Divya van Mahajan

Half Of Employees Still Hiding Ai From Their Bosses And Its Their Bos

ai-daily-brief-podcast

Half of Employees Still Hiding AI From Their Bosses — And It’s Their Bosses’ Fault

Overview

This episode of the AI Daily Brief (published May 2, 2025) examines a persistent workplace phenomenon: a majority of employees continue to conceal their AI tool usage from managers and colleagues. The host argues that this is fundamentally a leadership failure rather than an employee behaviour problem. The episode also covers headline news including NVIDIA CEO Jensen Huang’s comments on China’s AI capabilities, Microsoft and Meta’s earnings reports, and Google’s reported deal to bring Gemini to iPhones.

Source video URL not available.

Prerequisites

  • Basic familiarity with the current AI landscape (large language models, enterprise AI tools, AI assistants)
  • General understanding of corporate/enterprise technology adoption
  • Awareness of the US–China technology competition context
  • Familiarity with terms such as “agentic AI,” “co-pilots,” and “BYO AI”

Main Points

NVIDIA CEO Jensen Huang: China Is Not Behind in AI

  • When directly asked how far behind China is in AI, Huang stated plainly: “China is not behind.”
  • He framed AI competition as an “infinite race” with no short-term finish line.
  • Huang noted that 50% of the world’s AI researchers are Chinese and that China has both significant technical capability and wealth.
  • Context: NVIDIA was reportedly caught off guard by US restrictions on H20 chip exports to China; the company is simultaneously announcing $500 billion in planned US AI infrastructure investment over five years.

Microsoft Earnings: Solid Growth With Capacity Warnings

  • Microsoft reported 13% overall revenue growth and 33% annualised growth for its Azure Cloud division.
  • CFO Amy Hood forecast continued strong cloud growth at ~35% for the following quarter.
  • Hood warned of capacity constraints expected by June 2025, attributing them to higher-than-anticipated demand.
  • Reports of Microsoft cancelling some data centre leases circulated, but the company characterised these as routine balancing decisions rather than a retreat from AI investment; Microsoft reaffirmed an $80 billion data centre investment commitment for the year.
  • Infrastructure lead times of two to seven years make constant demand-balancing necessary.

Meta Earnings: Doubling Down on AI Infrastructure

  • Meta reported steady revenue growth and a slight beat on analyst estimates.
  • The company raised its 2025 AI infrastructure spending guidance from $60–65 billion to $64–72 billion, citing increased hardware costs due to global supply chain pressures from tariffs.
  • CEO Mark Zuckerberg expressed confidence in the advertising business’s resilience to macroeconomic uncertainty.
  • Wall Street responded positively; stock rose ~6% in after-hours trading — a shift from previous quarters when elevated AI spending caused investor concern.
  • Zuckerberg highlighted the Meta AI standalone app as strategically important for US market leadership, noting WhatsApp as the primary global access point for Meta AI.

Google Gemini Coming to iPhone

  • NVIDIA CEO Sundar Pichai, testifying at the Google antitrust trial, confirmed that Google is close to a deal to integrate the Gemini assistant into Apple’s iPhone ecosystem.
  • Pichai expects to close the deal with Apple by mid-2025 and complete integration before year-end.
  • The integration would likely allow Siri to defer complex queries to Gemini, similar to the existing ChatGPT integration.
  • Apple SVP Craig Federighi had previously hinted at the possibility of third-party model choice within Apple Intelligence.

The “Secret Cyborg” Trend: New Data Confirms Employees Still Hiding AI Use

  • A 2024 Microsoft Work Trends Index had found that 78% of knowledge workers using AI were not disclosing it to colleagues or bosses — labelled “BYO AI.”
  • A 2023 Slack study found 48% of workers were uncomfortable telling managers about AI use, fearing they would appear to be “cheating” or seem lazy/incompetent.
  • The new KPMG/University of Melbourne study (Trust Attitudes in the Use of Artificial Intelligence) surveyed 48,000 workers across 47 countries (November 2024–January 2025) and found:
    • 57% of workers still hide their AI usage at work
    • 50% present AI-generated content as their own work
    • 67% intentionally use AI at work
    • 69% say their organisation uses AI
    • 70% are using free, publicly available (non-enterprise) tools

Reported Benefits and Risks of Workplace AI Adoption

  • Benefits reported:
    • 54% report increased efficiency, quality of work, and innovation
    • 43% report increased revenue-generating activity
  • Risks and problems reported:
    • 28% report increased workload, stress, and pressure
    • Nearly half admit to using AI in ways that violate organisational policies
    • Many are uploading sensitive company data (financial, sales, customer) to public AI tools
    • Two in three rely on AI output without evaluating it
    • Over half report making mistakes in their work due to AI

The Root Cause Is Leadership Failure, Not Employee Behaviour

  • Only 60% of organisations have an AI strategy in place
  • Only 54% have responded with relevant policy
  • Only 42% of employees feel they have the skills and knowledge to use AI appropriately
  • Only 28% have received any formal or informal AI training
  • Only 52% feel they can use AI tools effectively
  • The host argues that the disparity between consumer-grade AI tools (which employees use at home) and enterprise-provisioned tools is large and growing — pushing employees toward unsanctioned tools.
  • The host places explicit responsibility on leadership for: failing to provide appropriate tools, policies, guidelines, training, guardrails, and a clear organisational vision for AI.

Upskilling Paradigm Is Already Outdated

  • Current AI upskilling efforts are still framed around the model of 9–12 months ago: helping individual employees be more productive using AI co-pilots.
  • The host contends that the relevant paradigm has already shifted to agentic AI: building, provisioning, and managing teams of AI agents; redesigning work at a structural level; and unlocking capabilities that were previously impossible.
  • No organisations are yet providing training aligned with this new agentic paradigm, which the host predicts will create significant problems in the near term.

Key Concepts

  • Secret Cyborg / BYO AI: The practice of employees using AI tools at work without disclosing this to managers or colleagues, often with personally sourced or consumer tools.
  • Agentic AI: AI systems that can autonomously plan, decide, and execute multi-step tasks, often as part of coordinated “agent teams,” representing a structural shift beyond simple productivity assistance.
  • Enterprise-grade AI tools: AI tools provisioned and managed by organisations with privacy, security, and compliance guarantees, as opposed to free consumer-facing tools.
  • AI upskilling: Organisational training programmes designed to build employee competency in using AI tools appropriately and effectively.
  • H20 chips: NVIDIA graphics processing units subject to US export restrictions to China, relevant to the US–China AI competition discussion.
  • AI CapEx: Capital expenditure by technology companies on AI infrastructure, including data centres and hardware.
  • Trust gap (AI context): The deficit of trust — both between employees and AI outputs, and between employees and organisational leadership — that impedes safe and transparent AI adoption at work.

Summary

Drawing on a large-scale KPMG/University of Melbourne survey of 48,000 workers across 47 countries, the host argues that the widespread practice of employees concealing AI tool use from their employers is not a behavioural or ethical failure on the part of workers, but a direct consequence of organisational leadership failures: the absence of clear AI strategies and policies, inadequate training, inferior enterprise tooling compared to freely available consumer alternatives, and a failure to articulate a coherent organisational vision for AI. With 57% of workers still hiding AI use and only 28% having received any AI training, the host contends that enterprises must move faster and must reframe their upskilling efforts away from individual productivity co-pilots toward the emerging agentic paradigm — one in which the critical skills involve designing, managing, and delegating to teams of AI agents. The episode frames this moment as consequential: how organisations respond now will substantially determine outcomes for workers and businesses over the next several years.