SentiSquare: No-Code NLP platform for text understanding
Turn emails, tickets, chats, call transcripts, or feedback into usable data. SentiSquare uses artificial intelligence (NLP) to understand unstructured text and automatically sort, categorize, and evaluate it (topics, keywords, sentiment)—without the need for programming and without your teams having to read everything. In low-code applications built on TotalAgility, we use SentiSquare as an AI input layer—text becomes structured data that the application can then process further.
No-code control
The platform is designed for business teams—you set up categories, rules, and scenarios without having to write code or be an NLP specialist.
Analysis of any language
SentiSquare evaluates texts across languages (including less common ones) thanks to a language-agnostic representation of meaning. You can use one model approach for multiple markets and teams.
Models based on your data
AI models learn from your real-world texts and adapt to your terminology, slang, and typos. Thanks to semi-supervised learning, it achieves high accuracy, often approaching human levels.
Stable operations and exceptions
Built-in features for handling exceptions and combining models help maintain quality in daily operations—even as topics and trends change.
How the SentiSquare platform works in practice
Text communication contains a vast amount of valuable information. However, there is often so much data that it is impossible for humans to process it all. Within companies, this data is frequently chaotic: language is full of ambiguity, irony, typos, and abbreviations; every company has its own vocabulary; and topics change over time. Manual processing is therefore unsustainable and unscalable—teams usually manage to review only a small fraction of content, and in the case of calls, often only about 1% of recordings are evaluated. The result is blind spots: recurring problems, growing customer frustration, churn signals, high-risk situations, or new trends that emerge before anyone can catch them.
The SentiSquare platform uses AI (NLP) to convert unstructured text into actionable data: it automatically recognizes topics, keywords, sentiment, and urgency, and supplements them with the business categories you need for your processes and reporting. SentiSquare is built on language-agnostic algorithms that work across different language groups, including those with rich morphology (e.g., declension, conjugation). This allows you to easily apply one approach across multiple teams and markets. And because it is a no-code platform, the basic setup of categories, rules, and scenarios can be handled by business teams without any programming.
We use SentiSquare as an AI supplement for digital processes and applications where text content is the deciding factor:
▪️Digital Mailroom – understanding email intent and sorting without manual rules
▪️Customer and Service Processes – request qualification, prioritization, and escalation
▪️Internal Requests and Approvals – identifying request types and routing
▪️Reporting and Management – trends, sentiment, and early warning signals for problems or churn
In practice, this means that emails, tickets, chats, call transcripts, and text feedback stop being an unreadable pile of content and become a consistent source of data across all channels. This translates into specific scenarios:
▪️In email and ticketing, SentiSquare filters out messages requiring no action (spam, out-of-office, empty queries) so operators can focus on what matters.
▪️For incoming communication, it performs qualification and routing: it classifies topics and urgency and forwards the request to the right team or specialist—in real-time, without manual sorting.
▪️For recurring queries, it prepares an automatic response or a draft response from a template for the agent to simply review and send.
▪️For management, it offers AI analytics covering 100% of communication instead of working with a small sample: trends over time, most frequent contact reasons, sentiment, topic evolution, early churn signals, fraud, or hidden product and service issues.
Technologically, SentiSquare is built on distributional semantics (meaning derived from context) and combines pre-trained language knowledge with unsupervised pattern discovery and supervised or semi-supervised learning. This allows AI models to adapt to your company's language and interpret text exactly as your operations require—from simple sorting to complex automation and reporting.
Text collection and preparation
We connect SentiSquare to your channels and systems where text communication originates—such as email, helpdesk, chat, CRM, call transcripts, forms, or surveys. Together, we also define your base unit of analysis: an entire email thread, a single ticket, a message, or a call.
Topic mapping
The system automatically identifies recurring patterns and topics in your data. This phase is useful when new issues and trends emerge or when you need a quick overview of what your customers are actually dealing with.
Model training
We fine-tune classification and interpretation on your data to match your specific operations. We set up business categories, request types, contact reasons, sentiment, urgency, churn signals, and spam detection. We also adapt the models to your company's language, including product slang, service names, and internal abbreviations.
Deployment and exception handling
We connect the models with rules and follow-up steps to ensure the outputs have practical utility. This typically involves automatically closing irrelevant emails, routing to the correct team, suggesting template responses, or reporting on trends. In operation, you then monitor changes in topics and the quality of results, continuously fine-tuning the models.
Frequently Asked Questions
Here you will find answers to the most frequently asked questions about the SentiSquare platform.
No-code means that categorization, rules, scenarios, and exception handling are set up directly by business teams without the need for programming. Technical integration with channels and systems is handled during deployment, while subsequent management and fine-tuning take place within the platform interface. NLP stands for Natural Language Processing—AI that can understand text, categorize it, and extract information (topics, reasons for contact, sentiment, urgency).
SentiSquare uses AI to understand meaning from context, and our models adapt to your specific data, not just general language. This allows the system to capture company-specific vocabulary, product slang, and the typical phrasing used by your customers. In practice, this reduces classification errors and improves the consistency of results in daily operations.
SentiSquare algorithms are language-agnostic, meaning they remain functional across different language groups, including those with complex morphology. This is crucial when you serve multiple markets or operate in multiple languages within a single organization. A single approach can thus be easily scaled across teams and regions.
This typically includes emails (including threads), helpdesk and ticketing communication, chats, CRM comments, call transcripts, form responses, or text feedback from surveys and reviews. Importantly, we can analyze both short messages and longer conversations. We also always define what constitutes a single case for you—for example, an entire ticket or an email thread.
We configure security based on how you currently handle communication: access levels and roles, audit trails, and clear rules on who sees what. For sensitive data, we typically address data minimization, access segregation, and, if necessary, anonymization in accordance with internal policies and requirements (e.g., GDPR). We always fine-tune the specific settings to match your processes.
Space for another question
Have any questions?
I am here to help
Jiří Plešek
Senior Sales Manager
