Gartner’s Core Warning
At its recent IT Symposium in Kochi, Gartner highlighted a growing and dangerous gap between enthusiasm for AI and the readiness to manage it. The firm warns that without strong governance, the rush toward autonomous AI systems is becoming a huge liability.
According to one Gartner survey, 95% of organisations have already faced at least one incident related to AI agents in the past year. Another projection is even more dire: by 2028, 40% of organisations scaling AI without proper controls will suffer a business-critical service disruption. This isn't just a technical problem; it's a looming business crisis rooted in a lack of planning and oversight.
First, What Are 'AI Agents'?
The term 'AI agent' can be confusing. It refers to more than just a simple chatbot. An AI agent is a type of AI that can operate autonomously to perform tasks, make decisions, and interact with other systems without direct human command for every step. Think of a system that not only answers a customer query but also processes a refund, updates inventory, and sends a confirmation email, all on its own. While incredibly powerful, this autonomy is also where the risk lies. Many businesses are deploying them rapidly; one survey found 71% of organisations expect to deploy agentic AI within the next 12 months. However, only 21% have the governance capabilities ready to manage them.
Unpacking the 18-Fold Cost Increase
The headline-grabbing figure from Gartner is that the cost to achieve positive returns from AI has risen 18-fold in just two years. This doesn't mean the AI software itself is 18 times more expensive. Instead, it reflects the total investment needed to turn an AI experiment into a successful, scalable, and safe business function. This includes the cost of technology, infrastructure, implementation, and, crucially, ongoing oversight and governance. As AI models become more complex and powerful, the resources required to manage their risks and ensure they deliver real value are climbing sharply, far outpacing the drop in cost for basic AI tasks.
The High Price of No Governance
Deploying AI agents without a strong governance framework is like giving the keys to your entire company to an intern with no supervision. The risks are enormous. Ungoverned agents can lead to 'runaway' processes that exhaust cloud computing budgets in hours or create cascading system failures. They pose significant security risks, as they can have broad access to sensitive company and customer data. A Gartner survey found that 57% of infrastructure and operations leaders admit their current AI governance is weak, and 55% say their existing IT risk tools are completely outdated. Without clear rules, accountability, and kill switches, companies are exposed to financial loss, compliance breaches, and severe reputational damage.
What Does Good Governance Look Like?
Effective AI governance isn't about stopping innovation; it's about enabling it safely. Gartner recommends a proactive approach built on several pillars. This starts with creating a full inventory of all AI agents and classifying them as 'nonhuman identities'. From there, it's about setting clear boundaries. This includes defining what data an agent can access, what actions it's allowed to take, and when it needs human approval. It also means having robust monitoring to observe what agents are actually doing and implementing 'circuit breakers' or kill switches to stop runaway errors before they cause major damage. Only 15% of companies currently have such automated backstops in place.
A Reality Check for Indian Businesses
For Indian companies eager to leverage AI for a competitive edge, Gartner's warning is a timely and critical piece of advice. The message is not to avoid AI, but to approach it with a clear strategy that balances innovation with risk management. The focus should shift from simply deploying AI tools to redesigning work and operating models to justify the investment. This means starting with a solid governance framework, even for small projects, and ensuring that accountability is clearly defined. Investing in oversight and controls from the beginning is no longer optional—it's essential to prevent AI initiatives from becoming costly failures.
















