Hotel reception call transcription: five ways to route the audio into CodeB.
Every hotel telephony estate is different. Some run a small hosted PBX, some a FRITZ!Box, some a full IP-PBX and a directly-provisioned SIP trunk from a carrier. This page enumerates every practical way to feed reception audio into CodeB for transcription and post-call analytics, and compares them on latency, complexity, portability of the reception number and monthly PSTN cost.
Pick the topology that changes the least about the existing setup. All five deliver the same searchable transcript, the same weekly “why do guests call the reception” roll-up, and the same GDPR-compliant archive — they differ only in how the call reaches CodeB.
What the reception team gets out of it.
Transcription is not the deliverable — it is the input to four operational wins. Position the technology by the outcome, not the codec.
Perfect recall after the fact.
The receptionist takes a booking on the phone, then forgets whether the guest asked for a cot, an early check-in, or both. The transcript is on the shift dashboard within seconds; the answer is one search away.
Weekly “why do guests call” roll-up.
Aggregate every reception call over a week, extract the intent (bookings, complaints, transport, laundry, restaurant, wake-up, spa). Discover the top three question categories and staff or automate accordingly.
Complaint retrieval and dispute defence.
Guest disputes a promise made on the phone. The signed archive lets the duty manager play back the exact 45-second window in front of the guest or the OTA.
Training and onboarding.
Anonymised transcripts become a call-handling playbook for new hires — the actual language guests use, not a fictional script.
Post-processing runs offline on your own IIS server. Neither the audio nor the transcript leaves the tenant boundary unless you explicitly wire a webhook to a downstream analytics tool.
Five topologies, ranked by disruption to the existing setup.
Each option is described with an ASCII call-flow diagram, then rated on the four dimensions the operator actually cares about: added latency, ops complexity, whether the reception number changes, and monthly PSTN cost delta.
Option 1 Full replacement — CodeB SBC in front of the SIP trunk.
Guest phone → PSTN → SIP-trunk → CodeB SBC → reception SIP phone
↓
transcript archive
- Best fidelity — single hop, no double transcode, cleanest audio.
- Also enables the sovereign SBC features (ACL, fraud-guard, per-trunk recording, browser softphones, Voice AI overflow) with no additional plumbing.
- Not always possible — hotel chains under a group contract or countries that require in-carrier number ownership cannot port cleanly.
- Retires the old PBX — a real project, not a weekend swap.
Option 2 Parallel DID — new number rings the reception phone via CodeB.
Guest phone → PSTN → SIP-trunk → CodeB SBC (new DID)
↓
reception SIP phone (via existing PBX)
↓
transcript archive
- Zero risk to the existing reception number and PBX. Roll back by unpublishing the new number.
- Great for pilots — publish the new number on the website / OTAs, keep the old one for legacy channels.
- Only calls to the new DID are transcribed. Calls that still ring the old number are invisible.
- One RTP hop of added latency — imperceptible in-region, noticeable trans-Atlantic.
Option 3 PBX call-forward loop — old number stays, PBX bounces to CodeB.
Guest phone → PSTN → SIP-trunk → hotel PBX → forward-to → CodeB SBC DID
↓
dial-back via SBC trunk → hotel PBX → reception phone
↓
transcript archive
- Number stays. Reception dials the same handset. Business cards, signage, OTA listings unchanged.
- Reversible in one PBX-config change — delete the forward and you're back to the pre-CodeB state.
- Two PSTN legs per call = double per-minute cost. Material for a busy front desk.
- Two extra media hops — the highest latency of any option here. Detectable on international calls.
- If either PSTN carrier misbehaves, the whole reception line goes down.
Option 4 SBC as SIP extension — CodeB registers to your existing PBX.
Guest phone → PSTN → SIP-trunk → hotel PBX → hunt-group [reception, CodeB SBC]
↓
CodeB answers only for archive/transcription;
reception phone rings and picks up as usual.
↓
transcript archive
- The recommended path when the hotel PBX or SIP provider can hand out one extra SIP account. Paste user, password and host into CodeB and it registers like any other softphone.
- Two behavioural modes — hunt-group listener (ring in parallel, answer late so the human picks up first, keep media for the transcript) or fork-and-record (accept immediately, join as a silent participant).
- No PSTN cost. No number change. No PBX-config surgery.
- Requires that the PBX / provider actually issues additional SIP accounts. Consumer boxes and some hosted PBX plans cap the extension count.
- Hunt-group semantics vary between vendors — a quick call with your PBX vendor confirms whether “late-answer” is respected.
Option 5 On-no-answer / after-hours — CodeB catches only overflow.
Guest phone → PSTN → SIP-trunk → hotel PBX → reception phone (rings N seconds)
↓ (no-answer / busy / after-hours)
forward-to CodeB SBC DID
↓
Voice AI answers, takes message, transcript archive
- Answered calls are unchanged. The transcript is only on the calls the desk would have lost anyway.
- Ideal for small properties where the desk is unstaffed evenings and weekends — the AI receptionist takes the message, emails the front-desk mailbox and files the transcript.
- Only overflow is transcribed. Answered-in-office calls stay unarchived unless a second topology is layered on top.
- Guest disputes about answered calls can't be resolved from the archive.
Which one fits your estate?
A one-glance matrix. Bold and green mark the recommended axis; amber marks a caveat; salmon marks a hard limitation.
| Topology | Number | Latency | Ops | PSTN cost | Coverage | Failure mode |
|---|---|---|---|---|---|---|
| 1. Full replace | Port to CodeB | 0 ms | High | Same | 100% of calls | Trunk down = line down |
| 2. Parallel DID | Unchanged | +20-60 ms | Low | +1 DID/mo | Only new-DID calls | New DID down = old still works |
| 3. CFU loop | Unchanged | +40-120 ms | Low | 2× per-minute | 100% of calls | Either trunk down = line down |
| 4. SIP extension | Unchanged | 0-10 ms | Very low | None | 100% of calls | SBC down = missing transcript only |
| 5. Overflow | Unchanged | 0 ms | Low | +1 DID/mo | Only unanswered | SBC down = calls die on overflow |
Consent, retention and who owns the recording.
In the EU (GDPR + PECR / TKG member-state overlays) and in most other jurisdictions, the caller must be informed before recording begins. CodeB ships a multilingual pre-answer notice that can be enabled per DID. Retention is per-tenant and per-purpose — a “seven-day rolling” policy for a low-value support line looks different from “seven-year archival” for a chain compliance officer. Recordings and transcripts live on your own IIS server; the hotel is the data controller. See the AI privacy page for the local-model options that keep the transcription itself on-premise.
Not sure which option fits your setup?
Tell us how the reception line reaches your carrier today, and which PBX or hosted provider you use. We'll come back with a one-page recommendation and a call-flow diagram.