Establishing AI Rooted in Practical Applications and Users

NEW YORK, Sept. 16, 2025 /PRNewswire/ — Sobot, an innovative AI customer contact solution provider, has released the latest testing results of its updated Sobot AI. With participation from over 100 customers, Sobot AI achieved an average response accuracy rate of over 90% and an independent resolution rate of 88%.

The driving force behind this improvement is what Sobot calls the synergy of Generative AI and Multi-faceted AI. Rather than simply pursuing sophisticated AI technologies, Sobot focuses on practical innovation in how AI is applied. This strategy is best reflected in Sobot’s “Five-AI” system, where Generative AI and Multi-Faceted AI play critical roles.

Sobot Generative AI Ensures Industry-Specific Accuracy

Among all applications of AI in customer service, Generative AI is one of the most widely recognized. But unlike conventional systems that rely on rigid rule-based responses, Sobot Generative AI delivers accurate, industry-specific and human-like answers.

RAG+LLMs: The Core Capabilities of Sobot Generative AI

At the heart of Sobot Generative AI is Retrieval-Augmented Generation (RAG), combined with advanced large language models (LLMs). While the regular RAG process follows familiar steps like chunking, embedding, retrieval and generation, Sobot AI introduces several innovations that set Sobot apart.

Intelligent Chunking: Instead of relying on rigid fixed-length splitting, Sobot AI supports semantic and structure-aware chunking to preserve complete meaning, ensuring that AI can understand the context more deeply and retrieve the right content more accurately.

Precise Retrieval: Sobot AI automatically rewrites customer inquiries for clarity, then matches them with the most relevant content that meets the pre-set “recall threshold”. Advanced retrieval and re-ranking further ensure that the best-fitting results are selected.

Enhanced Generation: Sobot AI has integrated with multiple advanced LLMs including Claude, OpenAI, Amazon Bedrock, DeepSeek and more rather than just depending on a single model, generating more fluent and natural answers. Also, with the fine-tuned talking tones and styles, the answers sent to customers always sound clear, professional and aligned with the brands.

All of these help Sobot AI reduce LLM hallucination and improve answering accuracy.

Specifically Designed SLMs for Targeted Services

Apart from traditional LLMs, Sobot has also introduced several Small Language Models (SLMs) to further enhance the industry-specific accuracy of the AI answers. Rather than conducting broad reasoning, SLMs make Sobot AI more targeted and task-oriented.

“SLMs are designed for solving specific problems in various industries. Take retail and ecommerce industry as an example, our SLMs support businesses with common scenarios like order tracking, product recommendations, returns, refunds and more.” said Yi Xu, CEO of Sobot.

With the combined strengths of RAG, LLMs, and SLMs, Sobot Generative AI delivers answers that are not only accurate but also rooted in real-world business contexts.

Sobot Multi-Faceted AI: Applicable for Different Roles

While Generative AI promotes the intelligence of Sobot AI, Multi-faceted AI empowers different groups of people to use Sobot AI more intelligently. Sobot Multi-faceted AI consists of 3 key components: AI Agent, AI Copilot and AI Insight.

AI Agent: Provide Customers with More Satisfactory Services

At the core of every customer contact solution is the customer experience itself. That’s where Sobot AI Agent comes in. Sobot AI Agent delivers fast and human-like services across various channels, including chat, voice, email, social media and more. It intelligently and independently answers customers’ questions, supports flexible settings such as tones, answering styles, and sentence length, and provides multilingual support to meet global business needs.

AI Copilot: Assist Human Agents in More Efficient Work

Despite the popularity and advancement of AI, human agents remain essential in customer service, especially for complex issues. To facilitate human agents, Sobot has introduced AI Copilot for comprehensive assistance. Sobot AI Copilot helps agents in conversation summary, content expansion and polishing, one-click ticket filling and more, covering the complete journey of customer service. Rather than replacing agents, Copilot enhances their capabilities, lightens their workload, and boosts their confidence in delivering an excellent and efficient service to customers.

AI Insight: Support Administrators for Better Adjustment and Decision Making

Beyond daily interactions, customer service requires ongoing measurement and refinement. Sobot AI Insight consolidates reports, analytics, intelligent quality inspection, voice of customer (VOC) analysis and more data in a single dashboard, empowering administrators to observe over 300 indicators at one time to evaluate the performance of customer service. It gives administrators the intelligence they need to make timely adjustments and data-driven decisions.

Sobot AI: Built on Real Applications, Designed for Real Users

Looking ahead, Xu emphasized that AI for customer service should not be seen merely as a tool of automatic question answering, but as a partner that reshapes customer experience throughout the service journey. By combining Generative AI and Multi-Faceted AI, Sobot has managed to deliver an integrated customer contact solution that is deeply rooted in practical applications and diverse roles of users.

With this approach, Sobot is not only addressing today’s service challenges but also laying the foundation for the next era of intelligent, user-centric customer contact.

“Sobot AI is not just the stacking of functions, but the deep integration with applications and users. That’s where we stand out.” said Xu.

For more information, please visit https://www.sobot.io/.