Yesterday Salesforce held its traditional Agentforce World Tour Madrid 2026 event in Madrid where the multinational has attracted more than 3,500 professionals who have been able to verify Salesforce’s vision around the company’s cornerstone pillar: agentic AI.
In this context, Marco Hernansanz, EVP & CEO of EMEA at Salesforce, was in charge of giving the main presentation. The basis of his speech has been based on the fact that business artificial intelligence is at a critical point. Despite the widespread enthusiasm, the reality is that the 95% of AI pilot projects fail to reach production. This figure, supported by an MIT study and Salesforce’s direct experience with its local customers, reveals a growing gap between expectations and tangible results.
According to Hernansanz, “this massive failure is not due to a lack of good ideas. The problem lies in a wrong approach. Many organizations have focused their efforts exclusively on language models (LLM), forgetting that these are only a small part of the total solution.” The manager explained that “an LLM alone is like a ‘Christmas brother-in-law’: someone who speaks with complete confidence on any topic, but who is often wrong because it lacks a real basis. A ‘disconnected’ LLM does not know the customer, does not know the product and does not have access to the company’s critical data and metadata. These models can be confidently erroneous,” Hernansanz warned, pointing out that this lack of precision and security is what prevents AI from being deployed in industries. regulated such as banking or healthcare.
Salesforce’s response: ora four-system architecture
To close this gap and allow companies to scale their projects, Salesforce proposes abandoning technological “DIY” and adopting an architecture based on four fundamental pillars integrated into its platform. agentforce.
The first pillar is Context System. For an AI agent to be useful, it must “know what it is talking about.” Hernansanz gives as an example an airline that manages the loss of a suitcase: the agent needs real-time data about the customer, the flight and the location of the luggage. Without this context, the AI is unable to offer a valid response. The second pillar is Work Systemwhich represents the logic and process flows that the company has perfected over years. This is what allows the agent not only to inform, but to act, such as activating the systems necessary to grant bank credit.
The two remaining pillars are the Agent Systeman infrastructure to create, test and monitor the behavior of AI, and the Interaction Systemwhich facilitates real-time collaboration between humans and agents. Hernansanz emphasizes that this is an “open architecture”: companies can integrate their own Data Lakes and their pre-trained language models. “The power is in the connection,” he says, urging companies not to leave their data locked in inaccessible “vaults,” but to put it at the service of employees who interact with customers.
Beyond technology: Values and a new economy of abundance
The success of this transition towards what Hernansanz calls the “agent enterprise” does not depend solely on the technical architecture, but on the foundations on which the organization is built. Salesforce bases its entire strategy on five core values: trust, customer success, innovation, equality and sustainability. In the age of AI, trust becomes a critical value that must be applied from the data level to the ethical use of technology.
The true revolution, as Hernansanz explained during his speech, is the paradigm shift at work. Historically, companies have operated under constraint economieslimited by rigid budgets, difficulties in expanding teams quickly and constant time pressures. Agentic AI breaks these barriers by functioning as an “infinite virtual workforce” capable of reasoning and orchestrating tasks. “We are moving from an economy of restrictions to an economy of abundance,” stated the manager.
This change is urgent due to an increasingly demanding macro and micro environment. At a macro level, geopolitical instability and competition from less regulated countries are putting pressure on European companies to act. At a micro level, customers no longer settle for less than what they get in modern conversational environments, and management teams have stopped being interested in innovation “for the love of art” to demand a real return on investment (ROI) that impacts the bottom line.
Hernansanz ended his speech by assuring that the experimentation phase with AI has ended. With competitors, customers and employees moving at high speed, the risk is no longer innovating, but rather taking too long to start and being left out of the new agentic economy. Those companies that manage to integrate context, work, agents and interaction will be the ones that lead this new era of abundance.
Put into practice
The success stories have been the main protagonists of the event. Representatives from CaixaBank, Fever and the European University participated in the main keynote. In the exhibition area, Agentforce City has been the epicenter of the Agentforce World Tour Madrid 2026, being the most prominent and ambitious space of this edition. Conceived as an immersive environment, this exhibition area has been dedicated to showing the reality of the Agentic Company with agents in production. The companies participating in Agentforce City were the following:
– AEDAS Homes is implementing autonomous agents in two key areas: Lara as a Promotions Advisor and Félix for the After-Sales Service. Agent Lara is trained to respond to queries from potential clients about any real estate project, inform about promotions and coordinate the first visit to the point of sale.
– CaixaBank is implementing AI agents to offer specialized assistance to customers and employees. A main use case is a remote hiring support assistant, which helps managers reduce interaction time and drive product hiring from digital channels (app and web). In Agentforce City, CaixaBank presents an agent to support call center employees in serving customers who request information about loans online.
– IBERIA is developing various use cases with Agentforce, including solutions to increase the productivity of its Contact Center agents and reduce service times. In the case of Agentforce City, the airline presents SofIA, an agent at the service of employees that makes it easier for People to answer questions about any aspect related to the company.
– European University uses Agentforce as an advisor for Educational Programs for students. Its use case focuses on the development of a recommender for degrees, masters or other customized training programs and answering frequently asked questions. The agent also allows you to pre-qualify the interest of potential students and escalate the case to a human agent.
– Pikolin is using Agentforce to develop an expert technical advisor that resolves complex questions about rest materials and technologies, facilitating an assisted purchasing decision, automating the response on delivery times and allowing the customer to decide with all the technical information at their disposal.
– Repsol uses Agentforce from Salesforce to transform the experience and communication channels with its customers, evolving from traditional bots to natural language interactions capable of understanding the context and executing complex transactions. This technology is already applied in Sales, with the conversational management of LPG (butane) orders via WhatsApp, which facilitates personalized self-service; in Customer Service, with intelligent support integrated into the contact center that helps managers and proposes adapted responses; and on the web, where customers have an AI chat to resolve queries.
