Here’s the translation into American English:
—
A recent report by Gartner has revealed that more than 40% of artificial intelligence (AI) projects targeting agents could be canceled before the end of 2027. While the technology itself poses no major issues, the real challenge lies in how organizations implement it.
AI agents are driving a new wave of automation, enabling systems to perform tasks with minimal human intervention. However, many companies experience a mismatch between their expectations and the results obtained when moving from pilot phases to production.
To ensure success in deploying AI agents, it’s crucial to consider five fundamental lessons. The first is the need to align strategy across the entire organization. Typically, companies approach AI from two perspectives: executing mandates from top management or conducting isolated experiments by teams. Both tactics often fail on their own. Bottom-up initiatives can lead to promising pilots, but they rarely scale without support from senior management. Conversely, an exclusive top-down approach can also result in failures. A strategic combination of both approaches, with clear objectives and success metrics defined from the outset, is vital to bridging the gap between vision and operational reality.
The second lesson emphasizes the importance of preparing data from the very beginning. AI agents do not generate knowledge on their own; they depend on existing information, which is often fragmented and unstructured. With over 30% of critical knowledge stored in employees’ memories and 80% of relevant data being programmatically inaccessible, organizations may not be ready to effectively implement an AI agent.
Setting realistic expectations about agent performance is the third lesson. While organizations accept a human error rate of 5-10%, they demand perfection from AI systems. This mismatch can jeopardize initiatives with potential. Companies should adjust their expectations based on human performance rather than seeking an unattainable level of perfection.
The penultimate lesson focuses on finding a balance between building versus buying technology. Many organizations face significant obstacles when developing internal platforms without the necessary expertise. Completely outsourcing processes may result in the loss of competitive advantage. A hybrid approach may be more effective, starting with commercial solutions to quickly validate their value before identifying unique areas that require custom development.
Lastly, the importance of operational infrastructure should not be underestimated. Many pilots may perform well in controlled environments but fail when implemented in production due to infrastructure failures. It’s crucial to consider operational demands from the start, asking how to validate an agent’s performance, what will happen in case of failure, and who will have the ability to reverse decisions.
Successfully adopting AI agents is not just about having perfect technology or large budgets; it’s about organizational learning capability. Companies that start before everything is perfect, build incrementally, learn quickly from their mistakes, and scale what works will be the ones leading future transformation.
—
Source: MiMub in Spanish
