> ## Documentation Index
> Fetch the complete documentation index at: https://docs.recepta.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# Your first 30 days

> How to go from a working setup to actually trusting the agent with your phone.

The [quickstart](/quickstart) gets you answering calls. This page covers what comes after — the four weeks between "it works" and "I've stopped worrying about it."

Most businesses that get poor results don't have a broken agent. They set it up, walked away, and never read a transcript.

## Week 1 — Read everything

Your only job this week is finding out what your agent doesn't know.

<Steps>
  <Step title="Start narrow">
    Use [forward-on-no-answer](/call-forwarding/conditional-forwarding) so the agent only picks up calls you were about to lose. Nothing is at risk, and every call it takes is upside.
  </Step>

  <Step title="Read every transcript, daily">
    Not a sample — all of them. Volumes are low enough in week one that this takes fifteen minutes. See [Call logs](/receptionist/call-logs).
  </Step>

  <Step title="Fix the input, not the wording">
    Wrong price → [knowledge base](/receptionist/knowledge-base). Missing detail → [caller types](/receptionist/caller-types). Wrong hours → [business profile](/receptionist/business-profile).
  </Step>

  <Step title="Verify a booking end to end">
    Confirm a real appointment reached the right calendar with the right duration, and that the customer got a confirmation.
  </Step>
</Steps>

<Tip>
  Have whoever used to answer your phone do the reading. They spot a wrong answer instantly, where someone reading fresh will accept it as plausible.
</Tip>

**What good looks like by Friday:** no factual errors in the last ten calls, and every call ending somewhere — booked, transferred, or a message taken.

## Week 2 — Close the gaps

<AccordionGroup>
  <Accordion title="Work through the failures" icon="circle-xmark">
    Filter to calls that ended without a booking, transfer, or message. Each is a lost job with a written explanation attached. Most cluster into two or three missing facts.
  </Accordion>

  <Accordion title="Audit your transfers" icon="phone-arrow-up-right">
    Every escalation is either correct or a knowledge gap. Sort them into those piles. A high transfer rate almost always means a thin [knowledge base](/receptionist/knowledge-base), not a badly written rule.
  </Accordion>

  <Accordion title="Test the emergency path" icon="triangle-exclamation">
    Call in, describe an emergency, and confirm a human actually answers — including the fallback if the first number doesn't pick up. An untested [transfer rule](/receptionist/transfer-rules) is a promise you haven't checked.
  </Accordion>

  <Accordion title="Connect your systems" icon="plug">
    If you haven't already, link your [calendar](/booking/calendars) and [field-service CRM](/integrations/field-service-crm) so bookings land where your team works.
  </Accordion>
</AccordionGroup>

## Week 3 — Widen the aperture

Once a week has passed without a factual error, give the agent more.

| Change                                                 | Why now                                             |
| ------------------------------------------------------ | --------------------------------------------------- |
| Move to forwarding all calls, or all after-hours calls | You've seen it handle your real callers             |
| Shorten the ring time before forwarding                | Callers wait less; the agent has proven itself      |
| Turn on SMS reminders                                  | Assuming [A2P](/phone/a2p-10dlc) is approved by now |
| Enable post-visit feedback                             | Starts building review flow                         |

Drop to reading failed and transferred calls plus a sample of the rest. Start grading against consistent criteria — see [QA & learning](/receptionist/qa-and-learning).

## Week 4 — Make it pay for itself

The receptionist stops calls leaking. The [outbound engine](/outbound/overview) creates new work from what you already have.

<Steps>
  <Step title="Start with unsold quotes">
    The highest-return campaign for most service businesses. These people asked for a price and got distracted. Work the last 30–90 days.
  </Step>

  <Step title="Run ten contacts first">
    Include your own number. Listen to a call and read the texts before releasing the rest.
  </Step>

  <Step title="Measure booked jobs, not answer rates">
    A campaign with a low answer rate that books six jobs beats a chatty one that books none.
  </Step>
</Steps>

<Warning>
  Before your first campaign, read the compliance section in [Outbound overview](/outbound/overview). Calling hours, consent, and do-not-call obligations are yours, and the penalties are real.
</Warning>

## After the first month

A sustainable rhythm:

* **Weekly** — grade a sample of calls, plus every call that ended badly
* **Monthly** — check [usage against your plan](/account/usage-and-overages) and your escalation rate
* **On any change** — new price, new service, new hours: update the knowledge base the same day, then test the scenario

<Note>
  The single best predictor of whether this works for you is whether someone reads transcripts in month two. Everything else is downstream of that.
</Note>