Divya van Mahajan

Why AI Leads to More Work, Not Less

ai-daily-brief-podcast

Why AI Leads to More Work, Not Less

Overview

This episode of the AI Daily Brief (dated 2026-02-10) discusses a study published in the Harvard Business Review by Berkeley Haas professors Aruna Ranganathan and Ching-Chi Maggie Yee, examining the real-world impact of AI adoption on worker behavior and workload. The central thesis is that contrary to popular expectations, AI tools — particularly generative and agentic AI — are not reducing work; they are expanding and intensifying it. The episode also covers related AI news including ByteDance’s Seed Dance 2.0 video model, OpenAI’s advertising rollout, Databricks’ revenue growth, and Anthropic’s agentic coding trends report.

Source video: No URL provided.

Prerequisites

  • Basic familiarity with generative AI tools (e.g., ChatGPT, Claude, Codex)
  • Understanding of agentic AI concepts (AI agents that operate autonomously or semi-autonomously on tasks)
  • General knowledge of the SaaS (Software as a Service) industry and enterprise software
  • Awareness of “vibe coding” — the practice of directing AI to write code through natural language prompts with minimal traditional programming
  • Familiarity with current AI model landscape (OpenAI GPT/Codex series, Anthropic Claude series)

Main Points

ByteDance Releases Seed Dance 2.0 Video Model

  • ByteDance launched Seed Dance 2.0, a new AI video generation model, without major announcement; early demos generated significant attention online.
  • Key differentiator: native audiovisual co-generation — audio is generated alongside video rather than added in post-production, resulting in accurate lip sync and environmentally contextual sound.
  • Supports multimodal input, 2K resolution, multiple cuts within a single 15-second clip, and a range of visual styles (cinematic, cartoon, product demos).
  • Reviewers noted strong character consistency and physics simulation; dialogue quality was the only noted weakness.
  • Includes a consumer-facing interface, unlike prior Chinese video models that were API-only.
  • Observers suggest this may accelerate the release timelines for competing models (e.g., Google Veo, OpenAI Sora).

White House Data Center Pact

  • The Trump administration is seeking voluntary commitments from major tech firms on community protections related to AI data center construction.
  • Draft agreement focuses on preventing data centers from raising household electricity prices, straining water supplies, or destabilizing the grid.
  • Tech companies would pledge to bear the full cost of infrastructure upgrades required by their data centers.
  • A formal White House announcement event is planned; administration officials noted the circulated draft was outdated.

SaaS Under Pressure: Monday.com vs. Databricks

  • Monday.com stock fell 21% after issuing weak guidance; 2026 revenue forecast was cut by one-third from prior investor day projections, and 2027 guidance was withdrawn entirely. Stock is down over 45% for the year.
  • A CNBC reporter demonstrated that Claude (Anthropic’s AI) could replicate a functional Monday.com-style platform in under an hour using vibe coding — illustrating disruption risk.
  • Databricks, by contrast, reported a $5.4 billion annual revenue run rate (up 65% year-over-year) and raised $7 billion in new capital.
    • 25% of Databricks’ ARR is now attributed to AI products.
    • 80% of databases on the Databricks platform are reportedly being built by AI agents.
  • Databricks CEO framed the divergence: companies that make the AI-first transition will grow; those clinging to legacy UX and SaaS models face compression.
  • The speaker interprets this as evidence that markets are penalizing companies that fail to show how they will compete long-term with AI, not just cut costs.

OpenAI: Advertising Rollout and New Model Rumours

  • OpenAI began rolling out advertising to free and lower-tier ($8 “Go”) ChatGPT subscribers; paid subscribers (Plus, Pro, Business, Enterprise, Education) are not shown ads.
  • Ads appear in a clearly labeled section in the lower third of the screen, not embedded in chat sessions.
  • Default settings allow targeting based on current/past chat content and stored memory; users can opt out, dismiss ads, or trade ad-free access for reduced usage limits.
  • Separately, reports citing internal Slack messages from Sam Altman indicated ChatGPT was back to exceeding 10% monthly growth and that an updated chat model (likely GPT-5.3 for general availability) was imminent.
  • Codex (previously released) saw 1 million downloads in its first week and 60% week-over-week growth in users.

The Berkeley Haas Study: AI Intensifies Work

  • Researchers embedded with a 200+ person technology company from April to December 2025; AI use was not mandated but employees were positively predisposed to it.

  • The study found three primary forms of work intensification among AI power users:

    1. Task Expansion — Workers took on responsibilities previously belonging to others. Product managers began writing code; researchers took on engineering tasks. AI made previously inaccessible tasks feel achievable, producing an “empowering cognitive boost” and reducing dependence on colleagues. These experiments accumulated into a meaningful widening of job scope.

    2. Blurred Work/Non-Work Boundaries — Because starting tasks with AI was so easy, workers began slipping small amounts of work into what had previously been breaks — prompting during lunch, meetings, or while waiting for files to load. “Just-in-time” prompting (sending a prompt before stepping away so AI could work in the background) became habitual.

    3. Increased Multitasking — Workers ran multiple AI agents in parallel, manually worked on one task while AI generated an alternative, or revived long-deferred projects because AI could handle them in the background.

  • Spillover effects on colleagues: Engineers spent more time reviewing, correcting, and coaching AI-assisted work produced by non-engineers, including finishing incomplete pull requests.

  • Boiling frog effect on recovery: Workers often only realized in hindsight that their downtime had been eroded and recovery was compromised. Expectations for speed increased organically — not through explicit demands — but through what became visible and normalized.

  • Organizational responses suggested by researchers: Intentional pauses, task sequencing, and “human grounding” as new management strategies.

The Speaker’s Interpretation: Expansionary vs. Efficiency Framing

  • The speaker argues the study is net positive because it demonstrates that the total amount of work to be done is not fixed — it expands to accommodate new capacity.
  • Winning organizations will not focus on “doing the same with less” but on “doing more with the same, or way more with a little more” — entering new markets, product lines, and revenue streams.
  • AI should be understood as an expansionary, opportunity-creating technology, not merely an efficiency technology.
  • In the short term, some organizations will use AI to cut costs and be rewarded by markets; long-term winners will be those who use AI for growth.
  • Non-technical use cases are expanding across organizations: coding capabilities are democratizing beyond engineering departments, with domain experts implementing solutions directly.
  • The speaker’s prediction: “vibe coders” will be hired specifically to serve non-engineering teams within organizations.
  • Multi-agent systems are replacing single-agent workflows (Anthropic Trend #2); the emergence of tools like OpenClaw is an early signal of coordinated agent teams becoming standard.
  • The shift from assisted AI to agentic AI is accelerating the “always-on” pressure already identified in the Haas study.
  • Referenced: a Lenny’s Podcast episode featuring Lazar Jovanovic, a full-time vibe coder at Lovable, as a preview of how this role might proliferate inside organizations.

Key Concepts

  • Task Expansion: The phenomenon where AI enables workers to absorb responsibilities previously held by others, broadening individual job scope.
  • Blurred Work/Non-Work Boundaries: The erosion of clear separation between work time and personal time as AI makes initiating tasks frictionless.
  • Agentic AI / AI Agents: AI systems capable of autonomously executing multi-step tasks, often running in the background or in parallel, with minimal moment-to-moment human direction.
  • Vibe Coding: The practice of building software by directing AI with natural language prompts rather than writing code directly; requires minimal traditional programming skill.
  • SaaSpocalypse: Informal term used to describe the current market pressure on traditional SaaS companies whose business models are perceived as vulnerable to AI disruption.
  • Multi-agent systems: Architectures in which multiple AI agents operate in a coordinated fashion, each handling different subtasks simultaneously.
  • Native audiovisual co-generation: Generating audio and video simultaneously within a single model pass, as opposed to adding audio to video in post-production.
  • Expansionary opportunity-creating technology: The speaker’s framing of AI as a tool for opening new markets and capabilities, contrasted with viewing AI purely as a cost-cutting efficiency tool.
  • Boiling frog effect (in AI context): The gradual, unnoticed erosion of worker recovery time and the invisible raising of performance expectations as AI-enabled work becomes ambient.

Summary

Drawing on an embedded longitudinal study of a technology company conducted by Berkeley Haas professors and published in the Harvard Business Review, this episode argues that AI is not reducing work but is instead intensifying it along three dimensions: workers expand into new task domains, they allow work to bleed into previously protected non-work time, and they take on more simultaneous workstreams via parallel agent use. The speaker frames this finding as broadly positive evidence that the aggregate volume of work is not fixed — it expands to meet new human and AI capacity — which he believes undermines the most pessimistic long-term job displacement scenarios, even as short-term displacement risks remain real. At the same time, the study surfaces genuine organizational challenges: spillover burdens on specialist colleagues, eroded recovery time, and invisibly rising performance expectations. The episode situates these findings within the broader 2026 AI landscape — ByteDance’s advance in video generation, Databricks’ AI-driven revenue growth versus Monday.com’s struggles, and Anthropic’s prediction that multi-agent systems and democratized coding will accelerate these trends across entire organizations — concluding that the central management challenge is no longer speculative but immediate and observable.