Requiring AI before hiring: the reversed burden of proof
On April 7, 2025, Shopify CEO Tobi Lutke required his teams to demonstrate AI's inability to perform a task before any request for additional staff.
How to require AI before any new hire without creating new resistance?
First, specify what AI must accomplish and who is responsible for the proof. Do this before announcing the decision. When 37 simulated voices reacted to Shopify's decision, a little less than half declared against it. Doubt primarily concerned how to implement this requirement.
The context, in plain terms
On April 7, 2025, Tobi Lutke, Shopify CEO, publicly released an internal memo. It established a new hiring policy. Under this directive, teams must now demonstrate that artificial intelligence cannot perform the work. This must happen before they request additional staff or resources. CNBC and TechCrunch reported this measure as sent internally in late March 2025. It then became public in early April. Bloomberg covered the topic on April 10, 2025. It treated this as an active policy change within the company. Publicly, the exact send date of the internal memo is not established. Reports only mention a late March 2025 distribution before the April 7 publication.
At Shopify, AI before any new hire
On April 7, Tobi Lutke's memo became public. Seen from the top, the rule puts the burden of proof on managers who hire. For teams on the ground, however, this requirement reads as a new constraint. It is an added burden even before considering growth.
Facing this change, a little less than half of the simulated voices declare against the decision. About one voice in five expresses doubt. About one voice in three supports it.
The group that counts for the most in this response is Shopify merchants, about one voice in seven of the panel. This is because the decision concerns them directly.
Before this panel, an internal hiring rule at Shopify is judged first by its merchants.
When AI proof shifts the burden
What holds it back first is doubt about the execution of this new policy. The question is not so much whether AI is capable. It is how teams must prove its inability. It is also what the consequences will be in case of failure or partial success. This uncertainty creates a new form of pressure on managers and employees.
Objections center on the administrative and cognitive burden this proof demands. Shopify engineers, about one voice in seven of the panel, are among the most critical. The simulated leadership voices, also about one voice in seven and carrying the decision, see a way to rethink workflows, provided adoption is real.
On this point, the Reversed Burden of Proof lacks clarity. One can support the principle of AI without believing the justification process will be fair or effective. At Shopify, the question switches sides: AI no longer has to prove itself; the hiring manager does.
For Shopify teams, showing what AI cannot do becomes a job in itself.
The data analyst who says yes, under pressure
Amid the reservations, one voice stands out among Shopify employees. This group otherwise leans against the decision. A junior data analyst, declaring for the measure, expresses a nuanced perspective: "I’m already automating my tasks and helping others, but now the bar’s raised and I’m the ‘teacher’s pet.’" The analyst says yes and, in the same breath, says the bar has gone up.
This voice shows that AI adoption can create unexpected team dynamics. Some position themselves as pioneers. Others fear being left behind. Conversely, a seasonal customer care temp, declaring against, anticipates a future where complex, non-automatable tasks will fall to them. This comes with a risk of job loss: "The bot handles the easy tickets, leaving the complex ones for us temps, if it ‘proves’ it can handle the load, I’m out before January."
We ran the exercise three times: same answer.
In this panel, the same rule reads as recognition to some and a threat to others.
For a policy that lightens, not burdens
For a leader facing a similar decision, the key is not to settle for a mandate. One must clearly articulate proof criteria. Define roles and responsibilities in this process. Most importantly, measure the real impact on team workload. Transparency about objectives and consequences matters. This prevents the Reversed Burden of Proof from becoming a brake on innovation rather than a catalyst.
Even the simulated chief executive's voice, which declares for the decision, sets a condition: "The mandate will force teams to rethink workflows, but I need to tie adoption metrics to bonuses to avoid shallow compliance." Without follow-up and incentives, the rule risks yielding paperwork rather than adoption.
Publicly, nothing establishes that the rule is still enforced in this form. But the panel's reactions suggest finer communication on execution methods could have eased initial fears. If the memo is an AI roadmap, it must first reassure about the Reversed Burden of Proof it imposes.
The success of an AI policy at Shopify depends on managing its new burden.
What you just read comes from a rehearsal, not a report. The exercise was run on the Kapari test bench, before a panel of 37 simulated voices. This rehearsal showed that Shopify's decision to prioritize AI created a new burden for teams. It showed that doubt centered first on implementing the proof. It also revealed that a junior data analyst can support the decision while highlighting increased pressure. The same exercise is done on a decision not yet announced. This helps identify friction points before it becomes public.
The questions readers ask
How to ensure managers do not circumvent the AI-before-hiring rule?
The memo from Shopify CEO Tobi Lutke places the burden of proof on managers. In the panel of simulated voices, the chief executive's voice sets a condition: tie AI adoption metrics to bonuses to avoid shallow compliance. This encourages real integration rather than mere justification.
What are the risks for temporary employees facing AI adoption?
AI adoption can redefine roles, especially for temporary employees. In the panel of simulated voices, a seasonal customer care temp expressed concern that AI might handle easy tasks. This would leave only complex ones for them, with a risk of non-renewal. Say early what happens to these roles.
Is this a poll or a prediction?
The voices in this panel are simulated. They are neither an opinion poll nor a prediction. The numbers cited are for a simulated panel of 37 voices, never a share of public opinion. Facts come from dated and named sources, verified beforehand. Kapari sheds light on the decision; it does not make it.
How Kapari computes and reads its signals: the method
Related cases
No other published case is about the same kind of decision. See all Hub cases
Your next decision deserves the same scrutiny.
Run it through the test bench before you announce it: a panel of voices reacts, you read the range and you see the frictions coming.
Start free