Playbooks
Reply in the prospect's language: the rules that ran in a 21 language programme
The Shelly Group programme wrote to installers in the language of their country, then followed the language each person replied in. The agent wrote in 21 languages, 51% of installers who replied wrote in a language other than English and 92% of the agent's replies matched the installer's message language. It was a listening survey, not a sales campaign.
Published by Sellinger AI. Numbers and settings checked on 19 October 2026 against the fact register. Report an error.
- countries in the language map the first email used
- 60
- languages the agent wrote in
- 21
- of 309 agent replies matched the installer's message language. 98% matched the message or the country language
- 92%
- of installers replied in the June to August waves, against 3.3% for the first email alone
- 13.6%
The scenario
A device maker with installers in more than 100 countries wanted to ask four questions of thousands of them. No team can write in dozens of languages. Read the case study and the survey playbook.
The language rules
- 1
First email
The country's language
Chosen from a language map of 60 countries.
- 2
A reply
Follow the language they wrote in
A person who answers in English to a German email gets English back.
- 3
Greeting
Local grammar
Greetings follow local grammar, such as the vocative in Czech and Polish.
- 4
Pricing or support
A short holding reply in their language
Then a person from the company answers. The agent made no promises on price, dates or certification.
The result
| Measure | Result |
|---|---|
| Languages the agent wrote in | 21 |
| Installers who replied in a language other than English | 51% (178 of 351) |
| Agent replies that matched the installer's message language | 92% (309 replies) |
| Matched either the message or the country language | 98% |
One Brazilian installer answered only once the Portuguese version arrived. 46% of converted installers wrote in a language other than English.
Limits
- The 92% figure counts matching language, not translation quality.
- The audience was an existing installer network that had registered with the company. A cold list may behave differently.
- The programme ran on email. No published study measures language matching on LinkedIn.
- Opt out detection is documented for English, Bulgarian and German among other forms. Elsewhere a weak or unclear phrase stops the agent and a person decides.
Common questions
Does the agent detect the language?
Do I need a native speaker to review?
Sources and method
A playbook here is a configuration that ran, with the outcome it produced and the base of every number. It is not a template promise: your audience and offer decide your own rates. Source: Shelly installer programme case study, Sellinger production data for the Shelly workspace, read only, to 9 October 2026.
Related: multilingual outreach, Shelly case study, four question partner survey.
