With their latest release of ChatGPT Atlas, OpenAI has brought the AI wars to the Mag7's doorstep.
By introducing Jobs, Certifications, Project Mercury, Fundamental Research, Sora, Codex, Agent Mode, Stripe & Shopify Integrations, AgentKit, and now Atlas, OpenAI is vying to become the one stop shop for everyone across business, technical, scientific, economic and consumer use cases.
Many of their offerings now press directly against Microsoft (LinkedIn, GitHub), Google (Search, Gemini, YouTube/Veo, Chrome), Meta (Instagram), and workflow stacks like Zapier/n8n...plus a long tail of private companies.
This week we dive deeper into the effect of OpenAI's dual-track Schumpeterian creative destruction of legacy software companies and its potential to capture a majority of the economic growth drivers in this AI supercycle.
Right after Fidji Simo joined forces with Vijaye Raji last month to become the second most important pairing in OpenAI after Sam Altman & Greg Brockman, we are seeing in real time this Applications team's vision build out at an unprecedented scale.
Before diving into OpenAI becoming the vanguard of this AI technological wave, let's take a look back at the four previous major technological waves that disrupted how the modern economy has shaped up from 1995 to 2025.
The web put economic demand online. With nearly 5.5B people now connected to the internet, the addressable market for any disruptive technology is simply bigger than it has ever been before.
When mobile came around, it made connection portable, thus allowing for instantaneous communication, unlocking a further step change in economic output. Today, 5G itself has nearly 2.5B subscribers worldwide, paving the way for a 6G with rich media & agents.
Then with the public cloud and the advent of today's AWS, GCP, Azure, capex was eliminated from most companies, enabling a new level of applications that was previously too expensive to operate. Public cloud spend is estimated to top $720B by end-2025 and growing, per Gartner.
Until this AI wave, SaaS had already greatly compressed and digitally transformed entire industries, with ads and subscriptions being the major revenue driver for B2B & B2C businesses. Global ad spend topped $1T in 2024 with the number slated to grow at nearly 10%/yr.
With ChatGPT Atlas, OpenAI's web browser with ChatGPT built in, and other products, they have captured today's equivalent of Google's search bar - the home page of the internet.
In doing so, they have gone upstream of search, and compressed the "answer to action" pathway. Where before we used to think, search, compare, and execute manually, ChatGPT essentially predicts what we would like and spits out the outcome in an easy to digest format.
With Instant Checkout, released for Etsy, Shopify, and other e-commerce stores, powered by Stripe, it goes a step further by closing the intent -> content -> purchase loop almost immediately.
On Ads and tailored content, Sora 2 lowers the time/cost to usable videos, further accelerating this flywheel of seeing something to purchasing the item we see.
On talent & business, OpenAI wants to modify traditional roles around AI-native skillsets that people would learn with Certifications, apply for and work using Jobs, feeding into the flywheel of accelerating outcomes for businesses and end users.
All in all, where previous technological waves prioritised usage and visibility to monetise, in the age of AI we are gradually moving to an outcome-based economy where the velocity of outcome generation, context input, and processing power will determine the winners.
Anthropic's Economic Index showed that usage currently is concentrated in specific geographies and industries, where context and outcomes have been the easiest to model, in addition to their proximity to established tech hubs.
OpenAI's GDPval evaluated real work being done across 9 sectors and 44 occupations and argues that in some specific cases, like in Government, Retail, and Wholesale, GPT5 is rapidly approaching parity in terms of outcomes generated.
(source: OpenAI GDPval study, arxiv)
Like in previous technological cycles, the adoption of AI is fast, uneven, and clustered around what is most legible to large language models and where there exists a large amount of explicit, repeatable information that can be learned by these systems.
This uneven adoption underlines the dichotomy that exists between the hyperscalers and the AI Infrastructure players that Coatue recently mentioned in their Public Markets Update: where AI Infra stocks have soared 146% since ChatGPT's launch, 95% of orgs haven't reported any return on their AI investments yet.
Nevertheless, the bull case says that $150B in AI revenues is growing rapidly, fuelled by the hyperscalers' operating cashflow rather than speculative leverage - even though these companies have bet the house on AI bringing better margins long-term.
Rather than waiting for broad market adoption, OpenAI is now betting it can accelerate specific verticals where incumbents, though well-entrenched, may be vulnerable to rapid disruption.
Fidji Simo built her reputation monetising Facebook's mobile business after its IPO and shepherding Instacart through the longest tech IPO drought in two decades.
Vijaye Raji brought other complementary skills to the table: a decade scaling consumer products at Meta before founding Statsig, the experimentation platform OpenAI already relied on, acquired for $1.1B.
Together, Fidji & Vijaye are bringing each of their product iteration & scaling skills for both vertical AI applications and horizontal platform plays. Let's dive deeper into what each vertical looks like today, and what it could mean if OpenAI's bet takes off.
Below, we map nine verticals where OpenAI is pressing into established markets. Each follows the same framework:
what OpenAI aims to do → who currently owns this space → what breaks if OpenAI succeeds.
OpenAI's ambition here is to compress the entire hiring funnel into a single AI-native workflow inside ChatGPT: sourcing, screening, scored assessments, and shortlisting.
Incumbents in this sector: Microsoft's LinkedIn ($16B+ revenue), Recruit Holdings (Indeed), SEEK, ZipRecruiter, plus enterprise HR players like Workday and ADP, and credentialing platforms like Pearson and Coursera.
If OpenAI succeeds, low-end contingency recruiting (15-25% of first-year salary) compresses toward commodity pricing. Applicant Tracking Systems become glorified scoring pipes. LinkedIn must prove skills-verified funnels, not just profile reach. The job board model, built on volume and SEO, faces existential risk when the chat interface becomes the default talent discovery layer.
OpenAI is bundling teaching, testing, proctoring, and credential issuance into a single conversational flow with ID verification. Learn a skill, take a proctored assessment, earn a verifiable badge, all without leaving ChatGPT.
Incumbents in this sector: Pearson, RELX, Coursera, Udemy, and Skillsoft have built defensible positions by owning distribution, content libraries, and accreditation partnerships.
Non-licensure bootcamps and generic certificate mills will reprice downward. Platforms with recognised assessments, especially those tied to hiring outcomes or regulatory gates, become the new toll booths. If OpenAI's certifications gain employer recognition, the ed-tech stack fragments between those who own legitimacy and those who merely own content.
OpenAI's creative vertical collapses production timelines. Generate video spots with Sora 2, test variants, buy media, all within ChatGPT Atlas.
Incumbents in this sector: Adobe dominates creative software, Shutterstock and Getty control stock libraries, while Google and Meta own ad distribution and measurement. When production costs fall by an order of magnitude and creative testing scales 10-100x, agencies built on billable hours face existential margin compression.
Distribution follows attention. If Atlas becomes the new home page of the internet, channels that can demonstrate measurable ROI keep budgets while those relying on attribution opacity lose share. This shift isn't just about cheaper creatives. It's about creatives becoming a real-time, test-driven function embedded directly in the purchase loop.
OpenAI aims to close the full loop: lead generation → outreach → quote → cart → purchase, and support. Agent Mode handles workflows, Instant Checkout (powered by Stripe) handles transactions.
Incumbents in this sector: Salesforce, HubSpot, UiPath, Appian, and Pegasystems own CRM and workflow automation, while Shopify, PayPal, Block, and Adyen own checkout rails.
Weak attribution layers and middleware compress first. CRMs must prove closed-loop conversions from model-generated flows, not just activity dashboards. E-commerce platforms with first-party transaction data defend, while those relying on middleware get disintermediated. The question becomes: do you own the action, or just the logging?
Codex aims to handle boilerplate, refactors, tests, and migrations from natural language, compressing the "glue code" hours that justify developer tooling seats.
Incumbents in this sector: Microsoft's GitHub Copilot, GitLab, Atlassian, Datadog, and Dynatrace monetise developer productivity, but if Codex can generate and maintain infrastructure code conversationally, the value shifts from writing code to governing it.
Platforms with policy enforcement, guardrails, audit logs, and compliance workflows defend their seat growth. Those selling pure productivity (lines of code per hour) face margin pressure. The developer experience fragments into those who own the rules layer and those who merely speed up execution.
OpenAI wants to handle basic customer service: automated responses, finding information, processing returns, billing questions, and scheduling, with humans stepping in only for complex issues.
Incumbents in this sector: Salesforce Service Cloud, ServiceNow, Freshworks, NICE, Five9, and Twilio charge for support staff seats and call minutes. But if AI assistants can resolve 70-80% of basic queries at pennies per interaction, the entire cost structure collapses.
Outsourced support centres must reprice to "AI cost plus quality checking." Value flows to systems that can actually complete actions immediately: process refunds, change orders, update accounts. Companies just providing dashboards and call routing lose pricing power. Support shifts from counting staff headcount to measuring problems actually solved.
OpenAI's research tools aim to triage literature, pull structured data, generate draft papers with citations, and execute reproducible code, all conversationally.
Incumbents in this sector: RELX, Wolters Kluwer, Thomson Reuters, and Clarivate own legal, scientific, and financial information, while Google and Microsoft sell workspace copilots. The GDPval study already suggests tasks with shorter horizons and clear templates will flip first.
Junior research support compresses literature reviews, data pulls and draft memos. Mid-tier publishing layers that don't add validation or peer review lose pricing power, while provenance, evals, and reproducibility gain weight. The question becomes whether the produced research is legally defensible or scientifically sound.
Reported but not yet launched, Project Mercury aims to bring high-fidelity financial modelling and drafting to ChatGPT, trained and evaluated by ex-bankers.
Incumbents in this sector: S&P Global, Moody's, FactSet, and Thomson Reuters monetise financial data and analytics, while Accenture and the Big Four sell service hours. Templated junior tasks are the first to compress: comps packs, model hygiene, document drafting.
Defensibility here lies in workflow lock-in and proprietary data. If Mercury can replicate analyst-quality outputs, the value shifts to data vendors who own exclusive feeds and compliance-approved workflows. Services firms must prove advisory value beyond execution speed.
Atlas is the key to all the above verticals working seamlessly, capturing intent at session start and compressing the search-to-action pathway into a single conversational pane.
Incumbents in this sector: Google owns search, Microsoft owns the second-place browser, Atlassian is a player with its acquisition of The Browser Company. If Atlas becomes a daily habit, economics shift upstream. SEO mills and affiliate linkware fade when users never leave the chat interface to compare options.
The browser becomes the cart.
This is the land grab that makes everything else work. If OpenAI controls when the session starts, they control intent. If they control intent, they control the distribution layer for every vertical above.
This play makes all the more sense with the latest acquisition of Software Applications Inc. who make a macOS native interface called Sky that is a supercharged version of the Spotlight Search bar on your Mac.
Three strategic patterns reveal OpenAI's deeper gameplan:
The masterstroke is Atlas or the upcoming OS as the universal context layer. Traditional software lives in silos. LinkedIn knows your CV, Google knows your searches, and Salesforce knows your deals.
If OpenAI becomes where all knowledge work starts, they inherit the full context graph. Every search, every document, every workflow then becomes training data for better personalisation. This verticalisation isn't about winning individual markets but a strategy to become the cognitive layer where all economic input and output begins.
(our selection of the top 20 companies impacted by OpenAI's applications stack)
UiPath (NYSE: PATH) - Agentic Automation Enabler
UiPath remains a category leader in enterprise-grade RPA, with AI-native copilots embedded across its automation suite. Their comprehensive attended & unattended automation platform with deep enterprise integration creates switching costs and workflow lock-in that OpenAI's conversational tools cannot easily displace.
Q2 FY2025 revenue reached $362M (+14% YoY) with record operating cash flow of $78M and non-GAAP operating income of $87M. Cloud ARR grew 25% YoY and now comprises 31% of total ARR ($1.38B). New launches include AI Trust Layer and Autopilot for developer workflows. With sustained margin expansion, UiPath is evolving into a platform bet on vertical AI agents for finance, healthcare, and HR.
Stride (NYSE: LRN) - AI-Enhanced Virtual Schooling
Stride is embedding generative AI into its K-12 and career learning offerings via Legend Library and adaptive tutoring. Its 25-year relationships serving 220,000+ students through district partnerships with state-certified teachers, hands-on materials, and physical infrastructure are regulatory moats OpenAI cannot replicate.
Q2 FY2025 revenue reached $575M (+22% YoY), driven by Career Learning (up 35% YoY) and strong enrollment. FY2025 EPS guidance was raised to $4.96-$5.00 (from $4.60-$4.90), with EBITDA margin expansion underway.
Bill Holdings (NYSE: BILL) - LLM-Tuned SMB Automation
Bill.com's integration of generative AI into its SMB workflow stack (invoicing, AP, AR, expense) is improving customer retention and onboarding velocity. Its network serving 400,000 SMBs with $344B payment volume and partnerships with 60+ top accounting firms creates embedded financial workflows that resist disruption from the OpenAI/Stripe network.
Q4 FY2025 revenue hit $383M with full-year core revenue reaching $1.30B (+16% YoY) and 84% non-GAAP gross margins. With 86% of core revenue from existing customers and 94% dollar-based net retention, management is optimising pricing to lift ARPU.
To dig further into smaller companies bound to benefit from the AI Infrastructure buildout, create your own StockScreener like we did:
To find out more about ETFs available to index the AI Infrastructure buildout, make your own ETF Screener like we did:
OpenAI has the best models, the best product velocity, and the best execution talent in the industry. Fidji Simo and Vijaye Raji are building the machinery to ship verticals faster than incumbents can respond. The vision is coherent. The execution is real.
But distribution eats product for breakfast. Google owns search, the browser, the mobile OS, and the compute stack. Apple owns the hardware layer and 2 billion devices. Microsoft owns the enterprise relationships and the cloud infrastructure. Each of these companies can be good enough and still win through defaults, integration, and patience.
OpenAI's vertical-first strategy is a calculated bet that product quality and velocity can overcome structural disadvantages in chips, distribution, and platform control. It's not impossible. Facebook beat MySpace. Google beat Yahoo. Disruptors win when incumbents move slowly.
But Google, Apple, Amazon, Meta and Microsoft aren't moving slowly anymore. They're all spending tens of billions on AI. They're all integrating models into every product. They're all competing on the same verticals OpenAI is targeting.
The question is whether great products are enough when you're fighting incumbents who own the defaults, the hardware, and the distribution layer.
