Use case

Automate the repetitive majority of support

Most tickets are questions your documentation already answers. Put an agent in front of them, and let your team spend its day on the conversations that actually need a person.

The problem, honestly framed

Support volume grows linearly with customers; support headcount can't. The result is a queue: customers waiting hours for answers that exist word-for-word in your docs, and agents burning their day on copy-paste answers instead of hard problems. Deflection — answering the repetitive majority automatically — is the only lever that scales, and it only works if the automatic answers are actually right.

Two deflection workflows

Product workflows — here is exactly what the software does

1. The pre-ticket answer

A customer opens your widget (or emails your support inbox) asking how to reset an API key. The agent retrieves the passage from your docs, answers with a citation, and the conversation ends — no ticket, no queue, full transcript in your dashboard. The same question at 2am gets the same answer at 2am.

2. The graceful handoff

A customer asks about a refund your policy doesn't cover. The agent doesn't improvise: the "not in the knowledge base" trigger escalates the thread, the AI mutes, your team is notified, and whoever takes it reads the full transcript before typing a word. The customer never repeats themselves, and the human minutes go where they're needed.

How it works

  1. 1
    Upload your documents
    Policies, FAQs, manuals — PDF, Word, text, Markdown, CSV.
  2. 2
    Write a one-paragraph persona and pick a model
    GPT-4o and GPT-4o mini on every plan; Claude Sonnet 4.6 and Haiku 4.5 on Pro.
  3. 3
    Go live where your customers are
    One script tag for the widget; guided flows for Gmail, WhatsApp, SMS, Telegram, Slack, Discord, Messenger and phone.
  4. 4
    Set the escalation rules
    Human-request detection is built in; add triggers for knowledge gaps, frustration, or your own keywords, with business-hours-aware messaging.

Measuring the impact

We won't hand you an invented "80% deflection" banner — your number depends on how much of your support load is genuinely answerable from documents. What the dashboard gives you is the honest arithmetic:

  • Deflection rate: conversations resolved without escalation ÷ total conversations — watch it climb as you fill knowledge gaps.
  • Escalation queue volume: what still reaches humans, with reasons attached — your roadmap for the next document to write.
  • First-response time: AI answers land in seconds on chat and about a minute on email, at any hour; your human queue shrinks to what deserves it.

Frequently Asked Questions

What kinds of questions can it deflect?

Anything your documents already answer: policies, how-tos, pricing details, setup steps, account questions. The agent retrieves the relevant passages and cites them — if the answer isn’t in your docs, it says so or escalates instead of guessing.

What happens to the questions it can’t answer?

They escalate: the AI mutes on the thread, your team is emailed, and the conversation waits in a queue with the transcript, priority and assignment. Nothing is dropped, and nobody repeats themselves to a human.

How do I measure whether it’s working?

Your dashboard shows every conversation and every escalation. The ratio of conversations resolved without escalation is your deflection rate; watch it and the escalation queue’s volume week over week.

Which channels does this cover?

All of them at once: the website widget, Gmail, SMS, WhatsApp, Telegram, Slack, Discord’s /ask, Messenger, and inbound phone calls — one agent, one knowledge base.

Deflect the repetitive. Elevate the rest.

10 minutes from sign-up to your first answered customer message. Free forever plan, no card required to build.