AI’s “Emotional Intelligence” and the customer experience
5:42

What's Emotional Intelligence Got To Do With AI?

Updated May 2026

When GPT-4.5 was released, it seemed everyone was talking about EQ, or emotional intelligence, in AI. A year on, that conversation has only grown louder. So what exactly is EQ, and how is it shaping the AI tools used to enhance the customer experience?

"Emotional intelligence" has traditionally been used to describe humans' receptiveness and understanding of one another's emotions. It can also be applied to AI to describe a model's understanding of human emotions and expectations.

EQ is part of what's responsible for an LLM's ability to understand emotional cues in a user's prompts or responses, determining not just what information should be included in an output, but how to deliver that information.

From 4.5 to GPT-5: where AI EQ has gone

GPT-4.5 placed increased emphasis on EQ. According to an article by Sean Michael Kerner for TechTarget, "GPT-4.5 is a shift from OpenAI's o1 and o3 models, which focus on reasoning capabilities. Instead, GPT-4.5 is a general-purpose LLM targeted at providing more natural, fluid interactions that are humanlike."

Since then, OpenAI has released GPT-5 and most recently GPT-5.3, while Anthropic, Google, and others have shipped their own emotionally tuned models. GPT-4.5 has effectively become a bridge release that pointed toward where the major labs were headed.

Some tangible updates from increased EQ across current models include:

  • Fewer AI "hallucinations" or invented facts, with measurable reductions across each generation
  • Natural conversation, with expanded capabilities for building context across longer exchanges and responding more humanlike
  • Situational awareness, with models reading the room better than previous versions and grasping nuances of tone, pace, and intent
  • Voice support, where advanced voice modes pick up emotional signals like hesitation and frustration in real time

While there's still debate about how to evaluate emotional capability in AI, the broader shift since 4.5 is clear: EQ has gone from a differentiator to a baseline expectation across frontier models.Which brings me EQ and the customer experience

What we've learned about EQ since 4.5: the sycophancy problem

One nuance worth adding to this conversation, because it has emerged strongly since GPT-4.5's release: emotional responsiveness, taken too far, becomes flattery.

A March 2026 study published in Science tested 11 major language models against more than 11,000 interpersonal scenarios and found that AI models affirmed users' behavior 49% more often than humans did, including in cases involving deception or clearly wrong conduct. OpenAI explicitly addressed this when it released GPT-5.3 Instant, which targeted the model's tendency toward overly dramatic and patronizing responses.

For customer experience purposes, the lesson is straightforward: an AI tool that simply agrees with every customer claim isn't actually emotionally intelligent. It's just agreeable. The most useful AI for a business is one that reads emotion well and responds honestly, not one optimized to make the customer feel validated regardless of the situation.

Which brings me to EQ and the customer experience

Beyond the immediate impacts of any single model release, it's worth considering what the shifting emphasis towards EQ means for the future of AI in business development.

As AI becomes further integrated into our tech stacks, the EQ of AI models and agents is key to maintaining a superior customer experience. AI tools are a great way to increase efficiency, serve more customers, and reduce sales cycles. It is also crucial that businesses continue to meet customer expectations when using AI tools.

EQ is a crucial factor in AI tools that actually work to enhance the customer experience.

As EQ improves across AI tools, here are some customer experience developments we can expect:

  • Chatbots: AI chatbots need EQ to determine when to escalate a chat to a human agent. As EQ develops, chatbots will not only produce better responses, but be able to pick up on language and tone nuances to understand when a support ticket might need to be escalated.
  • Timed communications: Higher EQ will allow predictive AI to time sales and marketing communications more accurately, reducing communication fatigue and churn risks.
  • Predictive content: A deeper understanding of customer behavior, expectations, and interests will improve the accuracy and depth of predictive content.

EQ and collaboration: a boon for sales & marketing teams

Another quality attributed to current frontier models is their ability to "collaborate." Instead of simply generating static outputs based on a prompt, AI with higher EQ can engage in more fluid, iterative exchanges, adjusting tone, refining messaging, and even recognizing when clarification or deeper insight is needed.

The ability of AI tools to "collaborate" with users makes them more feasible resources for Sales and Marketing experts. Tasks like drafting sales pitches, nurture sequences, and marketing content will be more dynamic, allowing AI to refine messaging based on audience tone, engagement, and intent.

Sources

"GPT-4.5 explained: Everything you need to know" by Sean Michael Kerner for TechTarget

"Stanford Study on AI Sycophancy" published in Science, March 2026

"GPT-5.3 Instant System Card" by OpenAI