Microsoft cuts AI agent sales targets as enterprises balk at unproven tech

Microsoft has slashed sales growth targets for its AI agent offerings after sales teams missed ambitious quotas, signaling weak enterprise demand. Customers appear unconvinced by the reliability and value of agentic tools, and many prefer ChatGPT over Copilot for general tasks. Despite record AI spending, much of Microsoft’s AI revenue still comes from AI firms renting cloud capacity rather than traditional enterprises deploying agents.

Key Points

• Microsoft cut AI agent sales growth quotas after widespread misses, with some Azure units reducing targets from 50% to ~25% or halving doubling goals to 50%.
• Despite hyping the “era of AI agents,” enterprise customers remain hesitant to pay premiums for agentic tools, and Copilot faces preference headwinds versus ChatGPT.
• Technical limits—confabulation, simulated reasoning errors, and brittleness outside training distributions—undercut reliability for autonomous, high‑stakes business tasks.
• AGI is portrayed as a potential remedy for agent weaknesses, but it remains nebulous and unrealized.
• Microsoft continues record AI capex ($34.9B in fiscal Q1) while much AI revenue stems from AI firms renting cloud, not mainstream enterprise adoption.

Sentiment

The discussion strongly agrees with the article's core thesis. HN commenters broadly validate that enterprise adoption has lagged because Microsoft's AI products aren't good enough and the underlying agent technology isn't ready for high-stakes autonomous work. Frustration with Copilot is pervasive and personal, extending beyond abstract skepticism to specific accounts of workflows being actively degraded. The minority view defending LLM value focuses on personal consumer use rather than enterprise agents, which most see as inadequate justification for current valuations and agentic promises.

In Agreement

• Microsoft Copilot's AI implementations are poorly executed, intrusive, and disruptive to existing workflows, with many users actively working around them or disabling them.
• AI agent technology is not ready for high-stakes enterprise automation due to reliability issues, hallucination risks, and brittleness—even modest error rates compound into unacceptable failure rates at production scale.
• The AI investment cycle is driven more by FOMO, herd behavior, and short-term financial incentives than by demonstrated product-market fit in the enterprise space.
• Enterprises' reluctance to pay premium prices for unproven agent tools is rational, not backward—the technology genuinely lacks the reliability needed for autonomous business-critical work.
• Microsoft's AI products feel like checkbox features designed for procurement tables rather than solving real user problems, explaining why employees prefer competitors like ChatGPT and Claude.

Opposed

• LLMs are genuinely useful to many people for everyday personal tasks, and dismissing them entirely is an overcorrection—the real question is whether current investment levels are justified, not whether the technology has any value.
• Microsoft's AI struggles reflect execution problems specific to their implementation rather than a fundamental flaw in the AI agent concept itself.
• The 'get them hooked then monetize' business model is familiar in tech and could still succeed once inference costs drop and capabilities improve, similar to how internet or mobile services evolved.
• AI failure in enterprise settings is partly attributable to companies deploying general LLM chatbots for tasks better suited to traditional ML approaches, not because agents are inherently flawed.