Finance Faces a New Kind of AI: Traffic Surges as Systems Start Thinking for Themselves

Last month, financial institutions saw a sudden shift. Traffic from autonomous AI agents aimed at their websites more than doubled. The jump was sharp and fast. It suggests that finance may be next in line for a wave of artificial intelligence that has already changed other industries.

Security firm Human Security tracked the increase in its latest report. In May 2026, this kind of traffic to financial services businesses climbed steeply, though it still made up less than 1 percent of all activity. The speed of the change matters. It marks an acceleration from earlier periods.

These agents do more than just fetch data. They plan. They carry out tasks that require multiple steps. They adapt without constant human oversight. Moody’s described the shift as moving from simple rule-based automation to intelligent decision support. Early users of related research tools consumed 60 percent more material and finished tasks 30 percent faster. Over 90 percent of interactions now target high-value analytics. Moody’s called this a structural change in how financial teams operate.

NVIDIA’s State of AI in Financial Services report captured the broader momentum. Customer-service applications of generative AI more than doubled in the past year, rising from 25 percent to 60 percent. More than 90 percent of respondents reported positive revenue effects. Banks are already automating document processing and report generation. Cost savings follow. Efficiency gains stack up.

But growth brings friction. FinRegLab examined the implications in a detailed report last fall. Data pulls by aggregators had already doubled at some banks between 2023 and 2025. Agentic systems could push volumes higher. They blur the line between legitimate automation and malicious bots. Bad bots already account for 38 percent of internet traffic to financial firms. Telling good agents from threats will require new investments. Legal questions may also arise.

Regulators are watching. The IMF explored how agentic models could reshape payments. Agents might initiate actions directly. New protocols like the Universal Commerce Protocol and Agent Payments Protocol have started to appear. Experimentation spans fraud detection, compliance, treasury functions and cross-border transfers.

Productivity projections add weight. KPMG estimated agentic AI could deliver $3 trillion in corporate productivity gains. Average EBITDA might improve 5.4 percent annually. Wolters Kluwer projected 44 percent of finance teams will deploy such systems in 2026. That represents more than a 600 percent increase.

Goldman Sachs expects token consumption tied to agentic AI to rise 24-fold by 2030. Monthly volume could hit 120 quadrillion tokens. Jim Schneider, a senior equity analyst at Goldman Sachs, highlighted the multiplier effect from both consumer and enterprise adoption.

Salesforce offered a concrete example from retail that touches finance. AI agents influenced $262 billion in U.S. holiday sales during the 2025 season. Banks that fail to make their systems readable by agents risk being left out at checkout.

Institutions are experimenting anyway. bunq built Finn, an in-app chatbot that now resolves over 90 percent of support tickets. JPMorgan Chase explores agents for fraud detection, personalized advice, loan approvals and compliance.

Risks receive equal attention. Reliability matters when agents execute transactions. Transparency gaps could undermine trust. Error correction grows complex. Financial stability questions surface if agents amplify market moves.

The trajectory sharpens. Traffic data provides one indicator. Productivity forecasts supply another. Real deployments at banks and fintechs offer proof points. The doubling seen last month may look modest in hindsight. Yet it offers a clear signal. Preparation time shrinks.