Tariffs are back, and they’re no longer just an economic concern. They’re introducing significant uncertainties that affect both e-commerce businesses and online shoppers. In fact, 81 percent of brands already foresee disruptions in their global strategy, while 60 percent of consumers are becoming more value-focused and price-conscious.

For online retailers, tariffs are more than just a financial hurdle. The average cart abandonment rate for the industry is 70 percent, with unexpected costs, slow deliveries, and unnecessarily complicated user experiences topping the list of reasons why shoppers bail, making transparency and experience the baseline for trust.

Where does AI fit into this? Too many businesses still use AI like a wrench. It’s great for tightening processes and workflows. Used as such, but it’s not so good at holding customer relationships together. Automating tickets and fine-tuning personal recommendations are useful, but when tariffs disrupt expectations, customers don’t just need speed. They need signals of empathy. In many cases, however, we’ve found that AI-powered solutions such as chatbots often don’t listen and respond well, even with something as basic as giving meaningful information about products or services.

It might be time to rethink your brand’s use of AI as an operational tool and start building it into your customer relationship infrastructure.

AI in the relationship layer across the customer journey

When treated as the emotional shock absorber of tariff-shaken e-commerce, AI can be tuned to detect signals of frustration, confusion, or hesitation. How would that look across each stage of the e-commerce journey?

Building confidence during browsing: Uncertainties start the moment customers land on your site or app. Shoppers grow wary when prices appear “too good to be true”, suspecting hidden fees or complex cross-border costs just waiting to surprise them. The result is hesitation, drop-off, and growing distrust in the browsing experience itself.

AI can detect this emotional undercurrent through real-time behavioral modeling, tracking micro-interactions that go beyond conventional metrics like time on page or bounce rate. Patterns such as repeated back-and-forth between product and cart pages, stalled scrolling, or mouse hovers over price elements are strong indicators of uncertainty or confusion, as demonstrated in previous experiments on human-computer interactions and the Massachusetts Institute of Technology’s report. Session-level sentiment analysis can take it further by using anonymized behavioral data to estimate emotional state.

Once these signals are detected, AI can activate dynamic personalization tools. This includes offering localized shipping guarantees, or even adjusting UI elements to foreground trust signals like badges, verified reviews, or local warehouse information to reassure hesitant shoppers in their purchasing decision. The right message, placed at the right moment, helps restore confidence and keeps the journey moving forward.

Using confidence scores to ease checkout hesitations: According to Deloitte, 70 percent of customers value pricing transparency more than discounts. Additionally, more than half of retail executives recognize that price clarity outweighs brand loyalty. In tariff-shaken markets, that gap becomes a dealbreaker.

AI systems can generate and display confidence scores that reflect the certainty of price calculations, eligibility for discounts, or the accuracy of product recommendations. When these scores are shown to customers—such as “We’re 95 percent confident this is your best price”— it provides a quantifiable assurance, making the process feel more transparent and less arbitrary.

Predictive abandonment models, trained on behavioral signals like sudden cursor hesitations, toggling between cart and price breakdowns, or backtracking from the payment screen, can identify when a customer is wavering. Rather than waiting for drop-off, AI can trigger context-aware intervention, such as offering a clear, personalized breakdown of tariffs or duties based on the shopper’s location and basket contents.

This is where large language models (LLMs) can come into play. Rather than serving vague legalese, LLMs can generate clear, contextual messaging based on real-time variables like shipping location, cart contents, and tariff thresholds. To support this, confidence-scoring systems can analyze checkout behavior, gauging how close a customer is to converting or abandoning. These scores can dynamically inform next-best actions, such as offer placements, chatbot prompts, and localized reassurance messages.

Proactive communication through shipping and delivery: Tariff-induced customs delays can spike frustration, especially when communication breaks down. More than half of delivery dissatisfaction reportedly comes from unclear or delayed updates, often tied to international shipping and customs bottlenecks.

While brands that rely on third-party logistics partners might not control the delivery chain, they can still steer the customer conversation. For example, AI-enhanced predictive delivery-risk models, which can be trained on real-time customs data, lane performance, local traffic conditions, delivery density, and carrier history, can help flag high-risk shipments early in the process. The insights would enable brands to trigger proactive notifications to customers. Even if a brand doesn’t own the delivery infrastructure, it can integrate these insights into its CX systems.

To make those updates meaningful, geo-behavioral analytics can calibrate messaging based on local expectations. What feels like a delay in Singapore, for instance, might be routine in Spain. AI can tailor communication cadence and content based on these tolerance thresholds. From there, generative AI can help craft personalized, tone-aware messages that reflect not just the delay, but the customer’s previous interactions. When a shipment requires a human touch, AI can flag it for agent-led outreach, equipped with context, sentiment history, and suggested next steps.

Preserving relationships during returns: Tariffs don’t just complicate the sale. They make returns more confusing, too, introducing uncertainty around refunds, duties, and restocking fees. Returns are often the final impression a customer has of your brand, and nearly 70 percent of shoppers say a single poor returns experience is enough to stop them from coming back.

AI can act as an early warning system. For example, when customers initiate return requests through channels like phone, email, live chat, in-app messaging, or self-service portals, emotion-tagging natural-language-processing (NLP) models, integrated into the brand’s CX platforms, can analyze these inbound messages to detect and flag emotional cues, such as frustration, disappointment, or perceived unfairness. High-risk cases can be escalated to human agents with relevant context, such as churn probability and sentiment history. This enables brands to apply emotion-aware triage, aligning urgency and intervention based on how the customer feels and the potential business impact of each return.

From there, predictive retention models can evaluate a customer’s past purchase behavior, return frequency, and estimated lifetime value to recommend proactive steps, such as full or partial refunds, store credit, or even personalized outreach by human agents.

AI as the new relationship infrastructure

Each stage of the buyer journey illustrates how AI moves from automating tasks to learning to read the room. Tariff changes and other market shocks inject friction, but it’s not the disruption that breaks loyalty. It’s how brands respond to and resonate with customers.

That’s the return-on-investment (ROI) that CX leaders need to rethink when using AI. It will increasingly measure beyond efficiency or cost reduction, and account for emotional responsiveness, effectiveness of interventions, and preservation of trust. When part of the relationship, AI becomes uniquely capable of recognizing and responding to emotional context at scale. It picks up the signals that human CX frontliners can act on. Together, they form a loop that strengthens every interaction: anticipating needs, adapting in the moment, and reinforcing trust where it matters most. In trade “wars” where customer emotions are the battleground, your AI needs to do more than automate. It needs to be the lending ear that absorbs the shocks.


#EcommerceAI #CustomerExperience #TariffImpact #AIinRetail #EmotionalIntelligence