The full conversation history, your campaign playbook, and compliance context — assembled fresh on every turn. The AI decides what to say, when to book, and when to stop.
On every turn the AI returns a structured JSON response: a message to send, a directive that determines what happens next, a reasoning field for the audit trail, and a confidence score. The runtime acts on the directive — the AI doesn't call APIs or update records directly.
Base guardrails sit above every campaign configuration. They enforce behaviour that must be consistent regardless of how a campaign is set up.
The AI uses only information present in the campaign fields when answering objections or questions. If no usable information is present, it deflects to a specialist — it never fabricates an answer.
How it works: Fields used in order: product_or_service, benefits, background, objection_handling, cross_up_sell.
When a contact asks directly whether they are talking to a human or a bot, the AI identifies itself as an AI assistant and offers to connect them with a specialist. It never denies being an AI.
How it works: Enforced via the base guardrail prompt — applies to every campaign without configuration.
UK English spelling, GSM-7 characters only, no markdown, no emojis, no exclamation marks. Stock filler phrases like 'Great!' and 'Thanks for sharing' are explicitly prohibited.
How it works: Acknowledgement is reserved for genuinely warranted moments — not used as a conversation opener.
All messages are capped at 320 characters — two SMS segments. This is enforced in code after the AI responds. Messages exceeding the limit are truncated at the last complete sentence.
How it works: Runtime truncation is a code guarantee — it applies regardless of what the AI returns.
The Campaign Tester lets you simulate any conversation against your live config before a single contact is messaged.