Who Does the Agent Work For?

Bob Hedges is a Digital Fellow at MIT’s Initiative for the Digital Economy and former Chief Data Officer at Visa. After four decades in leadership roles at retail financial services businesses and global strategy consultancies, Bob is today focused on research and advocacy work to establish the required transparency and consumer data empowerment for a trusted and successful digital economy.

With the accelerating adoption of agentic AI, the retail banking business is entering a period of unprecedented strategic challenges. Agentic AI has the potential to fundamentally restructure the customer experience, how data is collected and used, how consumers make and act on decisions, and to whom consumers’ loyalty is directed. Banking executives need to be clear-eyed and sober in understanding the challenges that this fast-evolving technology poses. 

Over the past 50 years, retail banking has survived waves of disintermediation threats. Past challenges attacked a product, a channel, or a margin. Each one fell short of its hype. Agentic AI is different — not because it attacks a banking function, and certainly not because of legitimate hype, but because it attacks the customer interface itself, and with it, control over the customer data that defines the relationship. As a first step in the strategy debate, retail banking executives need to clearly understand for whom the AI agent works. Spoiler alert — the AI agent does not work for the consumer.

Decisions by retail banking executives today regarding how they enable and support their customers’ interactions and data flows with AI platforms — with clarity regarding both the disintermediation challenges and consumer priorities — will define the future franchises that banks will or will not have.

Banks Have Seen This Movie Before 

For 50 years, retail bankers have faced strategic challenges that threatened to disintermediate banks from their customers. “Banks are dinosaurs” exclaimed Bill Gates in 1994 at a retail banking conference in Washington DC as he announced the Microsoft Money product. Reviewing history carefully is essential to understanding how today’s agentic AI threat is different, and profoundly more dangerous. 

Over the past 50 years, retail banking has faced five major strategic disruptions: 

1. Money markets challenged for deposits; but banks won on customer relationships (1971–1983)
   Merrill Lynch’s Cash Management Account drained more than $200 billion in low-cost deposits from banks within five years by offering market-rate yields that Reg Q-capped accounts could not match.¹ In 1982, with the passage of the Garn-St. Germain Act, banks succeeded in de-regulating deposit rates and established the right to offer their own market-rate money market deposit accounts. The challenge from money market funds was blunted. The primary customer interface and the data remained at banks throughout.²

2. Discount brokers captured the investment interface and began moving data outside bank walls (1975–1995)
   The SEC's 1975 deregulation of fixed brokerage commissions opened the door for discount brokers to strip the transactional activities of a consumer’s investment relationship away from traditional full-service banks and wire houses. Schwab and Fidelity built platforms where consumers formed investment habits beyond their traditional investment advisors.³ Transaction histories, holdings, and performance data began accumulating with third parties for the first time. Banks responded by acquiring brokerage capabilities, and by 2000, more than a dozen of the top 20 U.S. bank holding companies had retail brokerage businesses. The core banking relationship held, but a meaningful share of consumer financial data had migrated to a new home.

3. Internet banking players raced with traditional banks to build online businesses, but banks became digital and so did consumer financial data (1994–2005)
   Wells Fargo launched the first online banking service in 1994. By 1999, pundits were predicting that pure-play internet banks such as ING Direct, NetBank, and Wingspan would dominate the retail deposit market within a decade. The Harvard Business School case study, Fleet 2000, elegantly framed the strategic disintermediation risk faced by retail banks.⁴ The cover of Business Week on October 9, 2000 provocatively asked whether traditional banks could survive at all. In the end, none of the insurgent newcomers survived as independents.⁵⁶  The internet became a more efficient delivery channel for services that incumbents already provided, and today, 77% of consumers manage their accounts primarily through digital channels that banks built and own. Banks in parallel began generating a structured, machine-readable record of consumer financial behavior at scale for the first time.

4. Banks fumbled payments as a daily consumer touchpoint and the behavioral data that came with it (1999–2015)
   PayPal demonstrated that non-bank actors could own a meaningful segment of the financial services value chain without a charter, processing more than $280 billion in annual payment volume by 2015.⁷ Other start-up players followed — Venmo, Propel, Splitwise, Settle Up, Chime, Greenlight, and Wise. The traditional industry fought back with consortium-based responses, like Zelle, or direct commercial applications, like Popmoney. Banks lost transaction fee revenue but more consequentially lost the high-frequency daily interaction that payments represent. Consumers began evolving their primary financial routines and habits outside the bank interface, and payments behavioral data began flowing to platforms beyond bank control.

5. Fintechs attacked every function, but could not break the primary relationship; consumer data nevertheless kept leaking (2010–2022)
   Purpose-built fintechs targeted consumer credit, personal investing, brokerage, data aggregation, and merchant acquiring. None sought to displace the primary banking relationship. Banks retained the checking account, the direct deposit, and the regulatory standing that comes with deposit insurance. Consumer data, however, continued migrating through companies like Plaid to competitive platforms. The industry's defensive responses, including Akoya, the Financial Data Exchange, and Rule 1033 lobbying (and litigating) acknowledged the severity of the data vulnerability even as banks sought to contain it.

These five waves of disruption, and the focus of their impacts, are visualized in the below chart. 

Today’s agentic AI disruption has the potential to be the most holistic and invasive yet. As a result, banks’ ability to defend the primacy of their customer interface and data stewardship role has become more critical.

Why This Time Is Different

In the spring of 2026, something notable happened in American business. The Wall Street Journal reported that Starbucks, whose mobile app has long been held up as a gold standard of customer engagement, built a new presence inside ChatGPT’s platform. So did Lowe's, Wyndham Hotels, Little Caesars, and Zillow. Each company chose to build a presence inside ChatGPT rather than wait for consumers to come to their own AI-powered platforms.⁸

The past disintermediation waves that challenged retail banking attacked discrete functions performed by the bank. Agentic AI does not attack a specific, defined, function. It seeks to holistically replace the bank’s customer interface and become the core interface where consumers pose questions about their financial life. Moreover, agentic AI does not just replace the customer interface, it introduces a new party collecting and analyzing customers’ data and behavior. The two critical elements of banks’ enduring role with consumers that have resisted past disintermediation challenges are the primary target of agentic AI’s challenge. 

Five factors are different with today’s agentic AI challenge:

1. Friction, the silent protector of banking relationships, has been eliminated
   In the past, the inevitable consumer inertia produced by process complexity helped to defend customer relationships. In the near future, however, a consumer switching banks will no longer have to do paperwork, move direct deposits, or update their autopay instructions across multiple billers. The agent will. While some of the complexity of agent authentication and authorization remain to be ironed out, the friction will fade away. 

2. Consumers have already adopted and engaged with the new AI interfaces
   Previous disruptions required consumers to change behavior and learn a new platform. AI adoption requires no such transition. ChatGPT reached 910 million weekly active users by February 2026, and while smartphone adoption took seven years to reach 50% penetration, AI reached the same threshold in only three. OpenAI expects ChatGPT to become the primary way consumers interact with the key products and services in their personal and work lives.⁹ Banks will no longer be the starting point for a growing share of consumers’ financial questions.

3. AI agents are designed to act, not just advise
   Previous digital channels could present information, but required the consumer to complete a series of tasks to take any action. In contrast, when OpenAI launched Operator in January 2025, a user could photograph a grocery list and have the agent build the Instacart order, select a delivery window, and complete checkout without opening the Instacart app.¹⁰ The action layer that historically defined the bank’s operational relationship with the consumer is now available in a general-purpose AI interface. Retailers including Shopify, Walmart, and Target are already navigating the strategic choice of how much transaction control to cede to the platform. Banks do not have that option — a consumer who pursues a mortgage inquiry or refinancing discussion inside ChatGPT has conducted a financial interaction, and the data from it has been captured by the platform.

4. The data that defines the relationship is now readily accessible and migrating
   In every prior wave of disruption, primary account data (e.g., spending patterns, income cadence, credit behavior, savings trajectory) stayed with the bank. The AI agent changes this. A consumer discussing mortgage affordability, retirement timelines, or debt management with an AI platform is sharing precisely the behavioral and financial context that historically only a bank relationship manager would accumulate. When Wyndham Hotels built its ChatGPT app, it discovered it could not obtain basic customer usage metrics. When Little Caesars assessed its ChatGPT integration, it deliberately kept purchase transactions off the platform to preserve ownership of both the customer data and the relationship. If a pizza company and a hotel chain recognized the risk, retail banks need to take it very seriously.

5. Consumers are ready to switch
   Previous disruptions stalled partly because consumers were not prepared to trust non-bank actors with their primary financial relationships. That dynamic has shifted. Among Gen Z consumers, 77% report using AI for financial decision-making, and among Millennials the figure is 72%. These consumers will not return to a traditional bank interface for financial advice they can get from the same AI platform they already use for everything else. Moreover, after more than a decade of engaging with fintech product providers, these consumers are also unbounded from traditional product providers.

The risk of agentic AI compared to past disruption challenges can be seen in the below chart. 

Taken together, these five factors describe a disruption with no historical precedent in retail banking. The AI agent is already in the consumer’s hands, it is capable of acting, and it is already accumulating the data that historically belonged to the bank.

What Is Actually at Risk? Plenty

We can look at the potential agentic AI disintermediation risk along several critical dimensions — the nature of threat, the leading consumers adoption signals, the volume of deposit balances impacted, the net interest income impact, and the potential capital implications. Over the past 4 months, a wave of white papers by the leading financial services industry consulting firms — including McKinsey, Oliver Wyman, PwC, and Accenture — have examined Agentic AI disintermediation. The Federal Reserve Bank of Chicago has also published research. The accompanying chart presents recent major research studies on those topics, and each’s findings. 

The recent studies all come to the same conclusions — the Agent AI disruptors are focused on the customer interface; the consumer adoption risk is greater than in the past; bank deposits will move to higher-yielding providers; and bank profitability will be eroded. Across the leading consultants’ recent research studies, the “consensus estimate” calls for a 20% reduction in bank net interest margin. In parallel, the outflow of customer data to AI agents will significantly strengthen the disruptors’ ability to further expand their new customer relationships. We can consider each of the individual risk factors. 

The Consumer Adoption Risk Is Already High

Today, already 57% of bank customers, globally, say they would consider using a third-party gen AI financial agent if their banks don’t offer one.¹¹ When consumers are asked specifically what obstacles need to be overcome before they would delegate financial management to an AI agent, according to Visa’s research, the top concerns are data security (50%), privacy (44%), purchase accuracy (42%), reliability (40%), and loss of control (36%).¹² These obstacles are not permanent consumer attitudes. They are engineering and governance problems that can be solved. 

And despite these concerns, AI tools have the most compressed adoption curve of any technology in recorded history.¹³ Today, 99% of American adults use an AI-enabled product each week. Among Millennials, 61% are already using AI specifically to manage their finances.¹⁴

The Timeline Is Compressed

Last year, Tom Brown, the highly-regarded independent banking analyst from Second Curve Capital, described a near-term future in which a customer could configure an agent with simple parameters (i.e., maximize yield on cash reserves above $10,000, maintain FDIC insurance, minimize fees) and the agent would monitor rates continuously and move money automatically when the math makes sense. As Brown put it, “You could literally build it so that my free cash moves from Bank A to Bank B for five to ten basis points difference.”¹⁵ In this fast-approaching world, every consumer’s idle cash has become “hot money.”

AI agents could well become the default interface for financial decisions within three to five years. The AI platforms are already telegraphing this ambition with their strategic acquisition and partnership moves. OpenAI’s acquisition of Hiro and its partnership with Intuit point to AI infrastructure being embedded into financial relationships at scale.¹⁶ Reflecting the value that the AI platforms place on access to consumer financial data, Plaid has already become a consumer financial data source for major AI platforms including Perplexity, OpenAI, and Anthropic.¹⁷ Perhaps more significantly, on May 15, 2026, OpenAI announced a strategic partnership with Plaid to allow ChatGPT users to securely connect their financial accounts and receive personalized financial insights and guidance based on real transaction and balance data. 

Banks’ Financial Exposure Is Material

As AI agents become the primary interface for consumers’ financial decisions, approximately $23 trillion in retail deposits and investable assets are potentially in play. The vulnerable deposit and investable assets will be those held in low-yielding product categories (e.g., traditional savings deposit, low-yield money market accounts, and non-competitively priced CDs). The most vulnerable deposits will be those in banks lacking the granular deposit product systems and pricing models that enable complex and loyalty-inspiring pricing. Large national banks probably have made the requisite technology and analysis investments, while mid-sized regionals and community banks may not have.

Even a modest retail deposit share shift represents an existential revenue threat. Recent analysis suggests banks risk losing 20% or more of net interest income within three to five years if they do not respond effectively.¹⁸ Accenture finds that a relatively small disruption to loan and deposit rates could put a significant percentage of US banks’ pre-tax income at risk.¹⁹

All banks should model their agentic AI disintermediation risk. The analytical approach is reasonably straightforward and can leverage existing deposit management frameworks, incorporating practices like marketing analytics or funds transfer pricing. More specifically, banks should identify the key structural and behavioral attributes that could influence their customers’ deposit balance management actions, including digital engagement intensity, rate-paid sensitivity, balance tiers segments, tenure with the bank, depth of the relationship, which might be extrapolated from current “rate beta” modeling. Any impact analysis should include segment-specific migration curves that explicitly define the timing of varied migration patterns. Assume a 2-or-3-year adoption/migration curve where the deposit funds have migrated or been repriced. The agentic AI disintermediation math will vary by franchise composition and could well be alarming.

Deposit outflows and repricing will put pressure on many banks’ underlying profitability. For banks facing severe margin pressure, what starts as an existential strategic threat from AI could quickly become a profitability or liquidity issue. Community banks, whose margins are more dependent on low-cost deposits, may be particularly vulnerable to AI agents’ repricing of their retail deposit books. From the perspective of Will Callender, a partner at L.E.K. Consulting and leader in their banking practice, “AI agents have the potential to fundamentally alter the economics of retail banking. The erosion in funding capacity from deposit migration and the contraction in net-interest margin are the headlines, but agentic banking will also fundamentally reshape what businesses management views as profitable and worthy of investment.”

As AI agents begin addressing consumers’ financial questions, pursuing yield and cost comparisons, and making product recommendations, the bank's traditional central role in the consumer's financial life erodes. Given this ultimate relationship pressure, it is ironic that the current banking industry investment in AI is primarily for back-office efficiency while fintechs are embedding it in customer-facing experiences.²⁰ Banks’ overly narrow view of the “impact zone” for AI may be the ultimately myopic strategic choice. Now is the right time to step back to ask and assess the question, “who does the agent work for?”

How Banks Can Build Consumer Trust Relationships in the World of AI Agents

The pace of the current AI agent disruption is not accommodating of a deliberate “wait and see” approach. The banking industry may have been able to “wait out” the adoption of Microsoft’s Money product in the mid-90s, but that same approach will not work today with the AI platforms’ current ambitious agenda.

Banks’ franchises have long been protected and stabilized by the strong consumer trust that banks developed through generations of face-to-face relationships, the protection of federal deposit insurance, and a regulatory framework that holds them accountable to consumers in ways no technology platform has ever been required to match. At the same time, the fast-evolving role of AI has produced a trust dynamic where more than half of consumers are more concerned about AI than excited by it. Only 14% of US consumers say they meaningfully trust companies to use their data responsibly with AI. 64% believe companies benefit more from using their data than consumers do. 55% believe privacy policies exist to protect the company's legal interests, not to protect the consumer.

The retail banking players who move to address this unmet consumer need, and close this trust gap, will become the future trusted relationship players in this AI world. The bank that positions itself as the trusted steward of the consumer's data and financial identity — not the monetizer of it — will occupy a strategic position that other competitors cannot challenge.

Bank-Branded AI Agents: Technically Achievable, Strategically Fragile 

Building a bank-branded AI agent is certainly technically achievable. An AI agent that aggregates cross-institution data, maintains continuous consumer dialogue, and acts as the consumer’s primary financial interface would be enormously valuable. From an execution perspective, engineering capacity and data access are two important factors.

Ultimately, however, the most critical factor will be the bank’s level of consumer trust — and whether that consumer trust can be maintained when introducing a bank-branded AI agent. By building or sponsoring an AI agent that seeks to maximize cross-sell, optimize fee revenues, and potentially even monetizes customer behavioral data with third parties, the bank engages in the very conduct that leads to the suspicion of the AI platforms. 

Consumers are already suspicious that AI will be used to more aggressively monetize their relationship. A bank-branded AI model that optimizes its recommendations in ways that serve the bank’s revenue goals, above the consumer’s interests, will undercut the consumer trust relationships that it was developed to defend.

Past Visa research is unambiguous about what consumers want and cannot get: explicit control over data agents can access (85%), decisions made with their involvement (50% fear decisions made without them), and an agent they would trust enough to use even if they could not verify its recommendations in real time.²¹

By putting the consumer in control of their own data, a bank gives the consumer explicit decision rights over what data goes to which AI platforms and third-party AI players. By waving off the positioning of the bank as the lead data monetizer, the bank gains the strategic credibility and accountability to steward consumers’ data relationships with AI agents. The recent efforts by Consumer Reports to develop a Consumer Finance AI Standard start to detail the operating requirements for success on this front. 

Earning the trust of consumers to steward their data in a world of agentic AI is not a conceptual aspiration. It is achieved through a specific set of product and governance choices. Consumers prioritize addressing five specific needs to deliver against consumers’ desire for control of their data, anonymity in dealing with the AI platforms, and management of the context and history of their AI interactions. Executing against these specific needs, a bank can be a consumer-empowering data steward. 

Provide ID Shielding

Every interaction a consumer has with an AI agent shares device identifiers, behavioral fingerprints, and query metadata with the AI platform. These data signals are then used by the AI platform to build their own consumer profiles that the consumer neither sees nor controls. Banks that support the stripping out of their customers’ identifying signals, before those signals reach an AI platform, provide a structural privacy protection that no competing platform currently offers.

The consumer whose identity is protected while using their bank’s AI interface has a compelling reason to commit to their bank’s interface instead of migrating to a third-party alternative. This is not a marginal feature. It is the foundation for an enduring trust relationship.

Ensure Data Control and Consent

55% of US consumers believe privacy policies exist to protect the company, not the consumer.²² Terms of service can run to tens of thousands of words. Today’s data consent frameworks are designed to document consent, not to enable it. 

Consumers want control over how their data is used. Data control and consent, implemented as a stewardship capability, means that before any data is shared with an AI system or platform, the consumer receives a simple, plain-language summary of what is being shared, why, and what the benefit to them is, and seeks their approval. The consumer can approve or decline. It is what consumers will be looking for when every other actor is focused on monetization. The bank who enables this level of consumer data control and consent will establish a strongly differentiated position. 

Operationalize PII Deletion and Data De-identification

Consumers want to protect their personal information and thwart unwanted tracking and profiling. All Personally Identifiable Information (PII) is stripped from AI interactions before they are logged, stored, shared externally, or potentially used for model training.

To deliver against this consumer need, banks should implement PII deletion and data de-identification as a technical requirement, not an option, whenever a customer’s financial data is to be shared with a third-party. Protecting against the unwanted leakage of PII is the essence of data stewardship and a critical variable in maintaining consumer trust. 

Establish Cross-Platform Memory

One of the most underappreciated consequences of the agentic AI disruption is how assessing and understanding a consumer’s situation is evolving away from the traditional relationship manager. AI platforms are already developing their own storage or memory of a consumer’s financial interactions. Most consumers, however, want themselves to be the party with the strongest command of their history and context. No consumer wants to discover that their negotiating adversary has a superior understanding of their personal situation. 

Cross-platform memory, a consumer-controlled record of AI financial interactions that travels with the consumer across platforms, makes the consumer, not the platform, the owner of the consumer’s financial context. Banks who enable their customers to comprehensively manage and maintain the history and context of their financial situations will be delivering an immensely valuable service. 

Enable Intelligent Context

The ultimate power of agentic AI is its ability to analyze and reason across a consumer’s complete financial context with income, spending, debt, assets, and goals, providing genuinely personalized recommendations. Banks possess this data. 

Consumers want to be able to call upon their financial data and history when it helps them better understand their financial needs and potential solutions. Banks have the unique opportunity to be the player who automatically assembles the relevant data from the consumer's own accounts to improve the quality of the advice provided, with the consumer in control of what is shared and what is withheld. 

Consumers are clear about their desire to lever their data to better meet their needs. The choice banks need to make is do they want to enable consumers to manage their data to create the context for intelligent decision making, or is the short-term opportunity to monetize that data with third parties too great to resist? 

The strategic window to occupy this trusted steward position is open, but it will not stay open indefinitely.

The Question Every Bank Executive Must Consider

“Who does the agent work for?” 

The companies developing agentic AI tools to address consumer financial questions are not altruistic. They are assembling the most comprehensive consumer data infrastructure ever, and they intend to monetize it. The agent helping your customer compare mortgage rates, plan for retirement, or manage personal debt is not working for you or your customer. The agent is working for the platform that owns the interface, accumulates the data, and builds customer profiles for future marketing purposes. 

We are now three years into the hype and swirl of the potential for AI to transform the workplace, how we live our lives, and how consumers manage their financial affairs. The banking industry has made great strides in deploying AI for operational productivity gains across administrative, customer service, and software development functions. 

While 50 years of disintermediation waves have failed to displace the primary banking relationship, the current wave of agentic AI transformation is profoundly different. The agentic AI transition is the first disruption in which the two historical advantages of banks — primary interface and data access — are genuinely at risk.

So, banks now have a choice. While there is a short-term route where banks can cede data and AI accountability to third parties, and for sure capture some near-term monetization opportunities, the longer-term bank play is to serve as the consumer’s steward, supporting the management of their data and advancing their interests in the agentic AI environment. The trust-building bank position is to be the enabler and steward of consumers’ AI agent relationships and the data that fuels them. 

The bank that chooses the stewardship approach is not making an abstract “values” statement. It is making a strategic bet that in a world where consumers can choose their financial partners and agents, they will choose the one they trust. And trust, in the world of agentic AI, will be earned by whoever best can use their position to make AI agents work for the consumer by putting the consumer in control of their data when other players seek to monetize it for their own economic purposes. Forward-leaning banks should take the necessary steps to act on behalf of customers, invest in the required data governance and control capabilities that engender consumer trust, and by doing so, ensure a more transparent and better functioning marketplace for all. 

The opinions shared in this article are the author’s own and do not reflect the views of any organization they are affiliated with.

^[1]  Depository Institutions Deregulation and Monetary Control Act of 1980, Pub. L. No. 96-221, 94 Stat. 132; Garn–St. Germain Depository Institutions Act of 1982, Pub. L. No. 97-320, 96 Stat. 1469.

^[2]  Securities and Exchange Commission. “Regulation of Securities Markets: Fixed Rate Commissions.” Effective May 1, 1975.

^[3]  Wolff, Edward N. “The Asset Price Meltdown and the Wealth of the Middle Class.” State of Working America 2014. Economic Policy Institute. Table 6.9, “Share of Households Owning Stock, 1989–2010.” https://www.epi.org.

^[4]  Frances X. Frei and Hanna Rodriguez-Farrar, FleetBoston Financial: Online Banking, Harvard Business School Case No. 601-042 (Boston: Harvard Business School Publishing, 2000; rev. 2002).

^[5] “Bank One Pulls Plug on Wingspan.com.” American Banker, October 2000. Note: Wingspan was a subsidiary of Bank One Corporation, discontinued after failing to achieve independent viability.

^[6]  American Bankers Association. “Bank Customers Continue to Increase Use of Mobile Banking Apps.” News release, November 12, 2024. https://www.aba.com/about-us/press-room/press-releases/consumer-survey-banking-methods-2024.

^[7]  OpenAI. “Usage and Revenue Milestones.” OpenAI Blog, January 2026. https://openai.com.

^[8]  Bousquette, Isabelle. ”The ChatGPT-ification of American Business.” Wall Street Journal, May 6, 2026. https://www.wsj.com/cio-journal/the-chatgpt-ification-of-american-business-1d332c1f.

^[9]  OpenAI. “Introducing Operator.” OpenAI Blog, January 23, 2025. https://openai.com.

^[10]  Forrester Research. Predictions 2026: Banking and Investing. Forrester Predictions series, 2025.

^[11]  McKinsey & Company. “Global Banking Annual Review 2025: Why Precision, Not Heft, Defines the Future of Banking.” McKinsey & Company, October 23, 2025. https://www.mckinsey.com/industries/financial-services/our-insights/global-banking-annual-review.

^[12]  “2025 Agentic Commerce Consumer Research.” Visa, 2025.

^[13]  TD Bank. 2025 TD AI Insights Report. TD Bank Group, June 17, 2025. Business Wire. https://www.businesswire.com/news/home/20250617041073/en.

^[14]  “Americans Use AI in Everyday Products Without Realizing It.” Gallup News, January 15, 2025. https://news.gallup.com/poll/654905/americans-everyday-products-without-realizing.aspx.

^[15]  “From Copilots to Closers: How Agentic AI Accelerates Competitive Pressure.” Blend.com, November 19, 2025. https://blend.com/blog/thought-leadership/how-agentic-ai-banking-accelerates-competitive-pressure/. Note: Tom Brown, founder of Second Curve Capital, quoted therein.

^[16] Perplexity. “Computer Is Now Your Personal CFO.” Perplexity Blog, April 9, 2026. https://www.perplexity.ai/hub/blog/perplexity-finance.

^[17] “Connect Your Plaid Integration to Agents Built with OpenAI.” Plaid Blog, May 21, 2025. Plaid. “The Next Phase of Developer Tools with Anthropic and Plaid.” Plaid Blog, May 22, 2025.

^[18] “Banking Top Trends 2026: Unconstrained Banking—A New Age of Possibility.” Accenture, 2026. https://www.accenture.com/us-en/insights/banking/accenture-banking-trends-2026. Scenario modeled: 5% erosion of lending margins and 15% erosion of deposit margins produces a $19 billion quarterly decline in US banks’ net interest income, representing a 22% decrease in overall pre-tax income.

^[19] “Banking 2026: Unconstrained Banking—A New Age of Possibility.” 2026. PDF report.

^[20] “OpenAI and Intuit Announce Multi-Year Partnership.” OpenAI press release, 2026.

^[21] Visa. Consumer Empowerment Study / Consumer Trust Study 2019-2025.

^[22] McKinsey & Company. “How Gen AI Agents Threaten Retail Banks’ Customer Relationships.” Op. cit., footnote 19. Visa. “2025 Agentic Commerce Consumer Research.”

Open Banker curates and shares policy perspectives in the evolving landscape of financial services for free.

If an idea matters, you’ll find it here. If you find an idea here, it matters. 

Interested in contributing to Open Banker? Send us an email at [email protected].