AI in the Marketing Mix — Rebuild Lumio's Engine Without Burning the Brand

A four-round, advanced marketing-strategy simulation set inside Lumio, a London-headquartered direct-to-consumer beauty brand (GBP 240M revenue, GBP 38M marketing spend, 2.1M customers) built on an 'honest beauty' promise of transparency and clinically backed claims. On the eve of Black Friday, a viral screenshot exposes an AI-generated paid-social ad that fabricated a 'reverses ageing by 40%' claim Lumio never substantiated — shared 60,000 times in 48 hours, it has already cost an estimated GBP 1.2M and dropped the brand-trust tracker from 71 to 64. The board responds with a double mandate: rebuild the marketing operating model around AI within the quarter AND cut blended cost-per-acquisition from GBP 24 toward GBP 19 (−21%) within two quarters — without another trust collapse. Playing the CMO with a fixed GBP 4.0M transformation budget and a brand-safety guardrail (GBP 150,000 per verified incident; two further incidents freeze the AI programme), you (1) diagnose AI not as one switch but as a PORTFOLIO of use cases — targeting, content, and pricing — plotted on value created versus brand-trust risk, with a deploy/pilot/hold call for each; (2) deploy AI across the mix under constraint, setting the autonomy level (human-led with AI assist, AI-led with human approval, or fully autonomous) per use case and splitting the GBP 4.0M across AI tooling, human oversight, and brand-building; (3) write the human-oversight and brand-safety policy — claim-substantiation gates, fairness checks on targeting, pricing transparency, and an incident-response playbook — calibrated strong where claims and fairness are at stake and light where risk is low; and (4) commit and defend a two-quarter plan to the CEO and the board's risk chair, naming a CPA trajectory, a conversion-lift figure, an incident tolerance, and a brand-trust recovery path from 64 back toward 71. The math rewards a risk-weighted portfolio with peer-credible oversight and genuine brand investment, and punishes the five classic errors — treating AI as one decision, chasing CPA with autonomous content and personalised pricing, writing oversight so heavy it kills the efficiency, ignoring algorithmic bias in targeting, and starving brand-building to maximise tooling spend. Final KPIs track blended CPA (GBP), Brand Trust (0-100), Conversion Lift (%), and verified Incidents against the freeze trigger.