Up the Stack & Down to Business
Claude suits up for finance. ChatGPT starts shopping. The stack is collapsing - and the monetization games are just beginning.
This week, while Grok dropped semi-pornographic companion avatars and rolled out Grok 4 with no model card, no evals, and absolutely no restraint, Anthropic and OpenAI made serious moves up the stack.
In just one week, the two labs doubled down on four massive markets - and made incumbents everywhere very nervous.
Finance
Anthropic launched Claude for Financial Services - a fully verticalized, enterprise-ready AI offering for banks, asset managers, and insurers.
This isn’t just “Claude with a finance prompt.” It comes with:
Prebuilt connectors to PitchBook, FactSet, Snowflake, S&P Global etc
Domain-specific safety controls
Big Four consulting firms as implementation partner
They demoed it by asking Claude to benchmark a company against industry comps. It spat out a clean comps table in seconds. If you’ve ever done that manually in Excel, you’re already crying.
Dev Tools
On top of that, Claude Code is on a tear. Anecdotally, I’ve seen a surge in adoption across the ecosystem - and now there’s data to back it: over $200M ARR, growing fast. Usage surged 5.5x since May.
They recently launched a real-time enterprise dashboard that tracks lines of code accepted, suggestion acceptance rates, per-user activity, and team-level ROI. Anthropic’s pitch just shifted from “your devs like it” to “here’s the spreadsheet your CFO can present to the board.” They’ve solved the problem haunting every enterprise AI deployment right now: How do we measure ROI? Claude now comes with receipts.
Knowledge Work
Meanwhile, OpenAI is trying to become your junior analyst. They launched ChatGPT Agents yesterday. An autonomous assistant that runs on your behalf, takes feedback, completes multistep workflows across interfaces and tools, and wait for it.. generates fully editable PowerPoints and spreadsheets. They are inching closer to the promise of labor replacement as they move from autocomplete to autonomy.
Ecommerce
Separate from the agent push, OpenAI is also rumored to be building native checkout into ChatGPT, enabling users to discover, compare, and buy products mid-conversation. Shopify is in the mix. Merchants will pay a rev share. OpenAI’s product becomes the new point of sale.
Why it makes sense:
Native surface → no app-hopping
Mid-convo purchase → zero context switching
Checkout-as-interface → no middleman rev share
This follows Perplexity’s in-chat PayPal integration from two months ago. The transaction layer is collapsing into the discovery layer, turning soft intent into hard conversion. We’re not just reinventing search. We’re reinventing how people shop.
The Strategy Underneath
Foundation model labs aren’t stopping at the model. They’re compressing the entire stack: Infra → Model → App → Marketplace. In doing so, they’re rewriting enterprise software economics.
The new battlefronts?
Who owns distribution (browser, agent, workflow)
Who captures margin (API, vertical product, transaction)
Who builds trust (esp. in regulated domains like finance, health, gov)
So, What Now? For Startups?
In a world where intelligence is commoditizing, models have no choice but to go vertical - own the UX, the workflow, the margin.
The problem for startups is that you don’t know where the labs will go next. They can drop into your market like a piano from the sky. So picking your spot matters more than ever.
Option 1: Outrun the Labs
If you target broad domains like finance or coding, you have to bring your A-game: Superior UX, deep integrations, relentless shipping, and really hammer home that multi-model support and data privacy.
Option 2: Hide in the Niche
Not “AI for finance.” “AI for commercial mortgage underwriters”
Find a niche too small for the labs to care about - and serve it better than anyone else.
General-purpose AI is now a scale game. Play a different game. The more tailored your wedge, the more defensible your edge.
🧂One Last Thing
Every success story is an exception to the rule.
So take all startup advice - especially mine - with a grain of salt.
The stack may be collapsing, but the right niche still compounds.
Really well written! Agree w/ everything said.
Question: Do you believe building custom agents for enterprise workflows is a sizable market that will continue to expand? Or will general-purpose agents from the big model companies eat into that share as well?
Brilliant !