Martyn Rhisiart Jones

Oza-Cesuras 18th December 2025

The Rise of the Digital Minions: Why Agentic AI is the Tech World’s Latest Obsession

In the hushed corridors of Davos-style summits, one phrase has become inescapable. Agentic AI also dominates discussions in the feverish demo halls of San Francisco conferences.

These are not mere chatbots that parrot responses. Instead, they are autonomous “agents.” These are goal-oriented software entities that plan and reason. They execute multi-step tasks and even collaborate like tireless digital colleagues. Gartner, ever the oracle of enterprise trends, crowned it the top strategic technology for 2025. They predict that by 2026, 40% of enterprise applications will embed task-specific agents. This is a significant increase from under 5% today. Yet, in the same breath, the analysts warn of a different future. Over 40% of these projects will fail by 2027. They will become victims of soaring costs and murky ROI. The perennial sin of “agent washing” will contribute to rebranding yesterday’s chatbots as tomorrow’s revolutionaries.

This is classic Silicon Valley alchemy. Take generative AI’s lingering glow and infuse it with promises of true autonomy. Watch venture dollars flow. A Forbes contributor quipped that it’s the hype cycle on steroids. “Agentic” conjures visions of sci-fi independence. However, most implementations remain stubbornly tethered to human oversight.

From Hype to Half-Truths: Agentic AI isn’t vaporware.

At its core, it builds on large language models (LLMs). It includes added layers like planning tools, memory, and access to external APIs. Imagine an agent that doesn’t just suggest a travel itinerary. It books flights and reserves hotels. It also adjusts for delays – all without constant prodding. Real-world glimpses are emerging. Salesforce’s Agentforce orchestrates sales workflows. Microsoft’s Copilot agents handle enterprise drudgery. Startups like Aisera and Ema promise “systems of agents” for IT, HR, and finance. In customer service, Gartner foresees agents resolving 80% of routine queries by 2029, slashing costs by 30%.

McKinsey reports early adopters in tech and healthcare are scaling pilots. Additionally, 23% of surveyed firms are already deploying agents in functions like IT service desks. But peel back the gloss, and cracks appear. Critics decry “agent washing”: vendors slapping the label on glorified automation. Reliability falters – hallucinations compound in multi-step chains, security vulnerabilities multiply, and governance lags. As one Gartner poll revealed, while hype fuels investment (61% of firms dipping in), many remain in “wait-and-see” mode. Forbes has called it “insane” overpromising, normalising the irrational habit of misleading grandeur.

The Valley’s New Gold Rush – and Its Shadow. Wired would thrill at the prototypes. There are multi-agent swarms negotiating media buys. Physical AI is blending agents with robotics for warehouse autonomy. Yet the FT lens reveals the sober economics. The supply of agentic platforms (from AWS Bedrock to Google Gemini) far outstrips demand. This hints at impending consolidation. Microsoft reportedly struggles to sell enterprise clients on complete agentic suites; stock dips follow tepid uptake.

For executives, the allure is productivity nirvana – agents as a “goal-driven digital workforce” that never tires. Risks? Debates about job displacement are simmering. Forrester predicts 25% cuts in data teams. Looming on the horizon are ethical quagmires, such as bias in autonomous decisions. Fraudsters are eyeing agents for sophisticated scams.

In 2026, expect maturation: tighter guardrails, hybrid human-AI oversight, and wins in narrow domains like compliance or supply chains. Agentic AI won’t conquer the world overnight, but it’s no fleeting fad. As one observer noted, it’s less a revolution than evolution – from passive tools to proactive partners. The question for leaders: invest wisely, or get left chasing shadows in the trough of disillusionment.

The agents are coming. Whether they deliver utopia or just upgraded spreadsheets remains the multibillion-dollar question.

Thanks for engaging.