Why "I don't know" is the most important sentence for an AI system
The value of an AI answer is determined by what the system refuses to say, not by what it keeps endlessly asserting.
Read more
Home
No hallucinations
Sarrai doesn't make up answers.
Most AI chatbots are trained to always seem helpful, even when they don't know the answer. So they make something up. That's called a hallucination. That's the kind of chatbot we didn't want to build. When Sarrai can't find the answer in your documentation, it honestly says it doesn't know and brings a colleague into the conversation.
No answers from general knowledge
Sarrai isn't allowed to say anything that isn't in your knowledge base. The underlying AI model cannot use its own knowledge of the world.
"I don't know" is a valid answer
If Sarrai can't find a matching answer, it says so honestly and brings in a human. That person's answer flows back into your knowledge base via the approval inbox.
Every answer comes with a source citation
With every answer, you see where it came from. No black box: you can always check which article Sarrai based its answer on.
A made-up answer is worse than no answer at all
A chatbot that always has an answer ready looks impressive. Until it gets one wrong: a made-up answer sends your customer down the wrong path and eats away at the trust in your company.
That's why only one measure counts for us: how many customers were genuinely helped by Sarrai's answer. How many questions it answers comes second. In practice, that sounds like this: "I can't find this in the documentation, so I won't guess. I've passed your question on to a colleague, who will pick it up right here."
The customer knows exactly where they stand. And as soon as your colleague answers, that answer flows back into your knowledge base. So next time, Sarrai will know it, without you putting any time into it.
Try it out for yourself
Curious how Sarrai answers your customers' questions? Try it free for 30 days, or book a demo.
