Divya van Mahajan

AI Agent Deployments Quadruple in 2025

ai-daily-brief-podcast

Overview

This episode of the AI Daily Brief (published September 19, 2025) covers two primary topics: (1) a commentary on Microsoft’s strategic anxieties in the AI era, and (2) a detailed analysis of the KPMG AI Quarterly Pulse Survey (Q3 2025), which tracks enterprise AI and agent adoption among large organizations. The host is the producer/presenter of the AI Daily Brief podcast and video channel. No guest speakers are named. The central thesis is that AI agent deployments are rapidly moving from experimentation into full production across large enterprises, and that this transition is reshaping workforce dynamics, investment priorities, and how organizations measure ROI.

Source video URL: Not provided.

Prerequisites

  • Basic familiarity with generative AI concepts (large language models, AI assistants, copilots)
  • Understanding of enterprise software adoption cycles (piloting, deployment, production)
  • Awareness of major AI products: Microsoft Copilot, GitHub Copilot, Google Gemini, ChatGPT, Claude
  • Familiarity with the concept of AI agents (autonomous systems that take actions, distinct from AI assistants that respond to prompts)
  • General knowledge of corporate technology adoption patterns and ROI frameworks
  • Awareness of the competitive landscape among Microsoft, Google, and OpenAI

Main Points

1. Microsoft CEO Satya Nadella’s Existential Concern

  • At a company-wide town hall, Nadella expressed fear that Microsoft could follow the fate of once-dominant tech companies (e.g., DEC, Xerox, IBM) that failed to navigate platform shifts.
  • Nadella specifically cited DEC (Digital Equipment Corporation), a minicomputer giant of the early 1970s that collapsed after failed bets on emerging architecture.
  • He acknowledged that some of Microsoft’s most valuable product categories, built over 40 years, may no longer be relevant in the AI era.
  • An employee raised concerns that Microsoft’s culture had become “colder, more rigid, and lacking in empathy,” consistent with reports of all-time-low employee morale following multiple waves of layoffs.
  • Nadella’s framing: companies must “win the new rather than protect the past.”

2. Host’s Assessment of Microsoft’s AI Trajectory

  • Microsoft entered the generative AI era with a strong position via its unique OpenAI partnership, which positioned its products as early capability leaders.
  • However, the host characterizes Microsoft’s subsequent execution as “stumble after stumble”:
    • Copilot products have not kept pace with consumer alternatives (e.g., Gmail-accessible AI), fueling widespread shadow AI use inside enterprises.
    • The response to the Sam Altman firing/rehiring saga led to strategic hedging that diluted enterprise focus.
    • Hiring of an AI leader focused on consumer rather than enterprise markets is seen as a misallocation of Microsoft’s strongest asset: enterprise distribution and lock-in.
  • Positive signals noted: willingness to integrate Claude into GitHub Copilot based on perceived quality; AI agents being added throughout Microsoft Teams (note-taking, scheduling, channel assistance); a $3.3 billion data center in Wisconsin nearing completion, plus a planned $4 billion adjacent facility housing hundreds of thousands of NVIDIA Blackwell GPUs, intended as a frontier model training cluster.
  • Contrast drawn with Google, whose AI prospects the host views as having improved significantly over the past two years, while Microsoft’s have arguably worsened.

3. Google and Notion: Broad “Agentification” Trend

  • Google is embedding Gemini natively into Chrome as a default feature, with AI mode accessible from the Chrome search bar.
    • Near-term agentic features: automated flight booking, grocery shopping.
    • Planned cross-platform features: syncing with Google Calendar and Workspace.
    • Chrome’s distribution scale means Gemini will serve as many users’ first meaningful AI interaction.
  • Notion 3.0 replatforms entirely around agents:
    • Agents can create pages and databases, update data, and complete multi-step tasks autonomously.
    • Capable of completing up to 20 minutes of work across hundreds of pages simultaneously.
    • Integrates with external data sources: Slack, email, Google Drive.
    • Early user reports are strongly positive.
  • The host frames this as “the agentification of everything”—a normalizing macro-trend likely to define 2026.

4. KPMG Q3 2025 Pulse Survey — Longitudinal Context

  • The KPMG AI Quarterly Pulse Survey is now in its fourth edition (Q4 2024 through Q3 2025), covering 130 organizations with $1B+ in revenue.
  • Q4 2024 baseline: Most organizations were in the exploration phase; 37% piloting agents; 12% claiming deployment. Knowledge worker daily AI tool usage at 22%. Anticipated annual Gen AI spend: $89M.
  • Q1 2025: Piloting nearly doubled (37% → 65%); 99% of organizations intended to deploy agents; daily AI tool usage surged from 22% to 58%; anticipated spend rose to $114M. Strong preference (2:1) for buying pre-built agents over building custom ones.
  • Q2 2025: Agent deployments tripled (11% → 33%); 90% of organizations past experimentation; nearly half (46%) equally focused on efficiency and revenue growth (no respondents focused only on efficiency). Anticipated spend: $114M.

5. KPMG Q3 2025 — Three Key Findings

Finding 1: Agent deployments nearly quadrupled

  • Deployment rate rose from 11% (Q1) to 33% (Q2) to 42% (Q3)—nearly a 4× increase since the start of 2025.
  • These represent agents fully through the pilot phase and in active organizational operations.
  • New dominant challenge: complexity of agentic systems, cited by 71% of organizations (up from 39%), as organizations grapple with deploying agents at scale.

Finding 2: Workforce resistance to agents is collapsing

  • Employee resistance to AI agents dropped from 47% in Q2 to 21% in Q3—a massive shift in a single quarter.
  • Possible explanations offered (speculative): more concerted organizational upskilling efforts; employees experiencing agents as tools that remove disliked tasks rather than eliminate jobs; narrowing of the leadership-employee AI understanding gap.
  • Notable background context: adoption has historically been top-down (57% of C-suite using AI vs. only 15% of entry-level workers in formal/approved tools), partly obscured by widespread shadow AI use.

Finding 3: Traditional ROI metrics are seen as inadequate

  • 78% of leaders say traditional business metrics fail to capture AI’s full impact.
  • Simultaneously, 78% face significant board/investor pressure to demonstrate AI value—creating tension between needing quick wins and recognizing that AI’s benefits transcend standard ROI calculations.
  • 57% expect measurable ROI within 12 months.
  • Leaders are already tracking: improved productivity (97%), enhanced profitability (94%), higher quality work outputs (91%).
  • The host notes that measurement frameworks must differ by AI type (assistants vs. agents) and by department (e.g., coding agents vs. customer service agents).
  • Critically: absence of proven ROI is not causing organizations to pull back from Gen AI investment.
  • Organizations are beginning agent-specific training programs:
    • 50.7% implementing AI agent shadowing programs (employees observe experts working with agents)
    • 52% creating agent-specific sandbox environments for practice
    • 40% running workshops on human-agent collaboration mindset
    • Prompt engineering training has grown from 69% to 85% of organizations
  • The host critiques the broader market for remaining stuck in a “prompt engineering paradigm” rather than offering robust agent-specific upskilling tools.
  • Anticipated annual Gen AI investment has risen again: $114M → $130M per organization.

Key Concepts

  • AI Agent: An autonomous AI system that takes actions and completes multi-step tasks independently, distinct from an AI assistant that responds to individual prompts.
  • Agentification: The broad trend of embedding autonomous AI agents into existing software platforms and enterprise workflows.
  • Shadow AI: The use of AI tools by employees outside officially sanctioned organizational channels, often due to consumer tools outperforming enterprise-approved alternatives.
  • Exploration / Piloting / Deployment phases: A maturity spectrum for organizational AI adoption—exploration (learning), piloting (structured testing), deployment (production use at scale).
  • Efficiency AI vs. Opportunity AI: A framing distinguishing AI used to do existing tasks faster/cheaper (efficiency) from AI used to do things not previously possible (opportunity/revenue growth).
  • KPMG AI Quarterly Pulse Survey: A recurring survey of 130+ large enterprises ($1B+ revenue) tracking AI adoption behaviors, investment, and sentiment quarter-over-quarter.
  • Leader-led (top-down) adoption: The observed pattern in which senior executives adopt AI tools at significantly higher rates than entry-level or middle-management employees.
  • NVIDIA Blackwell: NVIDIA’s latest generation of high-performance GPUs used for AI training at scale.
  • Frontier model: A cutting-edge, large-scale AI model at the leading edge of capability, requiring massive compute to train.
  • Notion 3.0: Notion’s redesigned platform built around AI agents capable of autonomously managing knowledge-base tasks.

Summary

The central message of this episode is that enterprise AI adoption—particularly the deployment of autonomous AI agents—has undergone a dramatic and measurable acceleration in 2025, with deployments nearly quadrupling across large organizations since the start of the year. Drawing on the KPMG Q3 2025 Pulse Survey, the host documents a clear progression from exploration and piloting toward full production deployment, accompanied by a significant collapse in employee resistance to agents, rising investment commitments, and a growing recognition among leaders that traditional ROI frameworks are insufficient to capture AI’s transformative impact. This macro-trend of “agentification” is visible not only in enterprise survey data but in product announcements from Microsoft (Teams agents, massive data center investment), Google (Gemini embedded in Chrome), and Notion (agent-native platform redesign). Framing these developments against Microsoft CEO Satya Nadella’s public anxiety about corporate survival in the face of platform shifts, the host argues that the organizations and companies most likely to thrive are those willing to build toward what is emerging rather than defend what they have already built.