Receptionist Content Best Practices
Last verified
How to write the content your Orca AI receptionist speaks and works from. Good content is the difference between a receptionist that sounds like your business and one that fumbles basic questions.
This guide covers product behaviour and writing technique only. Legal obligations - including your responsibility to tell callers calls are answered by AI and may be recorded and transcribed - are set out in the Terms of Service and Privacy Policy, which are the canonical sources if anything here appears to differ.
Where content lives
In the AI config screen (web or mobile app), owner and admin roles can edit:
- Business name - required to publish. Spoken in the default greeting and used across your dashboard.
- Assistant name - optional. What the AI calls itself if a caller asks.
- Tone - friendly, professional or casual.
- Greeting - required to publish. The first thing every caller hears.
- After-hours message - collected in config, but note the limitation below.
- Business timezone - Brisbane or Sydney. Used for booking appointments and speaking times to callers.
- Knowledge lists - Services, FAQs, Policies, Do not say.
- Tools - what the AI is allowed to do (take a message, transfer, book appointments, email summaries).
Changes only reach your live receptionist after you Save draft and then publish. Until you publish, callers still hear the previous published version.
Writing the greeting
The greeting is forced as the AI's first utterance on every call. Write it the way you would answer the phone yourself.
Good example:
"Thanks for calling River City Plumbing, this is our after-hours answering service. How can I help you today?"
Weak example:
"Hello." (too short - gives the AI nothing to work with)
Tips:
- Say the business name clearly - callers should know they reached the right place.
- If you want callers told their call is answered by AI and may be recorded or transcribed (your obligation under the Terms), the greeting is the reliable place to say it. Orca does not add any disclosure line automatically.
- Keep it to one or two sentences. A long greeting gets in the way of the caller's reason for calling.
- Avoid jargon, abbreviations and in-jokes that a first-time caller would not understand.
Writing knowledge lists
Services
List each service as a short, plain phrase: what it is and any qualifier a caller would ask about (service area, typical response time, whether a call-out fee applies). The AI uses this list to answer "do you do X?" and to steer booking questions.
Policies
One policy per line, phrased as a rule the AI can apply: "We charge a call-out fee within 20 km of the CBD", "We don't do gas fitting", "Payment is due on completion by card or bank transfer". Vague policies produce vague answers.
Do not say
Phrases or commitments you never want spoken to a caller - for example competitor names, discount codes, or promises about pricing you have not approved. The list is injected as a "never say" instruction. Keep entries short and unambiguous.
FAQs
The FAQ editor takes question-and-answer pairs, and FAQ entries are read to callers as question-and-answer pairs in the current release. Add your top questions, publish, then run a test call to confirm the spoken wording before relying on FAQ answers on live calls.
Tone and assistant name
- Pick one tone and stay consistent with it across greeting and knowledge content.
- Give the assistant a name only if you want callers to be able to address it; "the receptionist" is fine too.
- The AI is instructed to be concise. If answers come back too brief for your liking, add detail to the knowledge lists rather than fighting the conciseness in the greeting.
What the AI can and cannot use
- The AI answers from the published config only. Saving a draft changes nothing for callers.
- The AI does not browse your website or look anything up online. If a fact is not in your knowledge lists, the AI either asks the caller or admits it does not know - it cannot check.
- Booking uses the timezone you set. If your business operates outside Brisbane or Sydney time, pick the closest option and state your actual hours in a policy so the AI speaks them correctly.
- Business-hours awareness is limited: the AI may not reliably know whether you are open or closed at a given moment, and the after-hours message field is currently not spoken by the AI. If after-hours handling matters, put the key information in your greeting or a policy (for example "If this is an emergency, call ...") rather than relying on the after-hours message field.
- There is no opt-out keyword that stops the AI mid-call, and unknown questions have no automatic escalation. Write knowledge that covers your common caller questions.
Test before and after every publish
- Use Test call (web or app) to place a synthetic call to your own number.
- Ask the questions your real callers ask: pricing, service area, hours, booking, "can I speak to the owner".
- Check the call record afterwards - status, escalation action, transcript - to confirm the AI answered the way you intended.
- If the test result shows a stale config version, publish again and retest.
Quick checklist before publishing
- Business name and greeting are set and read naturally.
- Greeting includes any AI or recording disclosure you want callers to hear.
- Timezone is correct for your booking area.
- Services and Policies cover your top ten caller questions.
- Nothing in the content promises outcomes the AI cannot control (guaranteed prices, arrival times, outcomes).
- Test call completed and the transcript checked.
Related guides
Need a hand?
If anything here does not match what you see in the app, contact us and we will sort it out.