# Ne oldu?

A recent BMJ (Clinical Research Edition) article draws attention to a practical problem in primary‑care clinics that have introduced AI‑driven reception desks. The report describes how the voice‑activated front‑desk system frequently fails to recognise patients who speak with regional or non‑native accents, causing the system to ask for repetition or to provide inappropriate instructions. This short‑circuit in communication is presented as a patient‑safety risk because it can delay triage, medication pick‑up or appointment booking.

# Şu anda bilinenler

The article explains that most speech‑recognition models used in these AI receptionists are trained on corpora dominated by standard British or American English. Consequently, phonetic variations common in many UK regions, as well as accents of migrant communities, are under‑represented. In practice, clinicians have reported instances where patients were asked to repeat their name or symptoms several times, and where the system mis‑interpreted a medication name, prompting a manual correction by staff. The BMJ piece does not supply quantitative data on how often these errors occur, but it flags the issue as an emerging bias that could undermine equitable access to care.

# Sırada ne var?

The authors call for immediate action on two fronts. First, developers are urged to expand training datasets to include a broader spectrum of English accents, and to validate performance across demographic groups before deployment. Second, health‑service managers are advised to retain human oversight at the reception desk until the technology demonstrably meets inclusive standards. Ongoing monitoring, transparent reporting of error rates, and collaboration with linguists are suggested as next steps to ensure that AI tools support, rather than hinder, patient communication.