This describes work that is in design. Everything below is what we are building, not something you can switch on today.
A firm in Montreal receives an inquiry in French and answers it in English, because the tool only speaks English.
A firm in Houston receives one in Spanish and answers it in English, for exactly the same reason.
Neither firm meant anything by it. In both cases somebody on staff could have answered in the language the message arrived in — that is not the gap. The gap is that by the time the message reached a person who could, the reply had already gone, or the day had.
And in both cases the applicant reads the same thing into it, which is: this is not for me.
The cost is paid at the first reply
This matters more than it sounds, because of where in the funnel it lands.
The first reply is the point at which somebody decides whether to keep going with you. An applicant who writes in French and gets English back has not been insulted — they have been told, in the most ordinary way possible, that dealing with your firm will be a small effort every time. People with three other tabs open do not accept small efforts.
Both firms above are in markets where a meaningful share of qualified applicants writes in something other than English. Not a marginal share, not a courtesy — people who would rent the unit and can pay for it. Every one of those who quietly moves on is a lead the firm paid to generate and lost at the cheapest possible moment to keep.
We are a Montreal company, and this is not an abstract topic here. Somebody writing to us in French and getting an English answer back is not a rounding error.
What we are building
English, French and Spanish for the applicant-facing flow, each of them in the firm's voice and under the firm's rules — which is the part that makes this harder than it looks, and the reason it is not a toggle.
The decision that shapes everything: no translation at send time
There is an obvious way to build multilingual support. Compose the reply in English, run it through a translation service on the way out, send the result. It is fast, it is cheap, it is what most software does, and we are not doing it.
Here is why, in one sentence: a policy line is a statement made on the landlord's behalf, so a machine-translated policy line is an unapproved policy in that language.
Read that carefully, because it is not a quality argument. Machine translation is very good now. The problem is not that the French might be clumsy. The problem is that a sentence describing your deposit terms, produced at send time by a service, and sent under your firm's name, is a commitment your firm has made in a language nobody at your firm ever read.
If it is subtly wrong — a shade of obligation the English did not carry, a term of art that does not mean what its dictionary equivalent means — you have made that commitment anyway. And you will find out about it from the person holding the reply, not from a review process, because there was no review process. Send-time translation is by definition a step that happens after the last human look.
So the design goes the other way. Each firm's pack carries its policy and tone sections in every language that firm enables, written or reviewed and approved by the firm, in that language. The French policy line is a French policy line your firm signed off, not an English one wearing a translation. When the software answers in French, it is quoting something a person at your firm approved in French.
Tone comes along for the same ride. How your office writes in English and how it writes in French are not the same voice, and a translated tone is a caricature of the original — slightly too formal or slightly too casual in ways native readers register immediately without being able to say why.
The cost, and it is the real one
Every language multiplies the acceptance tests.
The way we are building this, each firm gets its own set of tests — its own rules, answered back to it, checked before anything is switched on. Add a language and that whole suite runs again in that language, because a rule that is correctly applied in English tells you nothing about whether it is correctly applied in Spanish.
That is the actual expense here. Not the composition — models handle multiple languages well. The testing surface, multiplied by the number of languages, per firm.
We think that is the right cost, and it is worth saying why rather than just absorbing it: the alternative saves us engineering time and spends your credibility instead. A firm that has made a commitment in a language it cannot read has been sold a feature at its own expense.
What adding a language will actually involve
So this needs saying plainly, because it is not a toggle and you should not discover that afterwards.
Enabling a second language will mean going through your own rules and your own tone in that language and approving them. If you have staff who work in that language it is a morning's work for one of them. If you do not, it is a real question about whether you should be answering in that language at all — and that is a better question to hit during setup than during a tenancy.
Once it is done it stays done, until you change a rule, at which point the change needs approving in each language you have enabled. Same as the English one, with the same reason behind it.
For a firm that is genuinely operating in two languages, the arithmetic is easy — that work is already being done by hand, message by message, by whoever on the team speaks the language, and it is being done at the slowest possible moment. For a firm that gets three French inquiries a year, honestly, do not enable it. Route them to a person.
What we are not saying
Three things, stated deliberately.
We are not claiming language support ships today, or that the three languages arrive together. They are planned, they are not built, and if they land in stages we will say which stage.
We are not claiming parity across them. English is where the work starts. A firm enabling French or Spanish should expect the same rules and the same voice, and should not expect three identical maturity levels on day one.
And this is a market decision, not a legal one. We are building it because a Montreal firm and a Houston firm both lose good applicants at the first reply, and because answering somebody in the language they wrote to you in is the least you can do when they are asking about somewhere to live.