LOOKING BACK | Here Come the Financial Agents

The financial services industry has spent decades automating transactions, digitizing customer relationships and teaching computers to recognize patterns. Now, artificial intelligence is threatening to fundamentally change how financial institutions operate, how their employees work and how consumers manage their money. The arrival of agentic AI promises something considerably more disruptive than another generation of chatbots or analytical software: autonomous digital workers capable of researching, deciding and acting on behalf of financial institutions and their customers. Between August and early October 2026, announcements from major technology companies, financial institutions and regulators suggest that the financial industry is entering an era in which software will increasingly do more than recommend financial decisions. It will execute them.

That prospect is attracting enormous commercial interest while generating equally significant concerns about employment, cybersecurity, financial stability and the survival of traditional financial business models.

On October 1, European Central Bank President Christine Lagarde delivered a speech titled “Where AI Risks Meet”, warning that increasingly autonomous AI systems could introduce new vulnerabilities into financial markets.

Lagarde observed that nearly nine out of 10 significant euro-area banks already use generative AI, although most applications still involve limited autonomy. She warned that more sophisticated agents could independently devise trading strategies, exploit cybersecurity vulnerabilities and amplify market disruptions.

Meanwhile, Apollo Global Management Chief Economist Torsten Slok raised another possibility: AI agents could undermine banks by automatically transferring customer deposits to institutions offering better interest rates.

In a September 27 commentary titled “Is an Agentic Bank Run Coming?”, Slok suggested that autonomous financial assistants could fundamentally change deposit behavior, potentially threatening the low-cost funding on which traditional banking depends.

These developments raise a question that extends far beyond technological innovation: What happens when financial institutions and their customers begin delegating financial decisions to machines?

From Financial Assistants to Financial Agents

Agentic artificial intelligence represents an evolution from AI systems that primarily generate information toward systems capable of pursuing objectives and executing multistep workflows.

Traditional generative AI applications generally respond to prompts. A financial analyst might ask a chatbot to summarize an earnings report, compare investment strategies or explain a regulatory requirement. The system produces an answer, leaving the human user responsible for subsequent action.

Agentic AI introduces a degree of independent initiative.

An AI agent can receive an objective, develop a plan, access relevant information, interact with software applications and execute a sequence of actions. Depending on its permissions, it may also monitor results, adjust its approach and escalate problems requiring human intervention.

An agent tasked with preparing a commercial loan application, for example, could retrieve financial statements, verify borrower information, evaluate creditworthiness, identify missing documents and prepare recommendations for an underwriter.

More autonomous systems might eventually complete substantial portions of the lending process without continuous human direction.

As Google Cloud explains, agentic systems combine reasoning, planning, tool use and action to pursue objectives across interconnected applications.

Importantly, autonomy exists along a spectrum. Some systems merely organize information and recommend actions. Others execute approved transactions within established limits. More ambitious applications seek to operate with minimal supervision.

The distinction matters enormously in financial services, where mistakes can result in monetary losses, regulatory violations and consequences for consumers.

In a September 23 analysis, the Center for Democracy and Technology cautioned that the capabilities advertised for financial agents sometimes exceed what current deployments can reliably accomplish.

Nevertheless, the direction of development is unmistakable.

The financial industry's next major technological transition may be from software that assists financial professionals to software that performs financial work.

Financial Institutions Begin Deploying Their Agents

Perhaps the clearest illustration of this transition arrived August 25, when Google Cloud introduced Gemini Enterprise for Financial Services.

The platform was designed specifically for capital markets and corporate banking, combining specialized AI agents with financial data, research capabilities and enterprise software integrations.

Its Financial Research agent includes more than 50 specialized skills and connections to financial information providers, including FactSet, Moody's, MSCI, S&P Global and LSEG.

The system can assemble research, analyze financial information and prepare reports while providing citations and supporting documentation.

Google identified Deutsche Bank and CME Group among the institutions involved in developing or using the technology. Other financial organizations adopting Google's broader enterprise AI capabilities include BNY, Citi Wealth, Lloyds Banking Group and Macquarie Bank.

The announcement demonstrated how agentic AI is evolving beyond general-purpose productivity software into specialized financial infrastructure.

Two days later, Google published a review of eight financial institutions developing AI-agent applications, illustrating how the technology is spreading across banking operations, customer service and employee productivity.

Other institutions are pursuing similar initiatives.

On August 12, Kyndryl announced a collaboration with India's Suryoday Small Finance Bank to deploy agentic capabilities intended to improve customer service, compliance and operational efficiency.

Kyndryl said it would establish an Agentic AI Center of Excellence to identify applications and support implementation.

Such arrangements suggest that financial institutions increasingly view agentic AI as an enterprise transformation initiative rather than an isolated technology experiment.

However, announced partnerships and product capabilities should not be confused with evidence of widespread, fully autonomous deployment.

Many institutions are still developing the governance, integrations and operational controls required to use agents in consequential financial processes.

The Financial Agent's Expanding Job Description

The potential applications of agentic AI span virtually every major financial services function.

In banking, agents can assist with account onboarding, identity verification, loan processing, regulatory compliance and customer inquiries.

In fraud prevention, they can investigate suspicious transactions by collecting information from multiple systems, comparing behavioral patterns and preparing findings for human investigators.

In corporate finance, agents can support budgeting, forecasting, financial reconciliation and reporting.

Investment firms can use agents to monitor portfolios, analyze securities, evaluate risks and prepare research.

Insurance companies can deploy them to review claims, retrieve policy information and coordinate administrative processes.

The distinguishing characteristic is the ability to connect activities that previously required employees to move information manually between applications.

Consider a corporate finance department closing its monthly books.

Traditional automation might reconcile transactions according to predetermined rules. Generative AI might explain discrepancies or draft commentary.

An agentic system could identify an unexplained discrepancy, retrieve supporting records, compare transactions, request missing documentation and propose corrective entries.

Human accountants would retain responsibility for approving consequential adjustments, but much of the investigative work could be automated.

In an October 8 interview with CFO Dive, Prophix CEO Alok Ajmera argued that AI could help finance departments expand their capabilities without proportionately increasing staffing.

He estimated that current AI applications can save finance employees four to eight hours weekly on certain activities.

The more ambitious opportunity involves extending these efficiencies from individual tasks to entire workflows.

This is where agentic AI potentially becomes transformative.

Rather than merely accelerating existing processes, it could change how financial organizations distribute work.

Rebuilding Financial Infrastructure Around Agents

Agentic AI could also alter the technological foundations of financial services.

For decades, financial institutions have invested in interconnected systems for accounting, trading, payments, customer relationship management and regulatory reporting.

These systems generally assume that humans initiate decisions and software executes instructions.

Agentic AI introduces a different operating model.

AI agents may become intermediaries connecting applications, interpreting information and coordinating financial activity.

In its September 23 report, Boston Consulting Group argued that transaction banking is approaching a transition toward increasingly automated corporate treasury operations.

Corporate treasurers traditionally manage liquidity, payments, currency exposure and financing through multiple banking platforms.

Agentic systems could coordinate these activities continuously, automatically selecting payment routes, monitoring cash positions and managing transactions within established corporate policies.

BCG suggested that banks may increasingly compete to control the technological layer that coordinates these activities.

That represents a significant strategic change.

A bank might continue providing deposits, credit and payment infrastructure while losing direct control of the customer interface to an independent AI platform.

In this environment, the institution executing a financial transaction may not be the organization that determines where the transaction goes.

Financial infrastructure would increasingly need to support machine-to-machine interactions, including secure application programming interfaces, standardized data access, digital identity verification and automated authorization.

The emergence of AI agents also intersects with developments in real-time payments, tokenized deposits and programmable money.

Together, these technologies could enable financial transactions to occur continuously, with limited human involvement.

Yet they introduce complicated questions about authorization, accountability and interoperability.

Who is responsible when an autonomous system initiates an incorrect payment?

How does a bank verify that an AI agent genuinely represents an account holder?

And what happens when different agents disagree about the terms of a transaction?

These are becoming infrastructure questions rather than theoretical concerns.

The Customer Experience Becomes Autonomous

For consumers, the most consequential change may be the emergence of AI systems that actively manage financial relationships.

Traditional banking applications allow customers to check balances, transfer money and monitor transactions.

AI-powered assistants can explain spending patterns or answer financial questions.

Agentic systems could eventually perform financial management tasks automatically.

A consumer might authorize an agent to monitor checking accounts, compare savings yields, identify unnecessary fees, manage recurring payments and recommend opportunities to reduce borrowing costs.

With additional permissions, the agent could execute some of those recommendations.

This would represent a shift from customers periodically managing their finances toward financial software managing routine activities continuously.

The possibilities extend to mortgages, insurance, retirement savings and investment portfolios.

An agent could compare insurance policies, identify refinancing opportunities or alert a customer when their financial circumstances change.

However, enthusiasm among technology executives does not necessarily translate into consumer confidence.

An October 2 Forrester Research analysis found that approximately seven in 10 consumers in the United States and United Kingdom worry that companies will not use AI responsibly.

Nearly one-third of consumers nevertheless reported using AI assistance for at least some personal finance questions.

That apparent contradiction illustrates the challenge facing financial institutions.

Consumers may appreciate convenient financial assistance while remaining reluctant to surrender control over their money.

The distinction between receiving a recommendation and authorizing an autonomous transaction is particularly important.

A mistaken chatbot response can be ignored. An incorrectly executed transfer may require a lengthy dispute or recovery process.

Consequently, successful financial agents will need more than sophisticated reasoning capabilities.

They will need transparency, reliability, understandable permissions and accessible human support.

Are AI Agents Coming for Financial Services Jobs?

Agentic AI's implications for employment are becoming increasingly difficult to dismiss.

Financial institutions employ substantial numbers of workers whose responsibilities involve collecting information, processing documents, reviewing transactions and coordinating administrative workflows.

Many of these activities are natural candidates for agentic automation.

Customer service representatives, loan processors, compliance analysts, operations specialists, junior researchers and accounting personnel could experience substantial changes in their responsibilities.

The effects may involve job elimination, slower hiring, changes in required skills or the reassignment of employees to more complex work.

Recent announcements suggest that workforce restructuring is already underway, although it would be misleading to attribute every AI-related reduction specifically to autonomous agents.

On October 6, Reuters reported that credit-scoring company FICO planned to reduce its workforce by approximately 15% as part of an AI-focused restructuring.

The reduction could affect roughly 570 employees.

The same week, Norway's largest bank, DNB, announced plans to eliminate approximately 400 positions in its technology and services operations as it increased investments in AI and digital capabilities.

These developments demonstrate the broader employment pressures associated with financial AI adoption.

Nevertheless, other industry executives anticipate a more gradual transformation.

Prophix's Ajmera suggested that AI could lead to slower hiring rather than wholesale workforce elimination.

Both outcomes are plausible.

Financial institutions may initially use agents to expand output without increasing headcount. Over time, however, sustained productivity improvements could reduce demand for particular categories of employees.

The nature of remaining jobs would also change.

Financial professionals could increasingly supervise automated workflows, investigate exceptions, verify AI-generated analyses and exercise judgment over consequential decisions.

Technical knowledge, regulatory expertise and the ability to evaluate AI outputs may become more valuable.

Conversely, employees whose primary contribution involves transferring information between systems could face growing displacement risks.

One longer-term concern involves entry-level employment.

If agents perform much of the routine research and administrative work traditionally assigned to junior employees, financial institutions may need new approaches to developing future executives, analysts and advisors.

The industry's challenge will not simply be replacing workers with software.

It will be preserving the expertise required to supervise increasingly capable machines.

Could Agents Destabilize the Financial System?

The potential consequences extend beyond individual institutions.

On October 1, ECB President Lagarde identified agentic AI as an emerging financial stability concern.

Her remarks focused partly on the possibility that financial institutions could deploy agents using similar underlying models.

If those agents interpret market developments similarly, they might independently execute comparable trading strategies.

During periods of stress, their actions could reinforce one another, accelerating price movements and amplifying volatility.

Lagarde also discussed research involving simulated AI traders that exhibited deceptive behavior or learned potentially collusive strategies.

These findings do not establish that autonomous agents are already destabilizing financial markets.

They demonstrate possible failure modes that deserve attention before adoption becomes widespread.

Indeed, Lagarde cited research indicating that only 5% of surveyed asset managers currently granted AI autonomous or semi-autonomous authority over investment recommendations or trading.

That suggests the financial industry remains in the early stages of agentic deployment.

Nevertheless, the risks could increase as autonomy expands.

A second concern involves banking liquidity.

Apollo economist Torsten Slok's September warning illustrated how autonomous financial management could undermine traditional deposit relationships.

Consumers frequently maintain substantial balances in checking accounts that pay little or no interest.

Banks benefit from these inexpensive deposits, using them to fund lending and other activities.

An AI agent instructed to maximize a household's interest income might automatically move excess balances into higher-yielding accounts or money-market funds.

If millions of agents performed similar actions, banks could experience substantial deposit outflows.

An October 7 Financial Times report examined the potential consequences, including estimates that greater competition for deposits could threaten hundreds of billions of dollars in banking-sector equity value.

Such figures represent scenarios rather than established losses.

Still, the underlying concern is credible.

AI could reduce the inertia that historically allows banks to retain customers despite offering relatively unattractive interest rates.

Importantly, moving money to obtain a better yield is not necessarily equivalent to a traditional bank run.

A bank run generally involves withdrawals motivated by concerns about institutional solvency or liquidity.

Routine yield optimization represents competitive deposit migration.

However, automated withdrawals could become destabilizing if they occur rapidly, particularly during periods of financial uncertainty.

The distinction matters, but so does the possibility that agents could dramatically accelerate both phenomena.

Financial Agents Introduce New Security Threats

Cybersecurity represents another significant challenge.

An AI assistant that summarizes information creates one category of risk.

An agent capable of accessing accounts, retrieving confidential records and initiating transactions creates another.

On August 27, the National Institute of Standards and Technology warned that organizations deploying agentic AI were sometimes prioritizing functionality and immediate returns over security.

NIST emphasized the importance of establishing strong identity and authorization frameworks for autonomous systems.

Financial institutions must determine not only whether an agent is legitimate but also precisely what it is authorized to do.

An agent might be permitted to review account information but prohibited from transferring funds.

Another might execute transactions below specified limits while requiring human approval for larger amounts.

Such distinctions require carefully designed permissions and reliable enforcement.

Prompt injection creates an additional vulnerability.

Malicious instructions embedded in documents, websites or messages could manipulate an agent into performing actions its user never intended.

An attacker might attempt to persuade an agent reviewing an invoice to change payment details or disclose confidential information.

Because agents interact with multiple applications, a successful compromise could spread beyond the original system.

The Center for Democracy and Technology's September report highlighted these concerns, including the possibility that agents could inadvertently bypass organizational information barriers.

The threat is not confined to poorly functioning agents.

Malicious actors can also employ autonomous systems.

On October 8, Reuters reported that cybersecurity investigators had linked attacks against South Korean banks to a suspected individual who allegedly used AI-agent tools to support the intrusions.

The case illustrates how autonomous capabilities could assist attackers as well as defenders.

Financial institutions will therefore need comprehensive agent inventories, tightly restricted permissions, transaction monitoring, audit trails and mechanisms for immediately disabling compromised systems.

The central security principle is straightforward: an AI agent should never possess more authority than its assigned task requires.

Will Agents Threaten Traditional Banking?

Perhaps the greatest commercial threat from agentic AI is disintermediation.

Banks have historically maintained valuable customer relationships partly because consumers interact directly with their branches, websites and mobile applications.

Agentic AI could insert an independent technological intermediary between customers and financial institutions.

Instead of visiting multiple banking websites, a consumer might instruct an AI assistant to identify the best savings account, cheapest loan or most appropriate payment service.

The agent could compare competing institutions and recommend—or eventually execute—the optimal transaction.

Under this model, financial institutions risk becoming interchangeable providers of underlying products.

Their brands, interfaces and established customer relationships may become less influential.

The consequences could be particularly significant for institutions that depend on customer inertia, opaque pricing or complicated switching processes.

Banks would have stronger incentives to compete on measurable product characteristics, including interest rates, fees and service reliability.

Yet the technology could also benefit established institutions.

Banks possess regulatory licenses, customer relationships, capital, transaction histories and infrastructure that AI developers generally lack.

They can develop proprietary agents, integrate them into existing services and use automation to reduce operating costs.

The competitive outcome is therefore uncertain.

Agentic AI could weaken traditional banking relationships while strengthening institutions capable of adapting.

The most vulnerable businesses may be those whose profitability depends heavily on administrative friction or customers' difficulty comparing alternatives.

Wealth Management Faces Its Own Agentic Revolution

Wealth management presents a particularly interesting case because the industry combines highly automatable administrative processes with services traditionally dependent on human judgment and relationships.

AI agents could assist advisors with portfolio monitoring, investment research, meeting preparation, compliance documentation, client communications and financial planning.

More sophisticated systems could identify planning opportunities, analyze tax implications and coordinate workflows across custodial, portfolio management and CRM platforms.

In September, LPL Financial published a discussion featuring its technology executive Greg Gates and Anthropic's Peter Nolan examining agentic AI and the future of financial advice.

The conversation highlighted LPL's Cyan technology and the potential for agents to execute multistep workflows while retaining human oversight.

The significance extends beyond productivity.

If advisors can delegate administrative responsibilities to AI agents, they may be able to serve more households without proportionately increasing support staff.

That could improve profitability and potentially make comprehensive financial advice available to investors who previously could not afford it.

However, agents could also compete directly with traditional advisors.

A consumer-facing financial agent might eventually monitor investments, recommend portfolio adjustments, compare investment products and coordinate routine financial decisions.

For relatively straightforward financial needs, such capabilities could reduce demand for certain advisory services.

The threat is particularly relevant to business models centered on investment selection, basic asset allocation and standardized financial planning.

These services are increasingly susceptible to automation.

More complex advisory relationships may prove resilient.

Estate planning, family financial conflicts, business succession, retirement transitions and emotionally difficult investment decisions involve considerations that extend beyond computational optimization.

Indeed, Goldman Sachs reportedly welcomed its largest class of new private wealth advisors in October, emphasizing the continuing importance of human relationships and specialized expertise despite expanding AI capabilities.

The likely outcome is not the disappearance of financial advisors.

It is a restructuring of what clients expect advisors to provide.

Advisors who primarily sell information or routine portfolio management may face growing competition from automated alternatives.

Those offering sophisticated judgment, accountability and personal relationships may find that AI increases their productivity and expands their opportunities.

The industry could also experience significant pricing pressure as automated capabilities reduce the cost of delivering financial services.

Ultimately, agentic AI may accelerate a transition already underway: from financial advice as a collection of products and transactions toward financial advice as a continuously delivered service.

The Question of Control

The proliferation of financial agents presents an unusual paradox.

The technology promises to give financial institutions and consumers greater control over financial activities by automating complicated processes and continuously optimizing decisions.

Yet achieving those benefits requires delegating some control to systems whose behavior may be difficult to predict or fully explain.

This creates important questions about responsibility.

When an AI agent makes an unsuitable investment recommendation, who is accountable?

When it executes an unauthorized transaction, who absorbs the loss?

When agents operating across multiple institutions contribute to a market disruption, how should regulators identify responsibility?

Existing financial regulations do not disappear simply because a machine performs an activity previously handled by a human.

Institutions remain responsible for complying with applicable lending, consumer protection, privacy, securities and anti-fraud requirements.

However, the emergence of autonomous decision-making could complicate how those responsibilities are enforced.

The challenge will be designing systems that combine operational autonomy with meaningful oversight.

That means financial institutions must be able to establish clear authorization boundaries, reconstruct decisions, investigate failures and intervene before automated mistakes become systemic problems.

The industry will also need to confront concentration risk.

If large numbers of financial institutions rely on a handful of AI model providers, vulnerabilities or outages affecting those providers could propagate throughout the financial system.

Lagarde emphasized precisely this concern in her October speech.

Financial resilience increasingly depends not only on the strength of individual institutions but also on the reliability of the technological infrastructure connecting them.

Here Come the Financial Agents

The financial services industry has repeatedly survived technological transformations that initially appeared threatening.

Automated teller machines changed retail banking. Electronic trading transformed securities markets. Online brokerages challenged established investment intermediaries. Robo-advisors introduced automated portfolio management.

Each innovation eliminated some activities, created others and altered the economics of financial services.

Agentic AI could follow a similar trajectory, although its reach may be considerably broader.

Unlike earlier technologies designed to automate specific activities, AI agents can potentially coordinate many different processes across organizational boundaries.

They could become financial researchers, administrative workers, transaction coordinators, compliance assistants and customer representatives.

Eventually, they may become financial decision-makers operating on behalf of institutions and consumers.

For financial services executives, the immediate challenge is distinguishing genuine operational value from technological hype while building the infrastructure and governance required for responsible deployment.

For financial professionals, the challenge is adapting to workplaces in which increasingly capable software performs tasks that once defined entire occupations.

For wealth managers, it is determining which aspects of financial advice remain valuable when machines can deliver research, monitoring and routine recommendations continuously.

And for regulators, it is ensuring that greater autonomy does not translate into diminished accountability or heightened systemic vulnerability.

Agentic AI has not yet replaced the traditional financial system. Nor has it demonstrated that fully autonomous financial decision-making is consistently reliable.

But developments between August and October 2026 suggest that the industry's relationship with artificial intelligence is changing.

The question is no longer simply how financial professionals will use AI.

Increasingly, it is how much financial work—and ultimately how much financial authority—they will delegate to it.

The financial agents are coming.

The industry's future may depend on deciding how much freedom to give them.

---

Researched by DWN Staff