Designing for the Next Billion Users in AI
World
Research
Designing for the Next Billion Users in AI
Approaching accessibility and inclusive design as AI becomes the default interface for work.
Translating an interface is the visible part of localization and the smallest part of the work. The harder problems live in the data model: currencies, fiscal calendars, regional business conventions, and the assumption — buried in a hundred places — that a week starts on Monday.
Number and currency handling Formatting is straightforward. Conversion, rounding policy, and which currency an aggregate should even be expressed in are business decisions that need an owner. Calendar assumptions Fiscal years, weekends, and public holidays vary by market. Anomaly detection that does not know about a regional holiday will confidently report a crisis every year. Text expansion and direction Translated strings can run considerably longer than English, and right-to-left layouts expose every hard-coded alignment in the codebase. Model behaviour A model trained mostly on one market's data will be quietly worse elsewhere. That gap needs to be measured per region, not assumed away.
We treat localization as a platform capability rather than a per-feature task. Formatting, calendars, and regional model evaluation live in shared services, so a new feature inherits correct behaviour instead of reimplementing it.
"You haven't localized a product until its anomaly detector knows which days are holidays where the user lives."
No spam. Unsubscribe anytime.