AI In Marketing: Beyond 'Hi {First Name}' - Using AI for Personalization Without Creeping Customers Out

Chapter 2 of 5

The title is intentionally direct, because as customers ourselves, that is genuinely how it can feel at times.

"Your customer does not want a brand that 'knows everything'. They want a brand that remembers the right things, at the right time, for the right reasons."

Last month a local retail CMO told me: "We finally wired AI into our stack. Open rates are up. Complaints are up faster!"

The team had apparently nailed hyper-personalization. But they had completely missed the "creepy line."

Retargeting people with products they mentioned in private chats. Referencing life events the customer had never shared with the brand. Triggering five channels in 24 hours because the model said the "propensity score was high."

The AI in their marketing was working. The trust in their marketing was leaking.

Why AI Personalization Is Both the Promise and the Problem

AI in marketing finally solves a few things we have complained about for slightly more than a decade:

• Relevance
• Timing
• Next best action

When done well, McKinsey estimates that personalization can lift revenues by 5 to 15 percent and improve marketing ROI by 10 to 30 percent.

But here is where it breaks:

• Superficial, token personalization that feels lazy. The classic "Hi First Name" email, a generic offer, and completely wrong context.
• Crossing the "creepy line." Ads that reference sensitive topics. Remarketing that seems to follow you everywhere.
• Fragmented data and inconsistent journeys. One channel behaves like it knows you intimately. Another treats you like a complete stranger.

The deeper truth: AI does not create creepiness. AI amplifies whatever data ethics and journey design you already have.

A New Mental Model: From "Knowing Everything" to "Remembering What Matters, With Consent"

The question is not "How much can we technically know?" The question is "What has the customer clearly given us permission to use, in this context, for this promise?"

The Three Levels of AI-Driven Personalization

Level 1: Surface Details

Basic personalization using explicit data (name, company, role)

Level 2: Behavioral Signals

Using engagement patterns, browsing history, purchase behavior

Level 3: Context and Consent

Deep personalization based on explicit permission and clear value exchange

Each level adds value. Each level adds risk. You choose how far to go, on purpose.

Drawing Your Own "Non-Creepy" Line

1. Map What You Know vs What You Should Use

List the main categories of customer data you hold, then ask three questions for each:

1. Did the customer clearly know we were collecting this?
2. Did they understand why?
3. Did we explain how it would be used to improve their experience?

2. Apply Privacy, Consent, and Data Minimization

• Use the minimum data required to deliver the promised value
• Honor the purpose for which the data was collected
• Make it easy for customers to change their mind

3. Run the "Comfort Mirror" Test

Before green-lighting any AI-powered personalization tactic, ask:

• Would I be comfortable if another brand used this on me or my family?
• Would I be comfortable seeing a screenshot of this journey in the press?
• Would our Board be comfortable defending this approach in a public hearing?

Designing Journeys That Feel Helpful, Not Invasive

1. Abandonment Journeys That Respect Timing and Channel

• Frequency guardrails: Never touch a customer more than X times in Y days
• Channel priorities: If a customer responds to in-app prompts, cool the emails
• Sensitivity filters: Exclude categories where re-surfacing items could be emotionally risky

2. Post-Purchase Journeys That Add Value, Not Noise

Use AI in marketing to:

• Detect usage patterns and trigger education, not just upsell
• Offer "best practice" content based on role or industry
• Time requests for reviews when value has actually been experienced

The guiding question: "Does this next touch help the customer succeed, or just help us hit a target?"

Using AI to Predict Next Best Action Without Losing Control

Next best action engines are the crown jewel of AI in marketing. They are also the fastest way to create unintentional creepiness if you let them optimize blindly.

Practical guardrails:

• Prohibit actions that rely on sensitive or inferred attributes
• Require high confidence and clear consent for cross-channel triggers
• Suppress activity for customers showing frustration or fatigue signals

How to Measure If Your Personalization Is Working

Commercial Signals

• Engagement rates by segment and journey
• Conversion and revenue per personalized journey
• Time to value and adoption metrics

Risk and Trust Signals

• Unsubscribe and opt out rates after campaigns
• Complaint volume to support and social channels
• Brand sentiment in social listening and NPS verbatims
• Spam reports and deliverability issues

Qualitative Checks

• Listen to real customer calls or chats
• Run short interviews with customers in key segments
• Ask directly: "What felt helpful? What felt too much?"

Wrap Up

The job is no longer "use more data to target harder." The job is "use the right data to create helpful, human journeys that your customer would be proud to see."

Personalization that compounds trust beats personalization that chases clicks.