Mastercard Predicts 1 in 10 Online Shoppers to Use AI Agents by 2030

Key Takeaways

• Mastercard predicts one in ten online shoppers will use AI agents for purchases by 2030.
• 27% of teenagers expect to use fully AI-run shopping assistants compared to parents.
• Only 23% of consumers trust generative AI to handle payments, but 61% trust Visa.
• Visa CEO Ryan McInerney identifies trust as a major barrier to autonomous payments.
• Mastercard launched Agent Connect and Agent Pay to help merchants secure AI-driven commerce.
• Investors poured $435 million into enterprise AI security between April and September 2026.
• 88% of companies fail to launch AI projects into production due to safety and cost issues.
• AI spending per employee dropped nearly 10% at top firms in August as costs decreased.
• Companies are shifting to older models like OpenAI's ChatGPT 5.6-Terra and Anthropic's Sonnet.
• Chinese gig workers are training AI models for specialized tasks like writing building proposals.

Mastercard predicts one in ten shoppers use AI agents by 2030

A new report from Mastercard says more than one in ten online shoppers could use AI agents to buy products by 2030. These AI tools will recommend items, compare prices, and complete purchases based on rules set by the user. Teenagers are already more likely to use these tools than parents, with 27% of teens expecting to use fully AI-run shopping assistants. Retailers will need to make their product information machine-readable so AI agents can verify claims and policies. Trust and authorization will become critical as regulators may issue guidance on monitoring these AI agents.

Mastercard launches Agent Connect to help merchants scale AI shopping

Mastercard has launched a new solution called Mastercard Agent Connect to help merchants prepare for AI-powered commerce. This tool allows businesses to control how their products are represented and supports trusted transactions authorized by users. The company also introduced Mastercard Agent Pay to help merchants secure agentic transactions. Recent research shows merchants are investing heavily in building demand while controlling payments and pricing. These new capabilities aim to let merchants participate in a market where AI assistants handle product discovery and comparison.

Visa research shows consumers trust payment brands over AI for purchases

A new study from Visa reveals that while 72% of consumers use AI assistants, only 23% trust generative AI to handle payments on their behalf. Trust increases significantly when a known payment brand is involved, with 61% of consumers trusting Visa to manage agentic transactions. Most people will let AI search for products or manage loyalty programs, but only 35% are willing to share saved payment credentials. Experts say that broad adoption of AI shopping depends on having transparent controls and trusted payment experiences anchored in banks and card networks.

AI helps shoppers find deals but consumers still click buy button

Nearly 50 million U.S. adults now start their product research using AI, and this technology changes purchase decisions for 83% of users. While AI helps find better prices and new brands, 58% of consumers prefer to browse with the tool and complete the purchase themselves. People who follow AI recommendations to new retailers spend an average of $1,430, which is much higher than the $931 spent by those who stick to familiar sites. Visa CEO Ryan McInerney noted that while shopping adoption is high, trust remains a barrier for autonomous payments. Merchants are hesitant to let AI handle final pricing, with 46% saying this is the function they are least willing to delegate.

CyberDockside.ai and HexaShield partner for maritime cyber resilience

CyberDockside.ai and HexaShield have partnered to create a service that improves cybersecurity for the maritime industry. The initiative uses HexaOT, a passive monitoring platform that tracks operational technology assets without disrupting safety-critical systems. CyberDockside.ai will manage client engagement and conduct assessments before wider deployment of the solution. The service covers compliance with frameworks like NIS2, IMO requirements, and IACS rules. It also includes managed detection, response, and threat intelligence to protect systems such as quay cranes and shipboard controls.

New Router Prior Bias improves AI model accuracy

Researchers created a new method called Router Prior Bias to improve how AI models are trained after their initial setup. This technique helps keep the original structure of the model's routing system instead of forcing it to become uniform. Testing on the Moonlight-16B-A3B model showed this approach achieved 45.77 percent accuracy, which is much higher than other methods. The same results were seen with the Qwen3-30B-A3B-Base model family, proving the method works across different systems.

Investors pour $435 million into enterprise AI security

Venture capital investors spent $435 million on 12 funding rounds for enterprise AI agent security companies between April and September 2026. Most of this money went to firms building safety tools because 88 percent of companies fail to launch their AI projects into production. Experts warn that over 40 percent of agentic AI projects could be canceled by 2027 due to high costs and poor risk controls. The funding surge shows that businesses now prioritize governance and safety stacks to move AI tools from testing to real use.

AI spending per employee dropped at top firms in August

Data from Ramp shows that AI spending per employee fell nearly 10 percent at the top 1 percent of companies in August. The average cost dropped to $7,205 per employee as prices for AI tokens decreased to $0.68 per million tokens. Many businesses are now using older, cheaper models like OpenAI's ChatGPT 5.6-Terra and Anthropic's Sonnet instead of the newest powerful versions. This shift suggests that lower costs are driving companies to reduce overall spending rather than increasing their use of advanced AI tools.

Convo launches AI coach to help sales teams win deals

Convo released a new AI sales coach that helps sales representatives handle objections and answer questions during live customer calls. The tool listens to conversations and suggests responses based on the buyer's concerns without interrupting the meeting. Sales teams can upload their own pricing sheets and product documents so the AI can provide accurate answers with sources. After the call, the system drafts follow-up emails and captures action items to help close the deal.

Chinese experts work as gig trainers to build AI data

Skilled professionals in China are taking on gig work as specialized AI trainers to help build high-quality data sets for artificial intelligence. These workers, including architects and software engineers, use their real-world experience to teach AI models how to perform complex tasks like writing building proposals. They earn between 100 and 500 yuan per task but face challenges such as low pay and rejected work outputs. This trend is growing as Chinese tech giants compete with Silicon Valley to create better AI tools for specific industries like law and finance.

Sources

• More than one in 10 online shoppers could routinely use AI agents by 2030: Mastercard
• Mastercard Expands Agentic Suite to Help Merchants Scale AI Shopping
• Consumers Use AI Assistants but Hesitate to Hand Over Their Wallets
• AI Takes the First Step in Shopping While Consumers Keep the Buy Button
• CyberDockside.ai and HexaShield Partner for Maritime Cyber Resilience
• Router Prior Bias: Preserving Base Routing Structure in MoE Post-Training
• Enterprise AI Agent Funding Surges to $435M in Five Months
• AI spend per employee slumped at top firms in August — summer doldrums or a warning sign?
• Convo’s AI sales coach helps sales teams win more deals during live calls
• Now it’s China’s experts who are gig workers training AI data