Executive takeaway
AI should improve a customer journey—not conceal a broken process behind a faster response.
What this audit is designed to reveal
Customer communication rarely fails because one message is badly written. It fails across the spaces between acquisition, sales, service, operations and technology: a lead is not assigned, context does not follow the customer, a follow-up depends on memory, or a bot cannot recognise when the conversation has changed.
The audit examines five connected dimensions: entry, response, context, resolution and management. Together they show whether the organisation has a communication system—or only a collection of channels.
The 25-point scorecard
Score every question using evidence from one priority customer journey, one operating team and up to three channels. Avoid answering from policy documents alone; compare the designed process with what actually happens.
Demand and entry
Can the organisation see where conversations begin and what customers intend?
- Source visibility: Can each enquiry be traced to its campaign, channel, property, product or service?
- Consent and preference: Is the permitted purpose and preferred contact channel recorded?
- Intent capture: Is the customer’s immediate objective identified early in the journey?
- Duplicate control: Are repeat contacts recognised as the same person or case?
- Ownership: Does every new conversation receive a clear team and individual owner?
Response and qualification
Does the first response move the customer closer to a useful outcome?
- First-response time: Is it measured by channel, hour and enquiry type?
- Response relevance: Does the reply address the actual question rather than send a generic acknowledgement?
- Qualification consistency: Are essential questions asked without unnecessary repetition?
- Knowledge accuracy: Are pricing, availability, policy and service answers based on an approved source?
- Next-best action: Does the response create a clear next step for the customer and the team?
Context and continuity
Can the conversation continue without making the customer start again?
- Customer history: Can authorised teams see relevant previous interactions?
- Cross-channel continuity: Does context survive when the customer moves from web to WhatsApp, voice or email?
- Conversation summary: Is the latest status visible without reading an entire thread?
- Knowledge consistency: Do different channels and team members provide the same approved answer?
- Follow-up memory: Are promised actions recorded, scheduled and visible to the next owner?
Handover and resolution
Does the system know when, where and how to transfer responsibility?
- Human handover: Can automation transfer the conversation with context intact?
- Routing quality: Are cases routed by intent, urgency, language, value or skill where appropriate?
- Escalation logic: Are delay, complaint, risk and vulnerability triggers defined?
- Resolution ownership: Is one person or team responsible until the agreed outcome is completed?
- Closure quality: Is resolution confirmed with the customer and correctly recorded?
Measurement and governance
Can leaders see whether communication is becoming more useful and reliable?
- Journey metrics: Are response, conversion, resolution and drop-off measured together?
- Quality review: Are conversation samples reviewed against a consistent standard?
- Exception visibility: Are failed automations, unanswered contacts and repeated escalations visible?
- Change ownership: Is someone accountable for fixing the process, knowledge and system—not only coaching agents?
- AI controls: Are data access, approval, disclosure, monitoring and fallback rules defined for each AI use case?
Use a simple evidence-based scoring method
No defined practice or reliable evidence.
Depends on individuals or varies widely.
Documented and generally followed.
Consistent, monitored and improved.
The total score is directional—not a universal benchmark. Its real value is in showing which weaknesses combine to damage one customer outcome. A low score in cross-channel continuity may matter more than several minor gaps elsewhere if customers frequently switch channels.
Evidence worth collecting
Operational data
Volumes, timestamps, queues, transfers, reopen rates, abandonment, conversion and resolution.
Conversation samples
Anonymised examples across successful, failed, delayed and escalated journeys.
Process evidence
Routing rules, scripts, knowledge sources, ownership maps, handover and escalation instructions.
Frontline reality
Interviews with people handling enquiries, service, operations and reporting.
Use the minimum information needed for diagnosis. Remove or mask personal and confidential data wherever possible, control access and follow the organisation’s applicable privacy and sector requirements.
Turn findings into a prioritised improvement plan
For every material gap, record the customer impact, business impact, root cause, evidence, accountable owner and the smallest useful intervention. Then separate the remedies:
Ownership, sequence, escalation or policy needs to change.
Approved information is missing, inconsistent or difficult to retrieve.
A stable, rules-based task can be triggered or completed automatically.
Language, classification, summarisation or recommendation can support a human.
A bounded system may act, but only with defined authority, monitoring and fallback.
The safest sequence is usually to fix ownership and knowledge before increasing automation. Otherwise the organisation scales inconsistency.
What a useful audit should produce
- A current-state customer journey with evidence.
- A baseline of the most important performance measures.
- Root-cause findings—not only symptoms.
- A prioritised set of process, knowledge, automation and AI opportunities.
- A small pilot brief with acceptance criteria and human handover.
- Clear ownership and a review date.
Method note
The 25-point model is an original Visionary AI Studio diagnostic framework designed for practical discovery. It should be adapted to the organisation’s industry, customer journey, risk profile, channels and applicable obligations. It is not a certification or legal-compliance assessment.