Generative AI: The New Era of Customer Service
Companies across many sectors are already applying generative AI to boost efficiency, retention and service quality, but how did this tool make its way in?
Theo Tabach (Senior at GN Poli)

Considered the new technology of the decade, Artificial Intelligence is increasingly present in our daily lives, with a very simple purpose: to cut the time of tasks that would normally take a long time down to just a few seconds. Although it also has creative capabilities, its fast information processing and application of algorithms are what make AI a very useful tool, increasingly used by companies across a wide range of sectors.
Among the many sectors affected, one of those that benefits the most is customer service. Unlike traditional chatbots, based on fixed rules, generative AI understands natural language, adapts its tone of voice and generates contextualized responses. This makes it ideal for automating inquiries, updating information and assisting human agents in real time. The result is faster, more consistent and more scalable service. Companies that have adopted it are already recording tangible gains in productivity and customer satisfaction.
According to an IBGE estimate, 41.9% of companies with 100 or more employees use Artificial Intelligence in their daily operations, with the areas that make the most use of AI being Administration (87.9%), Sales (75.2%) and the Development of product, process and service projects (73.1%). But does the tool only bring operational benefits to companies, or are there risks and downsides associated with its use?
Case Study – Bradesco
BIA (Bradesco Inteligência Artificial) is Bradesco's virtual assistant, which was integrated with generative AI technologies to improve digital customer service. In 2024, the bank launched its platform called Bridge (“Bradesco Inteligência de Data Generativa”), used to unify AI services, such as voice, text and document processing, as well as to serve as a foundation for assistants and to implement responsible AI mechanisms. Its application in BIA has already reached very significant levels of usage: interactions with more than 24 million users in the bank's app and around 3 million customers served by the generative AI version of BIA, with an estimated resolution rate of 90%.
The results in efficiency and customer satisfaction are also significant. BIA retains between 85% and 90% of requests in digital chat service, which means that only 10% to 15% of requests need to be forwarded to a human agent. The satisfaction score is between 4.1 and 4.2 on a scale of 0 to 5 for these digital interactions, indicating that customers have perceived quality in the automated service. In addition, in the collections area, BIA with generative AI brought gains of 10% to 20% in credit recovery capacity, through agents that analyze conversations and guide operators with best practices.
On the other hand, there are challenges and risks identified in this application. One problem is that, in very specific cases or with technical, legal or uncommon vocabulary, the AI may not understand well, requiring escalation to human reviewers. There are also risks associated with data use and security, for example in fraud prevention: BIA was used to confirm suspicious Pix requests via WhatsApp, reducing fraud, but this requires care in handling customer data and in the channels used. Another point of attention is accuracy: for the high resolution rate to be maintained, constant monitoring of the models, exception handling and human feedback are necessary.
Case Study – Inspiro
The partnership between Tiger Analytics and Inspiro, a customer experience outsourcing company, involved the creation of a generative AI-enabled knowledge base tool, developed with Azure and OpenAI technologies, for a major airline in the Philippines. Previously, agents used a simple keyword-based search tool with rigid workflows, for example: typing “baggage” and viewing documents containing that term. With the new solution, all it takes is asking a free-form question (“What is the baggage allowance for my flight X?”), and the system returns not only the relevant document, but also the page number and the main references used to generate the answer, in a much more flexible way.
The practical results were very significant: the knowledge base response time dropped to 3 to 5 seconds, instead of minutes; average handling time was reduced by 30.5% for agents working via chat/text (not voice) and by 10.8% for voice agents. Answer accuracy increased by about 20%, while quality control (“quality scores”) improved by approximately 12.3% in chat service and 5.4% in voice service. This gain shows that the tool not only sped up service but also raised its quality.
Although the case does not mention serious failures, it is possible to identify some points of attention. It is essential to keep the knowledge base up to date to avoid incorrect answers; to clearly define when to call in human support; to protect internal data; and to ensure agents' trust in the answers generated by the AI.
Comparison – AI, Chatbot, Call Centers
Customer service has evolved significantly, moving from the traditional model of human call centers, centered on empathy but costly and time-consuming, to rule-based chatbots, which brought automation but with limitations in understanding natural language. Today, generative AI represents the next stage of this transformation: it understands context, adapts to the customer's tone and provides more complete and accurate answers. This technology combines efficiency and personalization, reducing handling time and operating costs while keeping the user experience smooth and close to human service.
Sources: Gartner, Microsoft, McKinsey

As the chart shows, the use of AI in customer service can bring significant gains in efficiency and scale as well as in cost and success rate. However, using this tool also involves significant risks when applied carelessly. One of the main dangers lies in the exposure and misuse of sensitive data. Since AI systems handle large volumes of personal, financial and behavioral information, any security failure or lack of clear privacy policies can lead to leaks, compliance violations and damage to the company's reputation.
In addition, poorly trained AI models, or those without adequate supervision, can generate inaccurate, biased or incoherent responses, hurting the customer experience and undermining trust in the brand. Another risk is excessive automation, when companies completely replace human service with artificial systems, resulting in impersonal interactions that are unable to handle emotional or complex situations. Finally, a lack of transparency and ethical oversight can lead to unfair automated decisions, algorithmic discrimination and legal consequences, reinforcing that the adoption of AI in customer service must always balance efficiency with responsibility, empathy and data governance.
In short, AI is a very powerful tool, already applied by the world's largest companies, but without rigorous oversight of its activities, it can significantly compromise the quality of customer service.




