Every AI action leaves a paper trail.
Coming soon: the IBIOS AI receptionist is not live yet. When it launches, every decision it makes will be recorded in a dedicated, append-only audit log. What was said, what action was taken, and what the outcome was — all traceable, all tamper-evident.
Comprehensive interaction log. Append-only design. Anchor-hash tamper evidence architecture.
Why does AI in a customer-facing role need an audit trail?
A customer calls and later claims the AI told them their bathroom installation would be completed within two weeks. Your team have no record of that being said. Without an immutable AI audit trail, it is your word against theirs.
AI receptionists make decisions on behalf of your business. Those decisions need to be traceable. Not just for disputes — but for quality review, training, compliance, and understanding how the AI is actually performing.
When it launches, IBIOS will record every AI decision in an append-only log. You will be able to retrieve the exact transcript, the emotional state detected, and the action taken for any call, at any time.
Coming soon: three layers of AI accountability
Designed from the ground up for auditability, not retrofitted as an afterthought.
Dedicated AI interaction log
Coming soon: the AI interaction log is a separate log from the main audit log, designed specifically for AI receptionist actions. Built for high-volume append workloads and fast search by session, caller, and outcome.
Append-only with tamper-evidence design
Records are never updated or deleted through normal application paths. The architecture supports anchor hashing — each record references the hash of its predecessor, making retroactive modification of the chain detectable without inspecting every record.
Cross-modality coverage
Coming soon: every AI receptionist modality will be covered: STT transcription, AI response generation, emotion classification, tool calls, TTS synthesis. A complete picture of every decision the AI made, at every step of every call.
What a single call record will capture (coming soon)
Every turn in an AI receptionist conversation will be logged as a separate record in the AI interaction log. A call with several conversational turns will create one record per turn, each one independently verifiable.
Single turn record — example fields
session_id: 9f3a8c21… | turn: 3 | caller_id: +44 7712 … | stt_transcript: “Do you have any appointments next Tuesday?” | emotional_state: neutral | tool_call: calendar.check_availability(2026-05-12) | ai_response: “Yes, we have slots at 10am and 2pm on Tuesday — which would suit you better?” | outcome: appointment_offered | latency_ms: …
Dispute resolution — how the trail is used
Customer claims the AI confirmed a Tuesday 10am appointment. Audit trail shows the AI offered the slot but the session ended before a confirmation turn was recorded. No appointment was logged in the calendar. The trail resolves the dispute from the record rather than from memory.
Who needs an AI audit trail?
Any business deploying AI in a role where decisions have real-world consequences.
Businesses with compliance obligations
Regulated industries require records of customer interactions. An AI audit trail with tamper-evident design supports compliance obligations without relying solely on call recordings.
Businesses that have had AI errors
When a customer claims the AI told them something they should not have been told, an immutable audit trail lets you trace exactly what was said, when, and with what confidence level.
Multi-operator deployments
When IBIOS serves multiple tenants, the AI audit trail provides per-tenant accountability. Each tenant’s AI interactions are recorded separately and accessible only to authorised users for that tenant.
Any business using AI in customer-facing roles
Deploying AI in a customer-facing role creates accountability questions. An immutable audit trail means you can always answer: what did the AI say, why did it say it, and what happened as a result.
Technical specifications (planned)
AI interaction log
Coming soon: dedicated append-only log for AI receptionist actions. Separate from the main staff audit log.
Fields recorded
Session ID, caller ID, timestamp, turn number, STT transcript, AI response, emotional state, tool calls, outcome, per-step latency.
Write mode
Append-only. No update or delete paths in normal application code. Deletion requires direct DB access and is logged.
Tamper evidence
Architecture supports anchor hashing. Each record can include a hash of the previous record. Tamper-evidence hashing is on the IBIOS security roadmap.
Access control
Audit viewing permission required for full log access. Call summary visible to any staff member with access to the relevant job or customer.
Export
CSV and JSON export from admin panel. Filtered by date, caller, session, or outcome. Export actions are themselves audit-logged.
Questions about the AI audit trail
What does the AI audit trail record?
When it launches, the AI interaction log will record every AI receptionist action: the call session ID, the caller identifier, the timestamp, the conversational turn, the STT transcript, the AI-generated response, the detected emotional state, any tool calls made (calendar, CRM, etc.), the outcome (appointment booked, callback created, query resolved), and the latency for each step.
How is the audit trail protected from tampering?
The AI interaction log is designed to be append-only. Records are never updated or deleted through normal application paths. The architecture supports anchor hashing — each record includes a hash of the previous record, creating a chain that makes retroactive modification detectable. Tamper-evidence via hashing is on the IBIOS security roadmap.
Who can access the AI audit trail?
Access to the AI interaction log is controlled by the audit viewing permission in the IBIOS role system. Only users with that permission can view raw audit log entries. The call transcript summary (without full audit detail) is visible to any staff member with access to the relevant customer or job record.
How long are audit records kept?
IBIOS does not automatically purge audit records. Retention is configurable per tenant and subject to your data retention policy. For most businesses in the UK, GDPR considerations apply to personal data in call transcripts — IBIOS supports configurable retention periods with anonymisation on expiry rather than hard deletion.
Can I export the AI audit trail?
When it launches, yes. Audit records will be exportable from the IBIOS admin panel in CSV or JSON format, filtered by date range, caller, session, or outcome. Export is restricted to users with the audit viewing permission and is itself logged in the main audit trail.
How does the AI audit trail differ from the regular audit log?
The main IBIOS audit log records human staff actions: who created a customer record, who approved a quote, who updated a supplier order. The planned AI interaction log (coming soon, with the AI receptionist) will record AI receptionist actions: what the AI said, what it decided, what tool calls it made. They are separate logs with separate schemas, designed for different purposes, but both are searchable from the IBIOS admin panel.
Which IBIOS product line includes the AI audit trail?
The AI audit trail is coming soon as part of the IBIOS AI Receptionist, which is not live yet. It is built for showroom retailers and UK tradespeople. See also: who changed what: an AI action audit.
IBIOS (Integrated Business Intelligence & Operations System) is made by IBIOS Limited (NI743167), Northern Ireland. Access is by waiting list; there is no free trial.
Coming soon: every AI receptionist decision traceable.
Append-only audit log. Comprehensive interaction record. Tamper-evidence architecture. Early access by waiting list.