Generative AI in Colombian banking is already the present: at a workshop with Microsoft in Bogotá, Delto presented a study with Mercoplus LATAM covering 353 banked users; only 19% had used a WhatsApp chatbot, yet 74% would use an AI agent and 38% would switch banks for that experience.
On June 11 in Bogotá, alongside our partner Microsoft, we had the pleasure of hosting leaders and representatives from Colombia's most important banks at our workshop: Transactional generative AI in banking: future or present? Our goal was clear: to move beyond the hype and analyze, with data and concrete use cases, how artificial intelligence is transforming banking operations. In this blog post we share the main insights and conclusions from the event. From Artificial Intelligence to Agents We opened the presentation by reviewing some foundational concepts of the technology in question. We are constantly bombarded with information, so it is important to understand how generative artificial intelligence actually works. What happens behind the "magic"? We reviewed: LLMs and their training, tokenization, transformers, the differences between base models, instruction-tuned models, and products that use generative AI such as ChatGPT or a banking AI agent. The great paradox: high priority, low satisfaction We continued the workshop with market data that exposes an undeniable reality in the sector. While 89% of C-level executives (CXOs) consider AI and generative AI one of their top three priorities , 66% say they are dissatisfied with the progress they have made in their organizations. This gap between expectation and reality was our starting point. Why, despite the investment and the interest, are results not reaching the expected level? The answer, as we explored during the event, lies in the fundamental difference between AI that converses and AI that resolves . The evolutionary leap: from a conversational chatbot to an AI agent The first generation of conversational interfaces in banking focused on solutions capable of answering questions. However, their limitations are clear: heavily structured responses, no long-term memory, difficulty integrating with other systems, and a limited ability to execute complex transactional tasks securely and at scale. This is where we introduced the concept of Agentic Banking AI , a paradigm shift that unites what once seemed like opposite extremes: customer needs (24/7 availability, relevance, speed) with bank objectives (revenue growth, cost efficiency, better experience). McKinsey data reinforces this view, projecting a potential efficiency gain of nearly 40% in customer operations and positioning banking as the second industry with the greatest potential economic impact from generative AI. The voice of the customer: what is really happening in Colombia? To ground these global trends in the local reality, we presented the results of an exclusive study on conversational banking in Colombia , conducted together with Mercoplus LATAM on a sample of 353 banked users. Some of the insights we found: Low adoption of current chatbots: only 19% of users have used an automated chatbot via WhatsApp in the last 6 months. The main reason: Users do not use them because they prefer human interaction (39%), believe their queries are too complex for a bot (24%), or have had bad past experiences (14%). Current solutions are not delivering on the promise of being practical, agile, and effective channels. A huge opportunity: When presented with a generative AI solution capable of understanding their context and solving problems, 74% of users said they would be interested in using it. Even more telling: 38% would consider switching banks to access an experience at this level. The message is clear: Colombian customers do not reject AI, they reject a bad experience. They are ready and waiting for intelligent agents that genuinely solve their day-to-day needs. Delto in action: the platform to capitalize on the opportunity The second half of our workshop focused on how to turn this opportunity into reality. We showed that building transactional agents does not have to be a multi-year project. At Delto, we have built a solution specialized in customer operations for banking , a platform that lets financial institutions deploy intelligent agents quickly and securely. The key to our approach is a modular architecture based on skills. With a playbook of more than 300 pre-built skills for banking, our clients can automate everything from balance inquiries and transfers to complex processes such as loan renewals or lost-card management from day zero. We showed concrete examples from real cases, such as automated loan renewal , a case that cut the unit cost of renewal from 15 USD to just 2 USD. All of this running on Microsoft's secure and scalable Azure infrastructure, ensuring compliance with banking regulations. The event at Microsoft's offices not only confirmed that transactional generative AI is the present of banking, but also made clear the urgency and the immense opportunity that exists in the Colombian market. We want to extend our deep thanks to all the banks that joined us and to our partner Microsoft for making this gathering possible together with us. The conversation is just beginning, and at Delto we are committed to continuing to push the limits of what is possible to build, together, the next generation of interaction between customers and banking. Related reading: the regulatory framework for conversational AI in LATAM · automated credit renewal · a specialized solution for customer operations in banking .
Why don't Colombian users use today's banking chatbots? According to Delto's study with Mercoplus LATAM (353 users), only 19% used a chatbot via WhatsApp in 6 months. They avoid them because they prefer human interaction (39%), believe their queries are too complex (24%), or had bad past experiences (14%). They don't reject AI, they reject a bad experience.
What sets an AI agent (Agentic Banking AI) apart from a conversational chatbot? A chatbot answers questions with structured responses, no long-term memory, and no ability to execute complex transactions. An AI agent resolves: it executes transactional tasks securely and at scale, uniting customer needs (24/7 availability, speed) with bank objectives (revenue, efficiency, experience).
How much can Delto reduce banking operations costs? Delto runs on a playbook of more than 300 pre-built banking skills on Microsoft's Azure infrastructure. In a real automated loan-renewal case, the unit cost dropped from 15 USD to 2 USD. McKinsey projects up to ~40% efficiency in customer operations with generative AI.
What role does the Superintendencia Financiera play in AI adoption by Colombian banks? The SFC actively audits AI use without waiting for a law: it created a Center of Excellence in Artificial Intelligence, runs the LaArenera and elHub sandboxes and launched its own money laundering detection tool. A bank needs data governance, documented human validation and full traceability of its agent conversations.
Do banking customers in Colombia want to use generative AI agents? Yes. In Delto study with Mercoplus LATAM covering 353 banked users, 74% said they would use a generative AI agent capable of understanding their context and solving problems, and 38% would consider switching banks to access that experience. The barrier is not AI itself, it is the poor experience of current chatbots.