A customer in Abidjan asks their AI assistant, in French, to book a meeting room for Thursday. Your booking form is in English. Your dates are written day first. Your prices are in another currency.
Somebody has to translate all of that, and right now it is the assistant, guessing.
Language is the easy part
AI assistants are generally good at understanding a request in one language and working on a page in another. That is not where most of the trouble comes from.
The trouble is in the small things that look the same across languages but are not:
- Dates. Is 04/05 the fourth of May or the fifth of April?
- Money. Does “1.500” mean one and a half, or fifteen hundred?
- Places. The same city can have more than one spelling, and the same suburb name can exist in two countries.
- Names of services. Your English menu says “consultation”. The customer’s language may have three words for it, each meaning something slightly different.
A person on the phone sorts these out with one quick question. An assistant filling in a form often does not ask.
How WebMCP tools help
With WebMCP, your website hands the assistant a list of actions with clearly described inputs. That gives you a place to remove the ambiguity once, instead of hoping each assistant gets it right:
- Ask for dates in one fixed format and say so in the field description.
- Say which currency a price is in, every time.
- Accept a list of known branch or city names rather than free text.
- Describe each service in a sentence, not just a label, so the assistant can match it to what the customer actually asked for.
None of this needs a separate website per language. It needs tools whose descriptions leave nothing to interpret.
Translate the descriptions too?
Sometimes. If most of your customers in one market use one language, offering tool descriptions in that language can help an assistant match requests more reliably. Where you sell in several languages, keep one careful set of descriptions as the master and treat the others as proper translations, checked by someone who speaks the language. Machine-translated tool descriptions are a fine way to book the wrong service in four languages at once.
Being found is a separate job
WebMCP helps assistants act on your website once they are there. Whether they find you in the first place, and describe you correctly in each market, is a different piece of work. Ankor Business Solutions covers that side in its guide to AI discoverability for small businesses.
The next step
List the languages your customers actually write to you in, not the ones on your website’s language switcher. Start there. If you would like to see where your current forms leave room for misunderstanding, read which AI connection to build first and then ask us for a free snapshot.
Free worksheet: Your first three WebMCP tools (PDF). Pick the three jobs on your website to make ready for AI agents first. No email needed.
Leave a Reply