Growth with AI is a trust budget — spend it carefully
AI in marketing, sales, and customer experience
The question: How do we use AI to improve growth without eroding trust or brand?
Marketing and customer service are the most visible AI functions — and the most exposed: personalisation creep, hallucinated promises, and channel spam are trust debits that compound. The Air Canada ruling made chatbot words legally binding.
What the lesson covers
AI now touches the whole customer journey: insight (segmentation, propensity), targeting, content creation at scale, conversational service, next-best-action, churn intervention, and learning loops. The funnel view — insight → segment → personalise → converse → convert → retain → learn — keeps individual tools connected to the journey instead of becoming disconnected gadgets.
Personalisation economics: relevance creates value; creepiness destroys it non-linearly. The line is crossed when customers realise you know things they didn't knowingly share, or when data collected for one purpose powers another (a GDPR issue as much as a taste issue). Consent, purpose limitation, and "would we say this to their face?" are the tests.
Generative content at scale needs brand and truth infrastructure: voice guidelines encoded into prompts/protocols, factual claim checking (products, prices, availability), rights management for assets, and channel governance — because the marginal cost of content collapsing to zero means channel spam is now a strategy failure mode, not a budget constraint.
Customer-service automation must be measured beyond deflection. The Klarna arc (Lesson 1) and the industry's hybrid convergence (roughly 60–70% AI on routine, deterministic issues; 30–40% human on complex, high-empathy cases) show that resolution quality, escalation design, CSAT, and liability are the real scoreboard. Workflow displacement — AI absorbing routine volume so humans handle complexity — is the sustainable pattern.
Liability is real: in Moffatt v. Air Canada (2024), a tribunal held the airline responsible for its chatbot's invented bereavement-fare policy. Your bot's words are your company's words. Guardrails, grounding in actual policy (RAG), escalation triggers, and audit logs are not optional in customer-facing deployment.
Key points
- Connect every AI tool to a funnel stage and a metric — and to the trust risk it creates.
- Personalisation value rises with relevance and collapses with creepiness; consent and purpose limitation are the tests.
- Zero-marginal-cost content makes channel spam a failure mode; brand voice and claim-checking are infrastructure.
- Measure service automation by resolution quality, escalation, CSAT, liability — not deflection. Hybrid (~60–70% AI) is the stable pattern.
- Your bot's words are legally your words (Air Canada, 2024).
Framework — Customer AI Funnel
insight → segment → personalise → converse → convert → retain → learn. For every AI intervention, name the funnel stage, the metric it moves, the trust risk it creates, and the guardrail that caps the risk.
The lab
Redesign a customer journey with AI — with guardrails designed before launch.
Open this lesson, its lab and its quiz
Sources and further reading
- Moffatt v. Air Canada (2024) — decision — BC Civil Resolution Tribunal
- Klarna AI assistant press release (read critically) — Klarna