From generative AI to agentic intelligence: What is really changing?
Most people today think of AI as a chatbot that answers questions or generates text. But that is only scratching the surface. Banco do Brasil has decided to go further and implement agentic AI. The difference between regular generative AI (like basic ChatGPT) and agentic AI is fundamental: while generative AI can create text, agentic AI can plan, use tools, and execute tasks within a defined workflow.
In the context of Banco do Brasil, this means that the NICE Copilot system is not just an "intelligent search engine." It is an active assistant capable of analyzing a client's history, summarizing previous interactions, detecting the customer's current sentiment, and suggesting in real time the most suitable next step for the banking advisor. This shift from "answering" to "acting" is the key trend that is now beginning to dominate the fintech sector.
How does NICE Copilot work in practice?
The implementation of the system within the NICE CXone platform brings bank employees several critical features that directly affect service quality:
- Automatic interaction summaries: Instead of spending hours going through notes from previous meetings, the manager gets an immediate, concise, and accurate overview of what the client needs.
- Real-time sentiment analysis: AI monitors the tone of voice or text of the client. If the system detects frustration, it can immediately signal to the advisor the need to change approach or escalate the situation.
- Contextual guidance: The system delivers relevant information exactly at the moment it is needed, minimizing time spent searching internal systems.
- Ensuring compliance: In banking, a documentation error or incorrect information is critical. NICE Copilot creates standardized and traceable records of interactions, making audits and rule-following easier.
Comparison with market competitors
NICE finds itself in direct competition with giants like Salesforce and Microsoft in this battle for AI dominance in customer experience (CX).
While Salesforce Einstein GPT focuses heavily on CRM and predictive analytics for salespeople, NICE has the advantage of deep integration into omnichannel communication (phone, chat, email) through the CXone platform. Microsoft Dynamics 365 Copilot bets on seamless integration into the entire Office 365 ecosystem, which is attractive for large corporations. NICE, however, profiles itself as a specialized tool for complex customer care, where the ability of agentic AI to work with data from many different channels simultaneously is key.
If we look at performance metrics (benchmarks), in the area of context understanding and accuracy in banking tasks, systems like NICE operate at the level of top models like GPT-4 or Claude 3.5 Sonnet, but with the difference that they are "wrapped" in a secure, closed sandbox environment that protects sensitive banking data.
Pricing and availability: What does it mean for businesses?
It is important to realize that NICE Copilot is not a "SaaS for individuals" tool with a monthly subscription of 20 USD. It is a comprehensive B2B enterprise solution. The price is usually determined by an individual contract tailored to the size of the bank and the number of users (agents). For middle-market companies, the costs of implementing such a system are significant, covering not only the license but also integration with the bank's existing legacy systems.
Availability in the Czech Republic: As a global provider, NICE has clients all over the world, including Europe. For the Czech market and Czech banks (e.g., ČSOB, Komerční banka, or Air Bank), implementing a similar system would mean integration into an environment that must be fully compliant with the EU AI Act.
Impact on the Czech and European market: Regulation is key
While Banco do Brasil operates under the supervision of Brazilian central banking, European institutions must address the strict regulation of the EU AI Act. Under this regulation, banking systems are often classified as high-risk AI systems. This means that every implementation of agentic AI, similar to NICE's, must meet extreme requirements for:
- Transparency: We must know why the AI decided or suggested this particular step.
- Traceability: Every interaction must be logged and auditable.
- Privacy protection (GDPR): Agentic AI must not uncontrollably "siphon" clients' personal data for training models outside a secured environment.
For Czech banks, this means that the journey to agentic AI will be slower than in Brazil, but at the same time much more focused on security and trustworthiness. Companies that can implement these technologies in compliance with European rules will gain a massive competitive advantage in efficiency while maintaining customer trust.
Can agentic AI replace human banking advisors?
No, not currently or in the near future. The goal is "augmentation," not replacement. AI handles routine tasks, summarizes data, and prepares materials so that the human advisor can spend more time building the client relationship and solving complex problems that require human empathy and ethical decision-making.
Is working with sensitive banking data in these systems safe?
Yes, systems like NICE Copilot are designed for the enterprise segment. This means that data is not used to train public models (as is the case with the free version of ChatGPT). Data remains within the bank's secured environment and meets strict cybersecurity standards.
What are the main benefits for the average customer?
For customers, this means shorter wait times, less repetition of information with every contact, and more personalized services. When a banking employee immediately knows what the client needs thanks to AI, the interaction is smoother and more efficient.
