Company
We started Huematte after watching a chatbot give a polite non-answer to a one-message problem.
Hue for the domain, because each vertical deserves its own understanding. Matte for the finish, because support should feel calm and dependable, not shiny.
The customer wrote one line: they wanted to change their order to a different size. Any experienced agent at that store would have checked the delivery date, confirmed the item was inside the return window, and started an exchange. It would have taken one reply.
The chatbot apologised, asked what they meant by “change”, offered a link to the returns page, and then created a ticket. The customer waited two days for a person to do what the first reply should have done.
The model behind it was fluent. What it lacked was the domain. It didn't know that in e-commerce, “change this” usually means an exchange; that in SaaS billing it usually means a plan change with proration; that at a clinic it means a new slot. And it had no access to the store's own rules, so even a good guess would have been a guess.
Huematte builds support agents the other way round. Each one is pre-trained on the patterns of a single domain, then grounded in a business's own policies, then bounded by limits the business sets. The goal is simple: the first reply should be the one an expert would have sent.
Principles
Three rules we build by.
Know the domain before answering.
A support agent is only useful if it understands what a customer means in that industry. We start from the domain, not from a general assistant with a longer prompt.
Grounded in your policies, never guessing.
Every answer should trace back to a rule the business wrote. When no rule covers a question, the right move is to say so, not to sound confident.
Hand off early when a person should decide.
Automation earns trust by knowing its limits. A clean handoff with full context is a good outcome, not a failure.
Where to find us
Our registered office details are on the contact page.
Which question fills your inbox?
Pick the one that sounds most like your customers. We'll set you up with that agent first.