polly is a governed ai assistant that works over whatsapp. she answers your customers in seconds, books appointments on a real calendar, replies from your own documents, and hands anything sensitive to a person. what separates her from an ordinary chatbot is the layer underneath: llmcontract, our own governance engine. what polly is allowed to do is enforced in code, on every action, not merely suggested in a prompt.

• lives on your whatsapp number, or ours, with nothing for your customers to install
• every tool call is checked before it runs, so she cannot break a rule even if a customer talks her into trying
• anything she should not do alone pauses and waits for a human in one shared inbox

polly holds a real conversation and turns it into governed action. she qualifies an enquiry, answers the recurring questions, checks availability against a live calendar, and books the meeting with an invite, all inside the chat. within the rules you set she works on her own; outside them, she is not allowed to move.

rules that are enforced, not requested.

her limits are checked in code before every single action. she cannot quote a price she does not have, promise a refund above your cap, or skip a step your process requires, because the platform will not execute it. it is not a matter of asking the model nicely.

a human, one tap away.

anything sensitive is parked for review and lands in your inbox. approve it, do it yourself and type back the result, or take the conversation over. the customer just sees things getting handled, and your team stops doing intake and starts making decisions.

everything on the record.

every message, tool call, gate decision and human action is logged to an exportable audit trail. useful for you, and built for the transparency and oversight the eu ai act now expects.

polly is a managed product, not a self-serve signup. we set her up on a whatsapp number with your tone, your rules and your knowledge, and she is live in days. each deployment is configured for one business, and the governance and audit trail are the same whether she is booking calls for us or running claims intake for an insurer.

the engine underneath, llmcontract, is available on its own for teams that need governed ai agents inside their own products. we published the core as a library, so the enforcement layer is inspectable rather than a black box.

if your customers already message you, polly fits. the conversation starts the way every engagement does: a call, a look at what you need, and a proposal with a scope and a price. or just message her yourself and watch her book it.