Block Is Testing Whether AI Can Rewrite Operating Leverage
Executive summary
Block is making one of the clearest public claims that internal AI use can alter the cost and speed of software development. In first-quarter 2026 materials, the company said production code changes per engineer had increased more than 2.5 times from January to mid-April and cost per production code change had fallen more than 70% in the quarter.
The relevant question is not whether engineers produce more changes. It is whether customers receive better products with fewer losses and less friction.
Financially, Block entered the test with momentum. Gross profit rose 27% to $2.91 billion, and adjusted operating income reached $728 million, a 25% margin. Cash App grew much faster than Square, making it essential to distinguish company-wide efficiency from segment-specific demand.
The market in context
Block combines a consumer financial ecosystem and a merchant platform. That structure creates shared capabilities in identity, risk, money movement, data, and engineering while exposing the company to different customer cycles.
| Q1 2026 metric | Result | Year-over-year growth |
|---|---|---|
| Block gross profit | $2.91B | 27% |
| Cash App gross profit | $1.91B | 38% |
| Square gross profit | $982M | 9% |
| Adjusted operating income | $728M | Not comparable to gross-profit growth |
| Adjusted operating margin | 25% | Company-defined non-GAAP |
Source: Block Q1 2026 shareholder letter. Adjusted measures are non-GAAP; refer to the company's reconciliation.
Block raised its 2026 gross-profit growth expectation to 19% and adjusted diluted EPS growth expectation to 62%. These are forecasts. The operational AI measures are also company-defined and have not been independently audited.
Code throughput is an input metric, not a customer outcome
More production changes per engineer can mean faster experimentation, automation of routine work, improved tooling, or smaller changes. It can also increase review burden, defects, or measurement gaming. Cost per change can fall because the denominator rises without an equivalent increase in product value.
A robust evaluation needs downstream measures.
| Layer | Useful measure | Failure mode |
|---|---|---|
| Development | Lead time, cost per validated change | Counting trivial changes |
| Quality | Rollbacks, incidents, escaped defects | Shipping faster with lower reliability |
| Risk | Fraud loss, false positives, control exceptions | Automating weak decisions |
| Customer | Activation, retention, support contacts | More features without utility |
| Economics | Gross profit per employee, margin, payback | One-time cuts presented as compounding productivity |
The distinction is especially important in payments. A defect can affect money movement, settlement, account access, tax reporting, or merchant operations. AI-generated speed must be paired with stronger validation and observability.
The segment mix can make the productivity thesis look stronger than it is
Cash App gross profit grew 38%, compared with 9% for Square. Cash App produced nearly twice Square's gross profit. Consolidated growth therefore benefits from a faster consumer engine even if merchant improvements take longer.
Square's economics depend on seller acquisition, gross payment volume, software attachment, hardware, and the health of small businesses. Cash App can grow through engagement, card use, deposits, lending, and monetization. Shared AI tooling may benefit both, but the commercial translation differs.
A better scorecard allocates productivity benefits by segment and product. Did seller onboarding shorten? Did dispute resolution improve? Did Cash App support contacts fall? Did fraud controls improve approval without increasing loss? Company-wide engineering throughput cannot answer those questions.
Workforce efficiency can create reinvestment capacity or service fragility
Block has argued that smaller teams equipped with AI can do more. The upside is lower coordination cost, faster decisions, and more capital for product or margin. The risk is loss of domain knowledge and thinner operational coverage.
Payments businesses depend on expertise that is difficult to encode completely: scheme rules, compliance, investigations, incident response, local market behavior, and edge-case servicing. Automation may remove routine work while making the remaining cases harder. Staffing should follow residual complexity, not average ticket volume.
A smaller team is only more productive if risk and customer work do not reappear as losses, outages, or unresolved exceptions.
Implications for operators
Executives evaluating similar programs should define a measurement chain before reducing capacity.
- Baseline the workflow. Measure current lead time, labor, defects, review, and rework.
- Instrument AI use. Record where agents propose, execute, or approve work and which model or policy version applied.
- Protect validation. Maintain independent review for financial, security, compliance, and irreversible changes.
- Track customer outcomes. Link internal productivity to activation, resolution time, reliability, fraud, and retention.
- Release savings gradually. Do not remove control capacity before observing residual workload across a full operating cycle.
For merchants using Square, the practical questions are product-specific. Faster release cadence matters only if it improves uptime, authorization, funding, reporting, or labor efficiency. Merchants should evaluate those outcomes rather than infer service improvement from Block's corporate margin.
For investors and strategists, gross profit per employee can be informative but incomplete. It improves after workforce reductions even if innovation later slows. Pair it with product launch adoption, incident rates, segment retention, and support performance.
Block's targets should be translated into operating measures that distinguish faster software delivery from better business performance. Engineering teams can track cycle time, deployment frequency, incident rate and rework, but management should connect those measures to product adoption, loss rates, support contacts and gross profit. More code shipped at lower apparent cost is not leverage if defects, fraud exposure or customer confusion rise with it.
The largest risk is denominator management. Cost per change can fall because the number of small changes rises, even when customer value does not. Productivity reviews therefore need paired quality measures and a stable definition of what counts as a production change. Teams should also separate savings from automation, role elimination, vendor consolidation and lower hiring. Those mechanisms have different durability and different effects on capacity.
For Cash App and Square, AI leverage will show up differently. Consumer products may benefit through support automation, personalization and faster risk decisions, but mistakes can affect access to money and trust. Seller products may gain from quicker onboarding, better fraud detection and automated back-office work, while unreliable decisions can interrupt a merchant's cash flow. A single company-wide productivity ratio will miss those asymmetries.
Capital allocation is the final test. If AI genuinely lowers the cost of building and operating software, Block can choose among higher margins, faster product investment and lower prices. Investors should look for consistent evidence across those choices: operating expense growth below gross-profit growth, stable product quality, better release velocity and sustained engagement. Temporary restructuring benefits or a favorable comparison period are not enough.
Operators outside Block should resist copying the headline target without the measurement system behind it. The defensible approach is to automate a bounded workflow, record the prior labor and error baseline, measure the customer outcome and keep a rollback path. Scaling should follow only when economic gains survive the inclusion of review, exception handling and model oversight.
Workforce effects also belong in the calculation. A smaller organization can make decisions faster, but only if product knowledge, control ownership and incident response remain intact. Management should monitor span of control, review queues and time spent correcting automated work alongside payroll savings. The clearest evidence of durable leverage would be a team that ships useful improvements more quickly while customer complaints and control failures stay flat or decline. If review work simply moves to fewer senior employees, the cost has been displaced rather than removed. That distinction should remain visible in quarterly reporting, especially while restructuring savings and AI adoption occur at the same time.
What would change the view
The first risk is definitional. Production code changes can vary in size and importance. Cost allocation may exclude review, infrastructure, or downstream support. The reported 70% decline may reflect early adoption and could normalize.
The second is causality. Revenue growth, restructuring, and product mix can improve margins independently of AI. A single quarter cannot isolate the effect.
The third is tail risk. Faster automated development can create rare but material failures. Average productivity measures do not price those events.
Evidence supporting Block's thesis would include sustained release speed, stable or lower incident rates, better customer outcomes, declining unit support cost, controlled fraud losses, and continued product adoption. Evidence against it would include rising rollbacks, service deterioration, slower merchant growth, repeated control failures, or re-hiring to restore lost capabilities.
Methodology
All financial and internal productivity numbers come from Block's Q1 2026 shareholder letter and 8-K. Gross profit is emphasized because Block revenue includes pass-through components. Adjusted operating income, margin, and EPS are non-GAAP.
The internal measures are management disclosures without a standardized industry definition. Comparisons with other companies would be unreliable unless change size, validation, cost allocation, and production scope were harmonized.
Chart build 1: Plot Cash App and Square gross profit growth for eight quarters, with each segment's share of Block gross profit. Source: quarterly shareholder letters.
Chart build 2: Create an AI productivity scorecard with January 2026 indexed to 100 for changes per engineer and cost per change. Add quality, incident, support, and customer metrics only when disclosed; leave unavailable cells blank. Do not infer missing outcomes from the two reported engineering measures.