In April 2025, Shopify CEO Tobi Lütke sent a memo that required teams to demonstrate why AI cannot do a job before asking for more headcount or resources. It also declared that “using AI effectively is now a fundamental expectation of everyone at Shopify” and added AI usage questions to performance and peer review questionnaires. Lütke published the memo himself after learning it was being leaked. More than a year on, the memo has aged into something more interesting than a policy: it is the cleanest published description of how an AI-native company runs its resource allocation.

The Operating Pattern

Read as an operations document rather than a hiring announcement, the memo does three specific things.

First, it inverts the burden of proof. In a conventional company, a team requesting headcount argues for the value of the work. Under the Lütke rule, the team must first produce evidence of an attempted automation and its failure. That converts every budget request into an eval: a defined task, an attempted agent implementation, a documented gap. The organisation accumulates a map of what AI currently cannot do — refreshed continuously, because the frontier moves.

Second, it makes AI usage measurable at the individual level. Putting AI questions into performance and peer review questionnaires turns “we use AI” from culture-deck language into an assessed behaviour, the same way code review or sales quotas are assessed.

Third, it prices labour against automation explicitly. The question “would we rather hire, or build the agent?” stops being philosophical once every headcount request carries an automation-failure attachment. The org chart becomes a routing decision between human and machine capacity.

Why It Matters

Most companies bolt AI onto an unchanged operating model and get unchanged results. The headcount gate is different because it changes the default: automation is presumed possible until demonstrated otherwise. For operators, that ordering is copyable tomorrow — it requires a policy, not a platform. For anyone evaluating companies, it offers a sharp diligence probe: ask a founder claiming to be “AI-native” when they last rejected a headcount request because an agent closed the gap. Companies running this loop can name the task, the agent, and the date. Companies that can’t are describing tooling, not operations.

The Charaka View

We run this pattern at its logical extreme: Manthan Intelligence operates with 49 agents in production against a two-person human team — a 24.5:1 ratio (as of 17 July 2026, per our internal registry), covering research, analysis, design review, fact-checking, and infrastructure. Every proposed new capability starts with the question the Shopify memo institutionalised: why can’t an agent do this? The honest answer changes month by month, which is exactly the point. A headcount gate is not a hiring freeze — it is a forcing function that keeps an organisation’s map of AI capability current. The companies that treat it as a one-time memo will get a one-time gain; the ones that treat it as a standing loop are building a different kind of firm.

This analysis draws on Fortune’s coverage of the Shopify memo, BetaKit’s reporting, and Lütke’s published memo. Human editorial oversight applied.

This analysis is informational and does not constitute investment advice, a research report, or a recommendation to buy, sell, or hold any security.

Charaka Notes by Manthan Intelligence. Subscribe