Equal Ear
A hearing test showing which customers your organisation cannot hear, and what it costs.
The job
Proves that bots, voice agents and human queues serve every accent, language and ability group equally well, prices the gap where they don't, and closes it with fixes whose targets are set in advance and checked.
The moment
Customer groups are plotted like a hearing test, showing where the organisation's bots and agents struggle to understand people.
The chart dips sharply for non-English speakers with billing questions. The voice agent hands them to a human 3.1 times as often, somewhere between 2.7 and 3.5 times, costing about $41,000 a month in agent time.
The bill also shows 11% of sampled non-English conversations switch to English by the fourth exchange: customers are translating themselves for the bot.
Four weeks after fixes to routing and vocabulary, the dip has clearly filled in. The committee chair plays the audio version, where the missing tone can now be heard.
What it does
- Open the verbatims behind a dip or bill line
- Pre-register a gap target
- Route the worst language x intent combination to a language-matched human queue, with an expiry
- Add misheard words as KB synonyms
- Send the monthly fairness exhibit as a PDF and a synthetic tone-sweep voice note
What you see
Audiogram (hearing-test chart) of customer groups, with an itemised tax bill per group and an audible monthly exhibit
What it moves
The gap in handover rates between the best-served and worst-served language groups, in percentage points, and the monthly dollar cost of the extra agent time this causes.
Built for
- Compliance, complaints & legal
- Frontline staff
- QA, training & knowledge
- Finance
