Personalisation as competitive strategy in the AI era

Executive overview

Most companies treat personalisation as a marketing tactic — slapping a first name on an email. The real opportunity is redesigning the entire customer experience around individual data, from discovery through support.

When done well, personalisation collapses decision-making for buyers, increases trust, and becomes a moat. When done poorly — or at volume without intelligence — it reads as spam or creepiness.

The core insight: personalisation is a cross-functional operating model, not a marketing feature.

Why most personalisation fails

• Companies default to volume: more touches, more content, more targeting — made easier now by generative AI tools
• At Aetna, analysis showed open rates and actions collapsed beyond four marketing touches per month — less is more
• "Creepy" framing destroys trust: telling a customer "because you call Germany" triggered backlash; "if you call Western Europe" did not — same data, less exposed
• Siloed functions mean personalisation breaks at handoff points: marketing personalises, but service, delivery, and billing don't
• Most CRM and marketing tools (Salesforce, HubSpot, Klaviyo) already have the capability — it's a mindset and process gap, not a technology gap

What meaningful personalisation looks like

• Sungevity sent physical mail with a personalised URL; clicking opened a Google Earth view of the prospect's roof with solar panels rendered on it, a precise energy-saving calculation, and a live video rep pre-loaded with all the data — the buyer didn't look at a competitor
• Sysco (food delivery) loads a personalised app view in 300ms: buyer identity, restaurant menu, consumption rates, local warehouse stock, and promotional offers for items they need to move
• Cisco (tech) built a "CRM on steroids" combining product usage data, content engagement, and firmographic signals to give reps a weekly hit list — who to call, with what, and why — driving up rep adoption, interaction rates, and cross-sell
• Aetna created personalised onboarding videos for new health insurance members: 70% watch rate on a 3.5-minute video, fewer call-centre calls, higher NPS, higher email open rates
• Marriott is building a preference centre for Bonvoy members (pillow type, drink preference, local experiences) and plans a chatbot capable of planning a 10-day trip with points optimisation

The customer journey as a personalisation map

• Every stage is an opportunity: learn, buy, get, use, pay, get support
• B2B companies with channel intermediaries can personalise to the channel (the rep, the purchasing manager) rather than the end consumer
• B2C packaged goods without direct consumer contact can still personalise to retail channel customers using the same B2B logic
• Zero-party data — information customers willingly provide — is among the most valuable; customers share it if they trust you'll use it

Data types and sources

• Behavioural data: what customers do, how often, which features they use, whether usage is trending up or down
• Interaction data: content viewed, emails opened, support calls made
• Firmographic data (B2B): company size, geography, financial health, M&A activity
• Third-party data: Zillow square footage, Google Earth roof orientation, neighbour references — used to build a credible, pre-personalised case before first contact
• Zero-party data: preferences explicitly stated by the customer (pillow type, project type, dietary restrictions)

What it takes to execute

• Personalisation at scale requires a cross-functional commitment — marketing, sales, service, operations, finance, legal
• Marketing is typically the catalyst and quarterback, but cannot own execution alone
• Sungevity took 18 months of test-and-learn cycles to get their system right — envelope design, rendering, rep interaction — before it worked at scale
• Smaller companies have an advantage: fewer functions to align, easier to sit at a table together and agree on a customer-data strategy
• Start by institutionalising what small-company salespeople do naturally — remembering preferences, family details, shared context — and scale that with tooling

Common pitfalls

• Dropping loyalty tier during a gap in activity (Marriott losing a high-value customer post-pandemic) signals the system doesn't understand customer lifetime value
• Blocking loyalty redemption at desirable properties undermines the entire loyalty proposition
• Generating AI content at volume without frequency caps creates the opposite of personalisation — it creates noise
• Personalising the front end (marketing) without carrying it through to delivery and support leaves customers feeling misled