Block should focus on being an AI-Enabled Fintech, Not an Applied AI Company
Block just beat earnings and raised its full-year guidance, and the stock slid anyway. That’s the backdrop for the post its head of investor relations put out afterward, and here’s the line that stuck with me:
“We became an elite applied AI company. We’ve launched numerous industry firsts, including an enterprise-grade agentic harness, Goose, and a frankly incredible agentic collaboration platform, Buzz, where agents are first-class participants.”
Read that again and it isn’t a victory lap this sounds like. It’s a company insisting it’s great to an audience that wasn’t buying it. “An elite applied AI company” is what you say when you need a story to be true, not when it plainly is. And I think the story is wrong.
It’s important for folks to realize that I don’t work at Block anymore, as my role was cut in this year’s restructuring. But I’m still an equity holder, I’m long Block as a fintech, and I want the company to win. I led global strategic partnerships finance there, and from that seat the reasons were obvious. Cash App and Square are two of the best distribution engines in the industry, the balance sheet is real and increasingly powered by Square Financial Services, and the team has already shipped genuinely useful AI into products millions of people touch.
That is exactly why the “applied AI” framing worries me instead of exciting me. What’s actually working at Block is AI-enabled fintech, not generalized AI. And on the three things that decide it, data, product, and go-to-market, Block isn’t built to win the applied-AI fight it’s picking. It should finish the job on its own turf first.
The distinction
These two words look like the same thing, and they aren’t so let’s make sure our definitions are clear.
Applied AI, the way Block is using it, means a general intelligence company. “An intelligence,” as Jack puts it. A horizontal platform pointed at many domains, with finance as just one application. An intelligence layer on top of music, consumer finance, business finance, Bitcoin. Goose and Buzz are purpose-built to be general, because being general is how you signal you’re more than a fintech. That framing is familiar to me: it’s exactly how the first startup I worked at, Sift, pitched itself, a machine learning platform whose first application was fraud. But as you go deeper you realize you have to specialize, and you become a fraud-prevention ML platform, not an ML platform with a fraud app. Unless you’ve actually built the underlying model, like OpenAI or Anthropic, you can’t point it everywhere. Just like Sift, it’s harder than you think to be a generalized AI company.
AI-enabled fintech is the other shape, and the one Block should own. This is financial intelligence: AI whose primary and defining application is money. Built on financial data, aimed at financial products, sold to financial customers, monetized through financial outcomes. Block isn’t running a horizontal platform that happens to do finance. It’s a fintech leveraging the AI muscle it built inside financial products. What it has, and what almost nobody else has, is financial intelligence. That’s the category it can own, instead of reaching for a generalized-AI framing, and the AI-lab multiple that comes with it, that it can’t back up.
Data
Block’s data is world-class, but it’s the wrong kind for this. It has enormous depth, billions of payments and commerce transactions, but no breadth: no codebases, no developer telemetry, no internals of a software business, none of that business literature shaved off the bindings and ingested into a global model like the terrifying videos you see on social media. Block is the operating system for a restaurant or a retailer, or it’s the consumer money app. That’s fantastic for financial products and close to useless for building developer tools.
It’s worse than that, because half of their business, Square, isn’t even built for software businesses. You can’t really sell SaaS on Square, I’ve tried. The subscription tooling is a commerce bolt-on for recurring product orders, not developer-grade billing, and Cash App Pay and Afterpay aren’t even supported for subscription payments. Compare that to Stripe, built for developers from day one, now shipping a payments foundation model trained on tens of billions of transactions. And even Stripe, with the developer ecosystem and the data to match, isn’t trying to be a frontier lab. It’s trying to be the best infrastructure for payments, which is exactly right. So I struggle to see Block spinning up frontier models from POS receipts and P2P transfers, let alone a credible GitHub competitor with no code data behind it. Even Cursor, built for developers and training its own frontier model on Colossus, has only a long shot at dethroning GitHub, and it’s far better positioned than Block.
Product
Two problems here: product differentiation and product precedent.
On differentiation, the missing data and missing developer base show up in the products. Take Goose. I’ll grant Block the “first coding harness” claim, but first means nothing without share, and the number they cite is internal: 60% of employees using it weekly, which from personal experience is surprisingly low given they were jamming it down everyone’s throats the year I was there. Who cares about internal usage? Tell me about real market share with real developers, and then we can talk. I never hear anyone outside Block call Goose their daily driver. It has broad tool use, but not the interface to beat a T3 Code, or the gravity to pull anyone off Codex or Claude Code. Buzz is lost in the sauce from the other direction: no foundational model of its own, it runs on Nostr and is self-hosted (a real onramp tax), and early users say it’s either too complex or just looks like Slack. And “agents as teammates” isn’t new, Salesforce and Slack shipped agents with their own identities in 2024. Mattermost is already an open-source, AI-enabled, enterprise-ready team platform Block could have forked. If Buzz is the most ambitious new thing Block has shipped lately, that doesn’t inspire confidence.
On precedent, set the generalized stuff aside and look at the financial AI Block should already be great at that it’s still struggling to get to exit velocity on. Moneybot is fun, but nobody’s using it as their primary personal finance manager, and it can’t see past the Cash App walled garden. Managerbot has real potential, a co-manager for a business, and already reaches a million Square sellers, but it’s treated as won when it’s barely begun. Square’s own voice ordering is being outperformed by voice agents built on LiveKit. If Block gets lapped on a comparatively solved problem like voice, why trust it on the genuinely hard ones?
What frustrates me most is the low-hanging fruit it’s ignoring, any of which would do more for its AI story than another open-source dev tool:
Transaction enrichment in Cash App, so every line item is legible instead of a cryptic merchant string.
AI underwriting for Square loans and Cash App Borrow, where Block’s repayment data is a real moat it never talks about.
AI-managed investing that turns a self-directed brokerage into a guided one.
Generative UI in Moneybot and Managerbot, so the interface adapts to what you’re actually trying to do.
Automatic bookkeeping for Square sellers, what Puzzle does for startups.
Agentic dispute resolution, one of the most painful and most automatable workflows in fintech.
Afterpay affordability personalization, underwriting far more than pay-in-four.
Bitcoin tax-loss harvesting, native to Cash App’s crypto rails.
Go-to-market
Monetization is where the story is weakest, and where Block’s own choices sink it. It made Goose free and open source, then handed its governance to the Agentic AI Foundation under the Linux Foundation, so it doesn’t even control the roadmap. You can’t have a roadmap to monetization for something you’ve given away. Even if Goose were the best agent out there, there are too many free harnesses, and with no foundational model underneath, nobody pays for the harness alone. Buzz is worse: no reason to pay today, and prying even one small business off Slack, which has real cross-org network effects, is brutally hard.
Distribution is the other half. Say Block builds something great. Who buys it? Its base is non-technical, POS merchants and mainstream consumers, so there’s no cross-sell into developers and no built-in demand. GitHub stars don’t mean anything, plenty of 50,000-star repos die. Block has never sold to that buyer, individual or enterprise, which is why it hasn’t happened and why the plans read so vague for a company that’s usually this precise.
The path forward
None of this is a knock on Block’s ambition. The winning move is just to not be something it isn’t. Don’t be the generalized AI company that also does finance. Be the definitive financial intelligence: the company that makes money legible, underwriting instant, disputes self-resolving, investing guided, and bookkeeping automatic. Do it with the taste Block is known for, because as function gets commoditized, form is what people choose, and form is exactly where Block wins. That’s a category with real data moats, customers who already trust it, and an obvious way to charge. Applied AI, the way Block is selling it, has none of those. So ship the roadmap only Block can ship, and let Goose and Buzz be what they already are, useful internal tools and open-source goodwill, not the headline.
Maybe I’m wrong. Strong opinions, weakly held, as always. If Block sees it differently, the fix is simple: hold another investor day and show, precisely, how it differentiates on data, product, and distribution.
Until then, I’m long Block as an AI-enabled fintech, a financial intelligence that should own its category. I’m not buying Block the applied AI company.