Support answers should come from your own words.

Most support chatbots fail the same way. They are built on a general model that has never read your documentation, so when a customer asks something specific — your refund window, whether a feature exists on their plan — the model produces the statistically likely answer rather than the true one. It sounds confident. It is wrong. And the business only finds out when a customer acts on it.

What we build

Ontroz is an AI chatbot that trains on your content — your website, your help docs, your PDFs — and answers only from that. When it does not have the answer, it says so and points the visitor somewhere useful instead of inventing something. That constraint is the product. It is what makes the answers trustworthy enough to put in front of customers unsupervised.

The rest follows from making that practical: training from a URL, a sitemap, or a folder of documents; a widget that drops into any site with two lines of code; lead capture; and a record of every conversation so you can see what people actually ask. You can read more about what it does or how retrieval-augmented generation works.

Who it's for

Solo founders, small SaaS teams, and agencies — people who need to reduce support load without hiring, and who want predictable pricing rather than a credit system that gets expensive without warning. Pricing is published in full on the pricing page; there is no “contact us for a quote” tier below Agency.

Who we are

We're the Ontroz team. We built this because we kept seeing the same failure: businesses putting a chatbot on their site to help people, and the chatbot confidently telling those people things that weren't true. A wrong refund window. A feature that doesn't exist on that plan. Support that makes a customer less informed than before they asked is worse than no support at all — and it costs the business the trust it was trying to build.

So the product started from a constraint rather than a feature list: answer from the customer's own content, or don't answer at all. Everything else — the training pipeline, the retrieval layer, the fallback message you write yourself — exists to hold that line. We'd rather the bot say “I don't have that” and hand off to a human than guess well enough to be believed.

The broader goal is reach. Most businesses can't staff support around the clock, and most people asking a question at 11pm aren't going to wait until morning — they leave. If the answer already exists in someone's documentation, it should be available to anyone who needs it, in their own words, whenever they ask. That's the part we find worth building: taking knowledge that's already written down and making it actually reachable, for as many people as possible.

We're a small team and we read what comes in. If something is wrong, unclear, or missing, tell us — we'd rather hear it.

Get in touch

Questions about the product, security, or whether it fits your case — contact us. How we handle your data is documented on the security page.