Block's AI Strategy: How Square and Cash App Are Rethinking Customer Discovery in 2026

TL;DR#

Block's AI strategy in 2026 runs on a single conviction from CEO Jack Dorsey — "we're going to build this company with intelligence at the core of everything we do" — and it shows up in two product surfaces: Managerbot, a proactive Square AI agent that reached roughly one million businesses by April 2026, and Money Bot, a Cash App AI assistant that drew one million active users in a single week with no marketing. Both are built on "codename goose," Block's open-source agent framework that ran internally for about 18 months and helped lift production code shipped per engineer by more than 40% since September 2025. Block posted roughly $24.19 billion in 2025 net revenue and about $10.2 billion in gross profit (up ~18% year over year), spanning Square's ~4 million sellers and Cash App's 58 million monthly transacting actives. Yet the company still learns why a seller churns off Square or why a Cash App user stalls largely through transaction data, dashboards, and surveys — which capture behavior but not motivation. The gap that quantitative signals can't close is the "why," and that is where conversational AI customer research, like the AI-led interviews Perspective AI runs, completes Block's two-sided listening loop.

Block, Inc. is one of the clearest examples of a company that has wired AI into both how it ships software and how it serves customers — and one of the clearest examples of where AI-as-automation outruns AI-as-listening. This is a customer research read on the block ai strategy: where Square AI and Cash App AI genuinely move the needle, where forms and transaction analytics still bottleneck customer understanding across two enormous networks, and what conversational interviews unlock that no dashboard can.

What is Block's AI strategy in 2026?#

Block's AI strategy in 2026 is to put "intelligence at the core of everything we do" — embedding agentic AI into both its internal engineering and its two flagship products, Square (its seller ecosystem) and Cash App (its consumer network), all built on its open-source agent framework, codename goose. The strategy is unusually integrated: the same harness that automates Block's own code review now powers Managerbot for merchants and Money Bot for consumers, making Block both a builder and a heavy user of agentic AI.

Block operates two distinct two-sided networks. Square connects roughly 4 million sellers — restaurants, retailers, salons, service businesses — to their buyers, processing about $67.2 billion in gross payment volume in Q4 2025 alone, up 10% year over year. Cash App connects 58 million monthly transacting consumers to peer payments, banking, lending, and investing. Block's 2025 net revenue reached approximately $24.19 billion, with gross profit near $10.2 billion, up roughly 18% over the prior year, per the company's investor disclosures. The AI question for a company at this scale is not "should we adopt AI" — it's "where does AI deepen the customer relationship versus merely automate the transaction."

For product and customer experience teams studying how large platforms operationalize AI, Block sits alongside the other fintech case studies worth reading: how Stripe approaches AI customer research as a $95B payments leader, how Shopify listens across 4.6 million merchants, and how PayPal is rethinking discovery across 430 million accounts. Block's twist is that it must learn from two audiences at once.

Where Block uses AI today: goose, Managerbot, and Money Bot#

Block uses AI today across three layers — internal engineering automation (goose), proactive merchant operations (Managerbot on Square), and consumer financial guidance (Money Bot on Cash App) — all sharing one agentic backbone. This is the most vertically integrated AI deployment among major fintechs, and it is the foundation of the entire block ai strategy.

Codename goose is Block's open-source agent framework, launched publicly in January 2025 after roughly 18 months of internal use. Goose connects large language models — Anthropic's Claude, OpenAI's GPT family, and others — to real-world actions like checking code for bugs, drafting communications, and orchestrating internal tools. Block reported a greater than 40% increase in production code shipped per engineer since September 2025 through agentic coding tools, and in December 2025 it contributed goose to the Linux Foundation's new Agentic AI Foundation alongside Anthropic's Model Context Protocol and OpenAI's AGENTS.md, according to TechCrunch.

Managerbot is Block's proactive Square AI agent, announced April 7, 2026 and moved to open beta on April 28. Rather than waiting for a seller to ask a question, Managerbot monitors a business's performance and acts: it watches stock levels, sales velocity, and external signals like weather and local events to flag when inventory is about to run out, and it generates optimized staff schedules from forecasted demand. It runs on frontier models from Anthropic and OpenAI, wrapped in Block's own agent harness derived from goose. Block reported about one million businesses using Managerbot, per VentureBeat's coverage of the launch.

Money Bot is Cash App's consumer AI assistant, which reached general availability in early 2026 and pulled in one million active users in a single week with no in-app or external marketing. Money Bot detects patterns and surfaces potential problems — a forgotten recurring charge, or a timing mismatch between a bill's due date and an income deposit — turning Cash App from a passive ledger into a proactive financial copilot. The learnings from Money Bot fed directly into Managerbot, showing how Block's two networks cross-pollinate. This proactive-guidance posture mirrors what we documented in how Chime replaced forms in AI customer onboarding and SoFi's member-first conversational financial discovery.

The two-sided listening problem: behavior is not motivation#

Block's core customer-understanding gap is that its AI reads behavior brilliantly but rarely captures motivation — and a two-sided network multiplies the blind spot across both sellers and consumers. Transaction data tells Block what happened; it cannot tell Block why it happened, and "why" is what drives retention.

Consider the seller side. When a restaurant's GPV drops 30% month over month, Square's dashboards see the number fall. They cannot tell whether the owner is testing a competitor's point-of-sale system, struggling with seasonal slowdown, frustrated by a fee change, or quietly winding down the business. Managerbot can react to the symptom — reorder inventory, adjust schedules — but it is optimizing operations, not diagnosing the relationship. The decision to leave Square forms long before the last transaction clears, and it lives in language, not in a chart.

Now the consumer side. Cash App's gross profit per monthly transacting active grew 25% year over year to about $94 in Q3 2025 — a healthy monetization signal. But aggregate ARPU hides the individual story: why did a user who direct-deposited their paycheck for six months suddenly stop? Why does a Money Bot nudge land for one user and feel intrusive to another? Behavioral analytics flag the churn risk; they don't surface the reason, and the reason is the only thing a team can act on before the account goes dormant.

The traditional fix is a survey or NPS pulse. But surveys inherit the same flattening problem we cover in depth in why AI conversations beat surveys for real customer research. A five-point scale and a comment box force a frustrated merchant to compress a complicated story into a dropdown, and most never respond at all — NPS response rates commonly run in the single digits to low teens. You learn the score, not the story. That is the structural limit of quant-plus-survey listening at Block's scale, and it is exactly the gap the voice-of-customer program guide for 2026 was written to close.

Why forms and dashboards bottleneck customer understanding at Block#

Forms and dashboards bottleneck Block's customer understanding because they capture structured fields and aggregate metrics, but customer motivation is unstructured, contextual, and often uncertain — exactly what a fixed schema cannot hold. This is the central thesis behind Perspective AI's view that AI-first customer research cannot start with a web form.

Three failure modes show up repeatedly at platform scale:

1. Forms flatten people into schemas. A churning Square seller has to translate "I'm overwhelmed, the fees crept up, and a competitor's rep called me twice this month" into a rating and a 200-character box. The richest part — the sequence of events that built the decision — never makes it into the data.

2. Dashboards see correlation, not cause. Managerbot and Block's analytics can correlate a GPV dip with a category or a region. They cannot interview the seller mid-decision to learn that the real trigger was a single bad support interaction. The "why now" lives in conversation, not in a time series.

3. The highest-value moments are the messiest. "It depends," "I'm not sure yet," and "I was thinking about switching but…" are precisely the answers a form discards and a skilled interviewer chases. Those hedged, uncertain moments are where retention is won or lost.

For a payments company, the stakes are concrete: a 1% improvement in seller retention compounds across millions of merchants and billions in GPV. The same logic that makes the complete guide to AI-powered customer experience worth reading applies to Block — the listening layer, not just the automation layer, determines lifetime value.

What conversational AI interviews unlock across both networks#

Conversational AI interviews unlock the motivation layer that Block's transaction data and surveys can't reach — by talking to thousands of sellers and consumers in their own words, following up on vague answers, and capturing the "why" behind every churn, stall, or expansion signal at survey scale and interview depth. This is the missing complement to Managerbot and Money Bot: those agents act on behavior; an AI interviewer explains it.

Here is how it maps to Block's two networks:

An AI interviewer agent can run hundreds of these conversations simultaneously, follow up on a hedge like "the fees got confusing" with "which fee, and when did you first notice it," and return synthesized themes — not just transcripts. For inbound moments, a concierge agent that replaces the intake form lets a new Square seller or Cash App user explain their goal in plain language instead of choosing from a dropdown. For at-risk accounts, an advocate agent keeps the conversation going across the relationship. The result is continuous, qualitative listening at the scale Block's networks demand — the same continuous-discovery discipline laid out in how to build a voice-of-customer program from scratch.

This is not a replacement for goose, Managerbot, or Money Bot — it's the layer that tells Block what to build next and which accounts to save first. Teams evaluating where this fits in their stack can compare options in the customer research tools modern product and CX teams actually use and the buyer's framework for AI customer engagement software in 2026.

How Block's approach compares to peers profiled this batch#

Block's approach stands out because it serves two distinct audiences — merchants and consumers — through one shared AI backbone, where most peers optimize for a single side of the relationship. That dual mandate makes the motivation gap twice as costly and the conversational listening opportunity twice as large.

Other platforms in this analysis face structurally similar gaps. Coinbase's conversational onboarding challenge and Toast's restaurant-platform AI work both wrestle with merchant churn diagnosis, much like Square. On the developer-and-issuer side, Marqeta's card-issuing platform and Bill.com's SMB finance automation face the same "behavior without motivation" limit. And the broader fintech cohort — Affirm's merchant onboarding and customer discovery, Robinhood's customer conversations, Brex's conversational discovery, and Plaid's open-banking listening across 8,000 fintechs — all illustrate that automation maturity has outpaced listening maturity industry-wide. Block is simply the most advanced example of the pattern, which makes the unfilled "why" gap most visible there.

Frequently Asked Questions#

What is codename goose, Block's open-source AI agent?#

Codename goose is Block's open-source AI agent framework, launched publicly in January 2025 after about 18 months of internal use. It connects large language models from providers like Anthropic and OpenAI to real-world actions such as code review, drafting communications, and orchestrating internal tools. Block contributed goose to the Linux Foundation's Agentic AI Foundation in December 2025, and it serves as the backbone for both Managerbot and Money Bot.

What does Square AI's Managerbot actually do?#

Managerbot is a proactive Square AI agent that monitors a business's performance and acts on it autonomously. It tracks stock levels, sales velocity, and external signals like weather and local events to flag inventory needs, and it generates optimized staff schedules from forecasted demand. Announced in April 2026 and moved to open beta later that month, Managerbot reached roughly one million businesses and runs on frontier models from Anthropic and OpenAI wrapped in Block's goose-derived agent harness.

How does Cash App's Money Bot use AI?#

Money Bot is Cash App's consumer AI assistant that proactively detects financial patterns and potential problems for users. It flags issues like forgotten recurring payments or a timing mismatch between a bill's due date and an income deposit, turning Cash App into a financial copilot rather than a passive ledger. After reaching general availability in early 2026, Money Bot drew one million active users in a single week with no marketing, and its learnings informed the Square-side Managerbot.

Why isn't transaction data enough for Block's customer research?#

Transaction data isn't enough because it captures what customers do, not why they do it — and motivation is what drives retention and expansion. A dashboard sees a seller's gross payment volume fall or a Cash App user stop direct-depositing, but it can't reveal whether the cause was a fee change, a competitor's outreach, a bad support experience, or seasonality. That "why," which lives in language, is invisible to analytics and largely uncaptured by short surveys.

How can conversational AI interviews improve Block customer research?#

Conversational AI interviews improve Block customer research by capturing the motivation layer that transaction data and surveys miss, at a scale neither manual interviews nor static forms can match. An AI interviewer can run hundreds of conversations with sellers and consumers simultaneously, follow up on vague or hedged answers, and synthesize the real reasons behind churn, stalled adoption, or onboarding drop-off. This complements automation agents like Managerbot and Money Bot, which act on behavior but don't explain it.

Conclusion: the listening layer Block's AI still needs#

Block's AI strategy is among the most ambitious in fintech — goose ships its software faster, Managerbot runs its sellers' operations, and Money Bot guides its consumers' money, all on one agentic backbone serving roughly 4 million Square sellers and 58 million Cash App actives. But automation maturity is not the same as listening maturity. Across both networks, Block still learns why a seller leaves or why a consumer stalls mostly through dashboards, transaction data, and the occasional survey — tools that read behavior precisely and motivation barely at all. The unfilled piece of the block ai strategy is the conversational listening layer: talking to customers in their own words, at scale, to capture the "why" behind every number.

That is exactly what Perspective AI was built for. Instead of forcing merchants and consumers into a form's dropdowns, Perspective AI runs AI-led interviews that follow up, probe, and surface the real reasons behind churn, adoption, and expansion — hundreds at a time. If you're building customer research for a two-sided platform, start a study in minutes, see how it works, or explore the platform built for CX and product teams. The companies that win the next decade won't just automate the transaction — they'll understand the customer behind it.