How Much is Too Much? Navigating the Fine Line With AI-Driven Personalization


Share on LinkedIn

It’s remarkable how seamless AI-driven experiences can be in helping us quickly find a product or service, make a recommendation or facilitate a purchasing decision. Some AI is so subtle that we may be unaware it is working behind the scenes to make our online lives easier and more manageable. At other times, however, AI can feel like an intrusive assistant, preemptively filling in the blanks based on vast amounts of data collected about us—often giving the unsettling impression that brands know more about our preferences and habits than they should.

For example, I was recently shopping online for a pair of red runners that had been sold out since Christmas. Lucky for me, I found the elusive shoes. While searching for a store nearby, I added a different address from my own. Without entering my name, the site’s location finder unexpectedly updated to reflect my street address, likely using information from a customer database or my IP address, which was surprising yet somewhat expected.

This is not unlike someone in the real world knowing your dog’s name, yet you don’t recognize them. As I continued my day, my social media feed was bombarded with ads for my newly acquired shoes, which continued for the next few weeks.

When over-personalization becomes overbearing

When you consider the vast amount of data that brands collect with every customer interaction, it’s not surprising that in some cases, AI can get a little too personal with us. Many of us have had the experience of opening our email to a full inbox of unsolicited emails that touch on a personal topic that might be a bit unexpected. While we uncovered in our recent survey that 62% of Americans prefer a more personalized customer experience (CX) over one that takes less time, it’s also not surprising that when brands use personalization strategies that use sensitive data without permission, this can actually turn customers away.

For example, targeting ads based on private health information can feel particularly invasive. In fact, according to the Cisco 2023 Data Privacy Benchmark Study, 81% of consumers agreed that the way an organization treats their data is indicative of how it views and respects its customers.

Generative AI (GenAI) has heralded a new era of personalized CX, but brands must be careful about where and how they use it. Too much personalization may feel intrusive to some, while too little can result in your brand missing opportunities to engage and connect with consumers. Consider the difference between a brick-and-mortar store clerk suggesting products to a regular customer based on past purchases and a GenAI chatbot that makes recommendations to a shopper based on data algorithms. How can brands leverage GenAI but yet walk that fine line?

Know your customer

One key factor to staying balanced with GenAI-enabled personalization strategies is to truly know your customer. Knowing your customers’ preferences, behaviors, and pain points allows brands to train their GenAI models more effectively, ensuring recommendations and engagements are not just accurate, but also contextually relevant, thereby enhancing customer satisfaction and loyalty. This includes understanding the topics of interest, method and frequency by which they wish to be contacted. Frequent, unsolicited emails suggesting similar items to a recent search can overwhelm and lead to disengagement or an unsubscribe.

It’s also important to understand your customers’ preferences when they do engage with your brand. For example, in our above-mentioned survey, results showed that 70% of consumers would rather speak with a human agent than an automated bot about product availability, but 56% preferred an automated bot when resetting passwords. But, if given the option between texting or speaking with a bot or human agent, 72% of respondents would prefer to chat or text with an automated bot, while 57% would prefer speaking over texting (43%) with a human agent.

These insights indicate varying levels of consumer comfort when engaging with AI, which is also dependent on the purpose behind the outreach and the perceived complexity of the task. Brands that align their GenAI use with their consumers’ preferences will enable a more streamlined CX while balancing their need for efficiency and information.

Integrating GenAI thoughtfully

Let’s look at how a financial institution might leverage GenAI practices effectively without over-personalizing to serve their customers’ journey. Starting with predictive analytics, a company can leverage GenAI with advanced machine learning algorithms to predict customer behaviors and preferences. This allows for anticipatory customer service, such as predicting when a client may need additional liquidity or might be interested in investment opportunities, often before they identify the need themselves.

Next, GenAI-driven recommendation engines can further personalize financial advice, suggesting products or services tailored to individual financial goals and risk profiles and recommending a diversified investment portfolio. Using an omnichannel and multimodal approach, the financial institution can create and deliver content in a variety of formats—from emails to webpages to calls and texts—to ensure the customer receives information where, when and how they prefer. Simultaneously, the GenAI platform learns with each interaction, adapting to real-time customer feedback while maintaining a consistent and brand-aligned message across all platforms.

Brands should also implement A/B testing while leveraging GenAI in their CX to help them determine the most effective interactions. This process enables them to test between two options simultaneously to determine which one performs better. For example, one customer might receive a product recommendation based on past purchases, while another gets one based on their browsing history. By using GenAI-powered A/B testing, companies can discover which approach appeals more to customers and drives sales while gauging the effectiveness of their personalization strategies.

Prioritize robust governance practices

Some customers might be uncomfortable not knowing how or why a service is overly personalized. For example, if a banking app offers advice on spending habits, users might worry about how their financial data is being analyzed and used. Brands must be clear and transparent with their customers on what data is being collected and stored and ensure they have easy access to adjust these settings as needed. What’s more, incorporating a robust governance framework to ensure the safety and security of customer data is crucial.

Brands must also ensure they incorporate a human-in-the-loop (HITL) approach throughout the GenAI lifecycle and employ diverse teams to train, monitor and review the platform’s outputs to mitigate bias and inaccuracies.

For example, when algorithms predict too much about an individual’s personal preferences or behaviours based on demographic data, it can lead to incorrect assumptions, such as suggesting products based on gender and ethnicity, which, in addition to exacerbating harmful biases, will result in customer attrition and potentially a larger reputation management issue if it is not stopped.

Working with an experienced CX partner with the right technology and expertise to thoughtfully implement AI-powered solutions can help brands improve the performance, adaptability and safety of their GenAI models. The potential for AI-driven personalization to revolutionize CX is immense. Brands that master the delicate balance between just enough and too much personalization will forge deeper connections with their customers.

Michael Ringman
Michael Ringman is the Chief Information Officer at TELUS International and has been with the company since 2012. As CIO, Michael remains focused on driving continuous innovation for both customers and team members, and has built his career on implementing technology services, especially developing public and private cloud solutions for retail, government, technology and finance verticals. Michael holds a Bachelor of Science degree in Aerospace Engineering and a Master of Science in Telecommunications, both from the University of Colorado.


Please use comments to add value to the discussion. Maximum one link to an educational blog post or article. We will NOT PUBLISH brief comments like "good post," comments that mainly promote links, or comments with links to companies, products, or services.

Please enter your comment!
Please enter your name here