CallMiner, Invoca, Level.AI, Observe.AI, Tethr, and MiaRec [Comparisons, Pricing, Features, & Reviews]

Sentiment Analysis: CallMiner analyzes sentiment using NLP. It identifies predefined keywords and monitors acoustics to determine a sentiment score. Emojis and special characters are included in omnichannel analysis.

Topic Analysis: CallMiner's Eureka Analyze tool offers keyword and NLP-based topical analysis. This method requires syntax expressions to define keywords and phrases, making it difficult to set up and less accurate, as it doesn't consider the entire conversation. However, we expect CallMiner to catch up with generative AI-based topic analysis soon.

AI Assist: CallMinor's RealTime Assist delivers real-time assistance for agents and supervisors. RealTime will listen for keywords in your conversations and alert or notify the agent or supervisor once it deems intervention is required, e.g., alerting a supervisor of a conversation detailing or identifying an upsell opportunity.

Auto Redaction: CallMiner also offers audio and text-based data redaction.

Prompt Designer: CallMiner does not offer prompt designer capabilities at the moment.

Call Recording: Invoca offers call recording. However, their website and other public sources do not reveal more details on the features.

Transcription Accuracy: Some users complained that "the transcription really needs to be improved" as the "transcriptions are often inaccurate and leave a lot to be desired." However, since then, Invoca released its new transcription engine which is AI/LLM based and was trained on over 700,000 hours of supervised audio files.

Live Monitoring: Not offered.

Screen Recording: Not offered.

Manual Agent Evaluation (Scorecards): Not offered.

Auto QM (Auto Scorecards): Invoca offers automated call scoring at an additional cost as it must be bought as a Quality Assurance & Agent Coaching add-on. However, it is worth noting that these capabilities are very new and publicly available information is very limited.

Sentiment Analysis: Invoca offers sentiment analysis at an additional cost through its Signal AI product. Signal AI "uses fine-tuned neural network models to categorize caller and agent sentiment separately, including how it changes throughout the call."

Topic Analysis: Invoca offers some AI-driven (GPT-based) topical analysis as part of its Signal AI product, but its primary focus is to uncover marketing and sales trends, topics, and issues using NLP-based keyword spotting, e.g., conversions by marketing channel. 

AI Assist: Invoca's Signal AI product offers GPT-generated call summaries using predefined prompts. Employees can also use a "GPT-powered assistant to get answers, suggestions, and content from calls" to improve marketing messaging, agent coaching, and more.

Auto Redaction: Signal AI includes automatic data redaction.

Prompt Designer: Invoca does not offer prompt designer capabilities at the moment.

Call Recording: Not offered.

Transcription Accuracy: Level.AI claims to have a transcription accuracy of 85%. which is comparable to MiaRec's accuracy rate.

Live Monitoring: Not offered.

Screen Recording: Level.AI offers a product called Agent Screen Recording that automatically records an agent's screen from the time a ticket has been assigned to the moment it is closed. You can also capture multiple screens and post-call wrap-up time.

Manual Agent Evaluation (Scorecards): Level.AI's product brochure mentions that its algorithm can assign conversations for manual review based on certain criteria, but it does not specify how the manual evaluations can be done within Level.AI.

Automated Quality Management (Auto QM or Auto Scorecards): Level.AI's Auto QA uses Natural Language Understanding (NLU) and a Large Language Model (LLM) trained on their customers' contact center data to automatically score an agent's performance across all phone calls, emails, and chats. Multiple customers have mentioned that they had to wait several months for the LLM to be trained before they could start using the solution. In addition, Generative AI is so precise that training an LLM on a contact center's data shows marginal to no advantage.

Sentiment Analysis: Level.AI uses Generative AI to score "multi-emotion" customer sentiment (meaning mildly positive vs. very negative) on a 0-10 scale. However, it does not score agent sentiment.

Topic Analysis: Level.AI's VoC Insights mines all customer interactions (calls, chat, email) using its proprietary Generative AI-powered topical analysis.

AI Assist: The company's AI Assist capabilities surface answers from a contact center's internal knowledge base to provide agents with answers in real time. It self-learns and can improve itself based on agent feedback.

Auto Redaction: Level.AI offers automatic data redaction.

Prompt Designer: Level.AI does not offer prompt designer capabilities at the moment.

Call Recording: Not offered.

Transcription Accuracy: Observe.AI's LLM is 35% more accurate in conversation summarization and 33% in sentiment analysis. (Source)

Live Monitoring: Not offered.

Screen Recording: Observe.AI has offered Screen Recording as part of its Conversation Intelligence platform since June 2023.

Manual Agent Evaluation (Scorecards): Observe.AI's manual scorecard capabilities are comparable to MiaRec's in terms of functionality and flexibility. QA managers can create fully flexible evaluation forms. They are filled out within context (call metadata, transcript, etc.). Observe.AI also adds coaching tips and additional contextual information.

Auto QM (Auto Scorecards): Observe.AI's Conversation Intelligence solution offers automatic call scoring. However, its demo video shows a setup using phrases rather than Generative AI. Also, users noted in reviews that there is limited reporting and no trend analysis.

Sentiment Analysis:  Observe.AI offers highly accurate sentiment analysis. The solution utilizes tone, speech rate, speaker volume, silence, and overtalk to decipher positive and negative customer sentiment.

Topic Analysis: Observe.AI's approach to topical analysis is slightly different. You can uncover why your customers are calling by looking at specific dashboards, e.g., sales and operations, but it seems there is no topical analysis as a stand-alone function.

AI Assist: Observe.AI provides strong, generative AI-based Auto Call Summary based on a proprietary LLM, real-time agent assist (walking through call scripts and playbooks to improve compliance and upsell), and the ability to automatically extract knowledge from your knowledge bases and previous calls.

Auto Redaction: Observe.AI offers automatic data redaction.

Prompt Designer: Observe.AI does not offer prompt designer capabilities at the moment.

Call Recording: Not offered.

Transcription Accuracy: Customer reviews mentioned that Tethr's transcriptions are only 80% accurate, and sometimes, "it might not catch phrases even though the phrases are relevant for a category."

Live Monitoring: Not offered.

Screen Recording: Not offered.

Manual Agent Evaluation (Scorecards): Tethr's Evaluation Forms allow QA managers to build forms for manual evaluation. Users can assign points to each question and answer, and these points integrate back into Tethr's score-based reporting.

Auto QM (Auto Scorecards): Tethr mainly relies on its two proprietary AI-derived metrics: Tethr Effort Index (TEI), which predicts customer loyalty and churn based on 250 independent variables and 10,000+ combinations of distinct phrases and utterances, and Agent Impact Score (AIS), which reflects the agent performance with 160 variables tracked.  However, according to multiple customer reviews, some customers like the KPI-based approach, while others find it too rigid and too much of a black-box.

Sentiment Analysis: Tethr's sentiment analysis model is trained to identify 28 emotions, including frustration, confusion, or apprehension, based on customer and agent phrases.

Topic Analysis: Tethr's solution offers NLP keyword-based use case-based topical analysis, including customer churn, revenue increase, etc. However, keyword-based topic analysis is often less reliable than generative AI-based analysis.

AI Assist: Tethr Live detects "response triggers" based on hundreds of words customers and agents frequently use. If detected, Tethr can alert supervisors or suggest knowledge-based articles, complaint checklists, and more.

Auto Redaction: Not offered.

Prompt Designer: Tethr does not offer prompt designer capabilities at the moment.

Call Recording: MiaRec offers compliant, secure, and reliable call recording as an add-on to its AI-powered Conversation Intelligence suite.

Transcription Accuracy: To provide the best possible transcription accuracy at an affordable price, MiaRec utilizes its own proprietary AI-based speech recognition engine for speech-to-text transcriptions with an accuracy of 85%.

Live Monitoring: MiaRec offers live monitoring as part of its lowest package, Quality Management Pro. It allows supervisors to monitor ongoing calls in real-time or consecutive calls of a particular agent over time.

Screen Recording: Customers can purchase MiaRec's screen recording as an add-on to capture and playback synchronized agent desktop and call activity for a comprehensive view of customer interactions.

Manual Agent Evaluation (Scorecards): Using MiaRec's fully flexible form editor, QA managers can create manual evaluation scorecards that can be used as a stand-alone solution for smaller contact centers or in addition to auto call scoring for particularly sensitive calls or interactions that require human oversight. MiaRec customers can accelerate their manual evaluation process by 40% compared to other tools.

Auto QM (Auto Scorecards): MiaRec offers accurate auto call scoring of 100% of your calls. In contrast to some other vendors, MiaRec allows you to write your own prompts, making it more targeted to your individual business needs. Detailed reporting allows you to determine coaching opportunities, measure the effectiveness of training, and much more.

Sentiment Analysis: MiaRec's Sentiment Analysis is NLP-based, which means that its model can be trained to read beyond mere definitions to understand things like context, sarcasm, and misapplied words. Both the customer's and the agent's sentiments are expressed in a numerical score as well as visually by color and as an emoji.

Topic Analysis: MiaRec's Topical Analysis can automatically categorize calls belonging to a certain topic or category to analyze call drivers, catch potential problems faster, and vastly improve customer experience. 

AI Assist: MiaRec, with its AI Assist capabilities, focuses on assisting agents post-call rather than in real-time. By using AI call summaries, agents can not only cut post-call work drastically but also focus more on the customer rather than taking notes. In addition, you can use AI Insights to extract any type of structured or unstructured information from your calls using natural language.

Auto Redaction: MiaRec offers highly accurate NER-based auto data redaction.

Prompt Designer: MiaRec's AI Prompt Designer is designed to enable users to create and refine prompts within a secure testing environment before deploying them into their live environment. The ability to customize prompts translates into enhanced call scoring accuracy, insightful call summaries, and targeted agent coaching, addressing critical areas unique to your business.