9.3 for the human, 7.0 for AI: how to close that gap in customer service

In the CX Brain survey with AntennasBI, human service got an average score of 9.3 and AI service 7.0. For us, the 2.3-point gap doesn’t come from a lack of technology, but from design: the customer misses empathy, an exit to a person and someone to solve the hard problem. Five adjustments attack that.

What the survey measured

How Brazilians see AI in customer service was created by CX Brain, the data unit of Grupo Padrão, and run by AntennasBI in 2024. It heard 1,003 Brazilians from every region and social class, who on average had talked to 5 companies in the previous three months. More than 2,000 videos they recorded on their phones were analyzed.

One limit: the article we consulted doesn’t state the exact question or the score scale. Read the 9.3 against the 7.0 as direction, not as an exact ruler.

Where AI loses points

Among the emotions people reported in AI service, 56% were negative and 34% positive. Boredom appeared in 16%, anger and indifference in 13% each, and distrust in 10%. Another 51% think AI will help little or nothing in customer service.

The most skeptical respondents said AI doesn’t solve complex issues and that, in those cases, the human is more trustworthy. They miss closeness and informality. In the article’s own reading, the central problem is the lack of empathy, not the technology itself.

Five adjustments to close the gap

This list is our reading of the survey. None of the five was tested by the study, so it’s worth validating in your operation.

1. Show the exit to a person right at the start. In the 2026 Globo survey, 79% want to be able to choose between human and AI. We cover the details in 74% accept AI in customer service, 79% want to choose.
2. Give AI only what it resolves. If the skeptical customer doesn’t trust AI for complex cases, serious complaints and disputes should start with a person. The split we propose is in hybrid service.
3. Pass the context along. When the conversation goes from AI to a person, the agent needs to see the history and the summary. A customer who repeats the problem feels indifference.
4. Write the answer to resolve. Boredom and indifference add up to 29% of reported emotions. In our reading, generic, repeated answers feed that. A good answer cites the customer’s problem and states the next step and the deadline.
5. Read the conversations where the customer asked for a person. Every week, pull a sample of those cases and find out what AI didn’t understand.

How to measure if the gap is closing

There’s no universal target. Measure the AI score and the human score separately, by channel, and track the distance between the two. Add the 7-day recontact and the share of customers who ask for a person. The step by step is in how to measure an AI service agent.

If your operation already has AI in customer service and the score is lower than the human team’s, the first step is a diagnosis. NorthStar does that in sales and service with AI.

Frequently asked

What do the 9.3 human score and the 7.0 AI score mean?

They are the average scores respondents gave to human service and AI service in the CX Brain and AntennasBI survey (2024, 1,003 Brazilians). The article doesn’t state the scale used.

Does AI have to reach 9.3?

In our reading, it doesn’t have to match the human at everything. It has to resolve well what it takes on and pass the rest without losing the context. The score rises when AI stops attempting what it doesn’t do well.

Why does the customer feel indifference with AI?

In the survey, indifference was cited by 13% and boredom by 16%. Respondents point to a lack of empathy and closeness. Standardized answers and repeated questions make it worse.

Where to start improving?

With adjustment 1: keep the exit to a person visible in the first message. Then measure the score by channel and read the conversations where the customer asked for an agent.

Sources

• Consumidor Moderno, on the survey How Brazilians see AI in customer service (CX Brain / Grupo Padrão and AntennasBI), Sep/2024. consumidormoderno.com.br
• Globo Gente / E&I Inteligência Setorial, Aug/2026 (the 79% figure used as context). gente.globo.com