Divya van Mahajan

The Top 50 AI For Work Apps You Haven't Tried Yet

ai-daily-brief-podcast

AI for Work: Top 50 AI Application Spending Report (A16Z × Mercury)

Overview

This episode of the AI Daily Brief (hosted by Nathaniel Whittemore, though not explicitly named in this transcript) examines an Andreessen Horowitz (A16Z) report on where startups are actually spending money on AI applications. The report was generated in partnership with Mercury, a fintech bank serving over 200,000 startup customers, using transaction data from June–August 2025. The central thesis is that startup AI spending patterns reveal which tools are maturing from individual/consumer use into enterprise-grade workflows — and therefore signal where larger enterprises will likely invest next.

The episode also covers two headline items: reported technical difficulties with the OpenAI/Jony Ive ambient AI hardware device, and an update on Sora’s early adoption trajectory and monetization direction.

Source video URL: (not provided)

Prerequisites

  • Familiarity with the AI application landscape (OpenAI, Anthropic, Cursor, Notion, etc.)
  • Basic understanding of startup funding and go-to-market dynamics
  • Awareness of the distinction between AI infrastructure (compute, model training) and AI applications (products and workflows built on top of models)
  • General knowledge of enterprise software procurement vs. bottoms-up SaaS adoption
  • Familiarity with terms: SOC 2, SDR (Sales Development Representative), vibe coding, agentic AI

Main Points

Methodology and Scope of the A16Z/Mercury Report

  • A16Z partnered with Mercury (a startup-focused fintech bank with 200,000+ customers) to analyze spending data from June–August 2025
  • The report identifies the top 50 AI application providers — explicitly excluding infrastructure providers (compute, model development)
  • The dataset is heavily concentrated in the startup sector, making it a leading indicator of tools that may later become enterprise-standard
  • The report answers: What are early-stage startups buying and using in AI right now?

Why This Matters for Broader Audiences

  • Most large enterprises are limited to a narrow set of AI vendors: Microsoft, Amazon, Salesforce, Google, OpenAI, Anthropic
  • Startups are running voracious, broad experimentation — the tools rising to the top of startup spending are the most likely candidates to become enterprise-grade
  • Two relevant audience segments: (1) professionals thinking about AI adoption inside companies; (2) individual entrepreneurs/solopreneurs seeking competitive tools outside standard enterprise stacks

Bottoms-Up Adoption Is the Dominant Pattern

  • ~70% of the 50 companies follow a bottoms-up adoption model: usable by individuals first, then scaled to teams without requiring enterprise licenses to start
  • 12 of the 50 companies also appear on A16Z’s separate top 100 Gen AI consumer apps list (measuring web/mobile usage, not revenue)
  • 11 companies started as individual/consumer products and later moved into enterprise
  • Key implication: Forward-thinking enterprises should monitor which consumer tools their employees are already using as a signal for organizational opportunity

The 11 Categories of AI Work Applications

The 50 applications were grouped into the following categories:

  1. Foundation models and general assistance — OpenAI, Anthropic, Perplexity, Manus
  2. Enterprise knowledge and AI search — Glean, Notion
  3. Meeting notes and comms capture — Otter, HappyScribe, Fixer (largest by company count, no clear single winner yet)
  4. Creative and content — Canva, Midjourney, FreePick, PhotoRoom, Descript, Gamma (overlapping functionality; Descript emerging as video editing leader; Gamma as the singular AI presentation tool)
  5. Sales and go-to-market automation — Multiple tools; direct ROI measurability drives spending
  6. Customer support — Customer.io, Crisp
  7. Legal tech — Crosby
  8. HR and recruiting
  9. Dev tools and vibe coding — Replit (#3 overall), Cursor, Lovable, Emergent, Cognition (10% of the full list touches code)
  10. IT ops, finance, and compliance — Delve (SOC 2/HIPAA), Serval (IT helpdesk agent), Combinely (accounting)
  11. Real-time meeting coaching — Clueless (the so-called “cheating app,” ranked #26; sole entry in this category)

Horizontal Tools Currently Beat Niche/Vertical Tools

  • Spend mix is approximately 60% horizontal, 40% vertical
  • Horizontal = usable by anyone across a company; Vertical = targeted at specific roles
  • AI is blurring the horizontal/vertical line: creative tools and vibe coding tools, once considered niche/vertical, are now used company-wide and increasingly qualify as horizontal
  • Glean and Notion exemplify horizontal tools that succeed by sitting on top of existing data and workflows

Vibe Coding Is a Major Theme

  • 5 of 50 companies are explicitly agentic coding tools: Replit, Cursor, Lovable, Emergent, Cognition
  • 10% of the entire list touches code in some form
  • Replit ranked #3 overall (even ahead of Cursor), compared to #41 on the consumer list — indicating enterprise/startup contexts reward production-readiness and controls over pure ease of use
  • Vibe coding has expanded beyond developers to product managers, designers, and generalists across companies
  • Closer to production = more willingness to pay

Distribution and Integration Into Existing Tools Still Matter

  • Even among the most experimentally-minded startups, tools that integrate with existing systems show strong adoption:
    • Glean: permission-aware search on top of existing SaaS
    • Notion: AI layered into where company documents already live
    • Cursor: bring-your-own-model flexibility plus enterprise controls
    • Customer.io / Crisp: multi-channel integration
  • Flexibility in model choice and customization to existing systems is rewarded

AI Employees Are Moving from Hype to Budget Line Items

  • A small but significant set of tools represent genuine agentic AI workers rather than augmentation:
    • 11X — AI SDR/sales agent
    • Serval — IT helpdesk agent
    • Crosby — legal agent
    • Alma — immigration workflows
    • Cognition — background coding agents
  • These tools are compelling when they deliver against measurable outcomes (meetings booked, tickets resolved, etc.)
  • Compliance and IT are early agent targets because they are mandatory functions with low strategic differentiation — startups want them automated and off their mental plate

Channels With Clearer ROI Attract More Budget

  • Sales and marketing tools dominate partly because their outputs (pipeline, conversion) are directly measurable
  • Back-office automation (compliance, accounting) is also growing because the ROI case is clear even if less glamorous
  • Tools with ambiguous ROI face harder budget justification

Additional Notable Signals

  • China-origin tools are present: Kling and CapCut both appear on the list — indicating cost or capability advantages are overcoming geopolitical hesitancy for some startups
  • Hardware has a niche enterprise use case: Plaud (#38), a hardware note-taking device designed for in-person meetings, demonstrates that unobtrusive physical devices can find startup customers — a potential signal for the broader AI hardware category
  • Manus (#33) is the only generalist AI agent on the list; Genspark is notably absent — the generalist agent category remains very early

Startup Confidence and AI Adoption Correlation

  • Mercury’s broader startup survey found significant anxiety: 89% worried about economic uncertainty, 83% about tariffs, 80% about stock market volatility
  • However, AI-adopting companies report meaningfully higher confidence: 42% of significant AI users said their financial confidence had significantly improved (2024→2025) vs. 28% of non-AI adopters

Headlines: OpenAI/Jony Ive AI Device Encountering Development Challenges

  • The Financial Times reports engineers have been unable to solve critical issues that may delay release
  • Device described as a palm-sized, screenless ambient AI that responds to audio and visual environmental cues without wake words
  • Reported problem areas: core software, privacy, model personality, compute
  • Key UX challenge: device is “too eager” — chimes in too often or doesn’t know when to stop
  • Personality balance described as difficult: not too sycophantic, not too direct, helpful without looping
  • OpenAI insiders characterize these as normal hardware development challenges; hardware cannot be rolled back like software
  • OpenAI’s broader thesis is always-on ambient AI as the future interface paradigm

Headlines: Sora App Trajectory and Monetization

  • Sora reached #1 in the App Store; ~200,000 downloads per day at time of reporting
  • A16Z’s Olivia Moore identifies a four-stage social app adoption arc and places Sora in Stage 3: users determining whether a compelling “status game” exists to retain them
  • Key Sora updates announced:
    • Cameo controls: rights holders can now restrict types of content generated using their likeness
    • Watermark and moderation balance being refined
    • Deleting Sora app no longer forces deletion of ChatGPT account
    • Copyright: moving toward opt-in model with granular controls for rights holders and character licensing
  • Revenue model: Sam Altman indicated revenue sharing with rights holders for character generation; host interprets this as an inevitable move toward advertising, and criticizes the indirect communication around it
  • Copyrighted content (e.g., Pikachu mashups) drove early engagement; tighter guardrails risk reducing the “cool factor”

Key Concepts

  • Bottoms-up adoption: A go-to-market pattern where a product is adopted by individual users first, then spreads organically to teams and organizations without top-down enterprise procurement
  • Horizontal AI tool: An AI application usable by any function or role across a company (e.g., Notion, Glean, Perplexity)
  • Vertical AI tool: An AI application targeted at a specific business role or department (e.g., an AI SDR tool, a legal AI tool)
  • Vibe coding: The practice of building software using AI-assisted tools with minimal traditional programming, now extending beyond developers to non-technical roles
  • Agentic AI / AI employees: AI systems that operate autonomously to complete multi-step tasks and produce measurable outputs, rather than simply augmenting human actions
  • Ambient AI: An always-on AI paradigm that continuously monitors context and responds without explicit activation (no wake words)
  • Cameo (Sora feature): A Sora feature allowing users to grant permission for their likeness to be used in AI-generated videos, with opt-in controls
  • Bring-your-own-model (BYOM): A product architecture allowing enterprise customers to choose and swap the underlying AI model used by a tool
  • Glean: An enterprise AI search platform that layers over existing SaaS tools with permission-aware retrieval
  • Plaud: An AI hardware brand producing a physical note-taking device for in-person meetings
  • 11X: An AI-native SDR (Sales Development Representative) agent designed to automate outbound sales tasks
  • Cognition: An agentic coding company positioned further toward autonomous background coding agents than traditional AI coding assistants
  • Manus: A generalist AI agent; the only such agent represented in the top 50 spending list

Summary

The central argument of this episode is that A16Z and Mercury’s joint analysis of startup AI application spending offers a credible leading indicator of where enterprise AI adoption is heading. The top 50 applications cluster into eleven categories, with meeting notes/comms capture, creative/content, and dev tools/vibe coding being the largest. The dominant pattern is bottoms-up adoption — tools that start with individual users and grow into teams — and approximately 60% of spending flows to horizontal tools usable company-wide, though AI is progressively blurring the line between horizontal and vertical. Vibe coding is a breakout theme, with 10% of the list touching code and Replit ranking #3 overall due to its production-readiness for enterprise contexts. A small but significant cohort of agentic “AI employee” tools targeting sales, IT, legal, and compliance represents the category with the highest expected future growth, as startups show willingness to pay when outcomes are measurable. The episode closes with a data point that AI-adopting startups report substantially higher financial confidence than non-adopters, reinforcing the practical case for broader AI tool experimentation.