewall rules for security.

    🎉 More Federal Holidays added to our Holiday table, so you never miss out!

    📞 Twilio integration: now better than ever!

    🖋️ Digital Signatures on Extranet get a sleek update.

    📑 Fee Schedule summaries made easy when a Lead’s address state changes.

    🔍 RPPS Code descriptions updated to keep you in the know.

    🏢 HSCP fields added to HUD module for extra functionality.

    👥 User management views are now packed with more helpful info.

    🚀 General report cleanups & optimizations for a smoother experience!

    New Feature: Waive NSF Fee based on Client’s Creditor

    There is a new System Setting: NSF_FEE_EXCLUDE_CREDITORIDS that allows listing CreditorIDs for which NSF Fees should not be charged.

    When importing the ACH failures, if a Client has an OPEN account that is associated with one of the CreditorIDs specified in NSF_FEE_EXCLUDE_CREDITORIDS, the following message is displayed “An OPEN Creditor prevents NSF Fee” and no NSF Fee is generated.

    New Feature: Custom SQL Reports

    We are excited to announce a new feature enhancement that will allow you to create custom SQL reports in our system. This feature will enable you to query data from various sources and generate reports that suit your specific needs.

    What is a Custom SQL Report?

    A custom SQL report is a report that you can create by writing your own SQL queries. SQL stands for Structured Query Language and it is a standard language for accessing and manipulating data in databases. By using SQL, you can select, filter, sort, group, and aggregate data from different tables and views.

    A custom SQL report can be useful when you want to:

    • Analyze data that is not available in the predefined reports
    • Combine data from different sources or applications
    • Perform complex calculations or transformations on data
    • Customize the layout or format of the report

    How to Create a Custom SQL Report?

    To create a custom SQL report, follow these steps:

    1. Go to System Management/Document Template Management
    2. Select Type = SQL
    3. Write your SQL query in the Query box. You can only use SELECT statements. Do not use INSERT, UPDATE, DELETE, or other commands that can modify data as they will generate an error.
    4. Click Save

    How to Access a Custom SQL Report?

    To access a custom SQL report, follow these steps:

    1. Go to the Navigation pane
    2. Select the name of your saved your report under Custom Reports
    3. Click on your report name
    4. If your report has filters, enter values for them and click Filter
    5. View your report on the screen or export it as a CSV file

    How to Limit Access to a Custom SQL Report?

    To limit access to a custom SQL report, follow these steps:

    1. Go to System Management/Document Template Management
    2. Select Type = SQL
    3. Find your report and click Edit
    4. In the Access Groups box, enter the permission groups that are able to view the report
    5. Click Save

    Only members of those groups will be able to see and run your report.

    We hope this feature enhancement will help you create more powerful and customized reports in our system.

    Thank you for using our system!