AI Can Move Money. Can It Make the Right Financial Decisions? | Opinion

The global expenditure on artificial intelligence (AI) is projected to reach $2.7 trillion by 2026, a significant 49.5% increase from the previous year, according to Gartner. Concurrently, financial systems are adapting to accommodate machine-driven transactions. In June 2026, Mastercard unveiled Agent Pay for Machines, a solution aimed at facilitating agent-assisted transactions, as reported by Yahoo Finance.

As AI transitions from offering financial insights to potentially executing payments and managing funds, a critical question arises: how can these systems make decisions that accurately reflect an individual's actual financial situation, rather than merely responding to numerical data? While AI agents can swiftly execute transactions, they must possess a deeper understanding of the context to ensure the decisions align with a person's financial reality.

This distinction is crucial, as a technically sound decision could inadvertently lead to financial complications. To address this challenge, the financial sector must develop AI systems capable of making nuanced decisions based on comprehensive context. An AI agent might approve a purchase based on a seemingly adequate balance, unaware of impending rent due dates, loan repayments, or fluctuating income patterns.

The issue lies not in inaccurate data but in the AI's limited interpretation. Financial systems have become proficient at recording and organizing transactions, categorizing them by merchant, amount, and spending type. However, categorization alone does not provide insight into the underlying reasons behind a transaction or its broader implications for a person's financial position.

For instance, a substantial transfer to a family member might appear as an unusual expense, potentially indicating an unconsidered recurring responsibility that must be factored into disposable income. Similarly, changes in spending could signal an emerging problem or represent a planned, temporary expense. These nuances are vital when AI systems are tasked with making decisions on behalf of individuals.

The future of financial technology will necessitate a greater emphasis on contextual information. This involves understanding the interplay between spending, income, savings, borrowing, and investments, rather than treating each transaction in isolation. Financial behavior often occurs across multiple systems, despite the interconnected nature of decisions.

An AI agent must account for these interrelationships to make informed decisions. As financial services become increasingly automated, the ability to reveal these relationships is becoming increasingly important. While AI agents can discern that someone spent money on groceries, travel, or dining, a broader financial context can help the system assess how that expense relates to the person's income, obligations, goals, and established spending habits.

The distinction between categorization and context is significant. Categorization describes the transaction, while context helps determine its significance within a specific financial scenario. However, this does not imply that AI agents should infer every aspect of an individual's life from their spending. Financial behavior can be complex, and a single transaction often provides insufficient evidence to understand someone's intentions.

Contextual systems must navigate uncertainty and avoid making assumptions that lead to decisions. Moreover, they must differentiate between useful and pertinent data. Context should be built upon relevant, consented data with clear access limitations, robust safeguards, and respect for individual privacy. Otherwise, a system designed to enhance financial decision-making could introduce new risks.

A practical illustration of this concept can be seen in an AI travel agent. A conventional agent might search for flights within a specified budget and select the most economical option. However, a more context-aware agent could also consider upcoming financial commitments and agreed spending limits. Instead of preventing the purchase outright, the agent could recommend a more affordable option, request confirmation, or adhere to previously established guidelines.

Similarly, an AI managing various obligations, such as bill payments and savings, would need to identify essential payments, anticipate incoming funds, and evaluate whether allocating funds to savings would compromise liquidity for future expenses. The effectiveness of the decision hinges on the quality of the available context.

Financial institutions face a similar challenge in leveraging contextual information for lending assessments, savings tools, and customer service interactions. While additional data could support these functions, financial institutions must avoid automatically translating this information into more automated decision-making. Clear guidelines are required on how data is evaluated, errors are addressed, and accountability is maintained when decisions yield adverse outcomes. This is particularly crucial in lending, where financial data can impact credit access.