Shopify CEO says employees’ ‘slop grenades’ are making more work for everyone else

Shopify CEO Tobias Lütke and Duolingo CEO Luis von Ahn warned that indiscriminate AI use is creating polished but error-prone "workslop," reversing earlier AI-first messaging. A survey of 962 U.S. desk workers found 52.7% had sent workslop and recipients spent an average 3.4 hours per month revising it, equivalent to about $186 monthly per employee and up to $9 million annually for a 10,000-person organization. The findings highlight productivity, quality-control and workplace-collaboration risks as companies scale generative AI deployment.

Analysis

This is not an AI-demand read-through; it is a monetization-quality warning. For SHOP, mandatory-use culture can shift costs from content creation to review, exception handling and customer-support remediation, limiting the operating-leverage benefit investors may be embedding in future opex. The relevant KPI is not AI adoption but whether merchant-support resolution times, engineering velocity and gross-margin-adjacent service costs improve without higher rework; absent proof, a productivity narrative deserves a lower multiple rather than an earnings haircut today.

DUOL faces the more direct risk because content quality is core to retention and brand trust. Incremental AI-generated lesson volume is only economically valuable if completion, subscription conversion and churn are at least neutral by cohort; a quality-control layer can erase much of the presumed variable-cost advantage and delay content cadence. Over the next 1-3 months, the issue is chiefly narrative and management-credibility risk; over 6-18 months, a measurable moderation/review burden would constrain margin expansion expectations.

The contrarian point is that this is potentially constructive for application software vendors that sell workflow controls rather than raw generation. If enterprises discover that unconstrained generation creates audit and rework costs, spend should migrate toward governance, evaluation, observability and human-in-the-loop orchestration. SHOP and DUOL could ultimately benefit if they demonstrate disciplined, bounded deployment, but the public comments alone are not independently verifiable evidence of a financial impact.

Key Decisions for Investors

• Do not initiate a directional SHOP or DUOL position solely on this signal; treat it as a multiple-risk watch item rather than a near-term earnings catalyst.
• For existing DUOL longs, reduce exposure or hedge through the next earnings print if consensus assumes AI-led content-cost leverage. Re-add only if management provides cohort-level evidence that AI-assisted content preserves engagement and paid conversion while sustaining contribution-margin guidance.
• For existing SHOP longs, require evidence in the next two quarterly disclosures that AI tools reduce merchant-support or engineering costs without deterioration in service metrics; a guidance cut to operating-margin expansion or higher support expense would falsify the productivity thesis.
• Watch long PANW or CRWD versus short a broad software basket only if enterprise surveys or vendor bookings show a clear shift toward AI governance/security budgets; this article alone does not establish sufficient demand data for a trade.