Fast and wrong is expensive.
Not expensive like a wasted afternoon. Expensive, like it can scuttle a deal.
What does a small mistake actually cost in real estate?
In most industries, a software error costs you the time it takes to catch and redo it. Annoying, recoverable, forgotten by Friday.
A real estate transaction doesn't price mistakes that way. A missing detail, an unsigned page, a deadline that slid, can cost the deal you've been working for months. The listing appointments, the showings, the negotiation, the inspection back-and-forth, the client's trust: all of it can hinge on one detail being right at the moment it matters.
That's the asymmetry that should shape how we build and judge AI for this industry. The error takes seconds. The cost is measured in months.
Why isn't "almost right" good enough?
Because "almost" is where the risk lives.
An obviously broken tool is harmless; nobody relies on it. The dangerous tool is the one that's right often enough to be trusted and wrong at the wrong moment. I've argued the order of operations before, in Fast and Wrong Is Still Wrong: correctness comes before speed. This is the reason why. The downside of wrong isn't symmetrical with the upside of fast. Saving an hour is nice. Losing a deal is a story the client tells for years.
Does that mean speed doesn't matter?
Speed matters. AI can be fast, and that's genuinely great; the hours agents spend on transaction paperwork are real, and getting them back changes the job.
But speed is a bonus on top of right, never a trade against it. Nobody remembers that the paperwork was fast. Everybody remembers that the deal closed, or didn't.
How should this change the way agents evaluate AI?
Ask about the failure case, not the demo. When this tool is wrong, how would I know? Does it flag what it's unsure about, or does it present everything with the same confidence? Is it consistent enough to be trusted on the hundredth deal, not just the first?
When it matters, it has to get it right. That's the bar. Everything else is a footnote.
The full Ethica blog library lives at heyethica.com/blog.