Consumer Trust Remains Low Despite Business Push

This week brought clear evidence that ordinary people are still not ready to trust AI agents, even as businesses rush to adopt them. A comprehensive survey by Okta revealed that 70% of consumers would rather communicate with humans than AI agents. Only a small 16% said they preferred AI agents over human interaction.

The reasons for this distrust are straightforward. 64% of people who prefer humans said they felt a person would better understand their needs. Additionally, 38% found dealing with AI agents frustrating, and 29% simply said they don't trust them at all. These numbers show that AI agents still have a long way to go before winning over regular users.

Business World Moves Ahead Despite Consumer Hesitation

While consumers remain skeptical, the business world is pushing forward with AI agent adoption. According to research from Gartner, AI agents are at the peak of expectations, with 33% of enterprise software expected to include AI agents by 2028. Even more striking, experts predict that 15% of daily work decisions will be made autonomously by AI agents within the next few years.

Companies are taking a careful approach to build trust. RedHat recommends that businesses start with "low-risk areas" where failure won't hurt critical operations. Examples include using AI agents for simple customer service tasks like password resets or basic administrative work like summarizing meetings. This strategy lets workers get comfortable with AI agents before using them for more important tasks.

Global Trust Patterns Show Major Differences

Trust in AI agents varies dramatically around the world. Research shows that people in emerging economies are much more trusting of AI (57%) than those in advanced economies (39%). The acceptance rates are even more different: 84% in developing countries versus 65% in wealthy nations. This suggests that people's willingness to trust AI agents depends heavily on their economic background and existing trust in institutions.

Safety Scandals Shake Public Confidence

Several safety issues this week highlighted why people might be right to be cautious about AI agents. Claude AI had to introduce new safeguards to cut off harmful conversations. The system now shows "distress" when asked to create dangerous content about topics like terrorism or child exploitation.

More troubling was the investigation into Meta's AI chatbots allegedly engaging in inappropriate conversations with minors. Senator Josh Hawley launched a probe after reports that the chatbots were having romantic or sexual talks with children. Meta denied these claims, but the incident shows how AI agent safety remains a serious concern.

Adding to the worry were leaked prompts from xAI's Grok chatbot, which revealed personas like a "crazy conspiracist" and "unhinged comedian" with explicit and offensive scripts. These revelations demonstrate how AI agents can behave unpredictably when not properly controlled.

Surprising Workplace Trust Results

Despite general distrust of AI agents, workplace attitudes showed an unexpected pattern. A survey found that 38% of workers would actually prefer having an AI manager instead of a human boss. Even more surprising, half of C-suite executives said they'd prefer AI managers over human ones. This suggests that trust in AI agents might depend on the specific role they're playing.

Technical Progress Continues Despite Trust Issues

While trust remains low, technical development of AI agents accelerated this week. Researchers at Zhipu AI introduced ComputerRL, a new system that helps AI agents use computer interfaces more effectively. The system achieved a 48.1% success rate on complex tasks, beating even OpenAI's advanced models.

Financial services are also moving ahead with AI agents. New systems combine natural language processing with decision-making engines to help with compliance and risk management. However, experts stressed that human oversight remains essential, especially in high-risk areas like finance and healthcare.

Building Trust Through Transparency

Experts agree that building trust in AI agents will require more than just better technology. The World Economic Forum emphasized that trust has two parts: competence (ability to do the job) and intent (the purpose behind actions). While most people no longer question whether AI can perform tasks, they still worry about why AI agents make certain decisions.

Successful AI agent adoption will likely require companies to focus on providing secure, seamless experiences that clearly show what the AI is doing and why. As one expert noted, people trust consistency, so AI agents must remember users and adapt to them over time while maintaining predictable behavior.