Gartner: Current AI is not ready for the enterprise environment

The mirage of maturity in generative AI

During the recent Gartner IT Symposium, the consulting firm issued a warning that resonates strongly in boardrooms around the world: artificial intelligence, as we know it today, is not ready for the enterprise. Analysts Daryl Plummer and Kristin Moyer have dismantled the unbridled optimism surrounding major AI providers, pointing to a fundamental disconnect between the priorities of AI labs and the needs of corporate resilience. This diagnosis is not minor; it represents a shift in discourse in an industry that, until recently, treated the adoption of Large Language Models (LLMs) as a long-distance race where speed was the only indicator of success.

Historically, technological adoption in companies has followed a predictable maturity cycle: proof of concept, pilot deployment, and scaling. However, generative AI has broken this cycle. Unlike the transition to cloud computing of the last decade, where companies could manage infrastructure with clear SLAs (Service Level Agreements), current AI behaves like an unstable black box. Gartner argues that companies are trying to build skyscrapers on foundations of sand, entrusting their critical processes to providers that have not yet defined their legal or technical responsibility for systemic failures.

The problem of instability: Short-lived models

One of the most critical points is the lack of consistency. According to Plummer, current providers do not understand the terms of liability or the continuity standards required by professional-grade enterprise software. The practice of constantly updating models, often referred to as performance 'drift,' without considering how this might break existing applications, is a symptom of technical immaturity. With an average lifespan of barely six months before a model is replaced or deprecated, companies integrating these technologies face an unsustainable maintenance cycle reminiscent of the worst years of 'spaghetti code' in early software development.

This volatility forces organizations into constant updates, what Plummer calls a 'runaway pace of innovation.' Unlike a traditional ERP, where updates are planned and backward-compatible, the AI ecosystem operates under the logic of 'move fast and break things,' a philosophy incompatible with financial, healthcare, or legal environments where determinism is a non-negotiable requirement.

The threat of 'sloppy AI'

Kristin Moyer highlights a worrying phenomenon: 86% of CIOs perceive that the risks associated with AI are growing faster than the value it provides. This imbalance is fueled by two critical factors:

• Sloppy consumption: Employees using AI tools without criteria, increasing the attack surface and unnecessary spending. Moyer compares this behavior to the uncontrolled adoption of expensive cloud instances in previous years, where developers chose the most powerful option without considering the return on investment (ROI).
• Hidden cost (AI slop): The proliferation of low-quality AI-generated content requires hours of human labor to be corrected or filtered. According to Gartner data, 40% of workers have encountered this 'slop,' and correcting it consumes, on average, two hours per incident. In an organization of 1,000 employees, this translates into an opportunity cost of 9 million dollars annually, a figure that rarely appears in innovation budgets.

Towards centralized governance: The 'AI central bank'

Beyond the criticism, Gartner proposes a paradigm shift. The recommendation is to establish an "AI central bank" within organizations. This structure would not only serve to oversee technological deployment but to ensure accountability, something that is currently almost impossible due to the fragmentation of tools and the invisible integration of AI into existing SaaS products. AI is no longer an external application; it is an invisible component within the tools that companies already use, making it extremely difficult for a CIO to quantify how many autonomous agents are operating within their corporate network.

The historical lesson is clear: no technology has survived in the enterprise without a solid layer of governance and long-term support. Gartner suggests that companies must regain control by revaluing alternative and proven automation technologies, which, although less 'shiny' than the latest generation of LLMs, offer the stability necessary for critical systems. Business innovation cannot be an act of faith; it must be a calculated bet where resilience is, at the very least, as important as processing power.

In conclusion, the final message is a call for sobriety: companies must stop viewing AI as a magic solution and start treating it with the same rigor they apply to their critical ERP or CRM systems. The era of free experimentation is over; the era of corporate responsibility has just begun.