The $20 Billion Gamble: JPMorgan’s “Long-Running” AI Agents Are Coming – And They Will Work While You Sleep

 

 The $20 Billion Gamble: JPMorgan’s “Long-Running” AI Agents Are Coming – And They Will Work While You Sleep

**Subtitle:** *From 20% sales boosts to 50% larger client loads, the bank is proving that autonomous agents are finally safe for corporate use. Here is why “hours-long” agents are a bigger deal than ChatGPT.*

**Reading Time:** 8 Minutes | **Category:** Artificial Intelligence & Finance

## Introduction: The Two-Hour Workday

For the past two years, the AI conversation has been dominated by chatbots. You type a prompt. You get an answer. The interaction lasts seconds.

Later this year, that model is going to change dramatically.

JPMorgan Chase, the largest bank in the United States by assets, plans to deploy AI agents capable of operating autonomously for **hours at a time**, according to Derek Waldron, the bank’s chief analytics officer . “We’ve entered now the era of long-running autonomous agents,” Waldron told CNBC. “That means that agents don’t just run for two or three minutes to carry out a goal or some instructions of a human, they can run for an hour or two” .

This is not a research project. It is not a pilot. It is a deployment – scheduled for 2026 – that signals a fundamental shift in how Corporate America thinks about AI . The technology has finally cleared the security and governance hurdles that have historically kept it out of large organizations .

The numbers are compelling. In private banking, AI systems that analyze overnight market data, client holdings, and research reports have already driven a **20% increase in gross sales** . Waldron expects the technology could eventually enable each banker to serve a client base roughly **50% larger** than what is currently manageable .

But the story is not just about efficiency. It is about a redefinition of work itself. “For enterprises to win with AI, it’s not about cutting the maximum number of jobs,” Waldron said. “It’s all about trying to create a sustainable competitive advantage” .

In this deep-dive, we will break down the “intellectual coherence” that makes long-running agents possible, explain why 2026 is the year of deployment, and reveal the “build vs. buy” shift that threatens traditional software vendors.

## Part 1: The “Intellectual Coherence” Breakthrough – Why Hours-Long Agents Are Finally Possible

The difference between a two-minute agent and a two-hour agent is not just about time. It is about **quality**.

### The “Team Manager” Shift

Waldron coined a term for the capability that makes long-running agents possible: **“intellectual coherence.”** It describes whether a model can sustain productive, independent operation over an extended period without losing the thread of its task .

Recent improvements in AI reasoning have pushed systems away from the role of solo executor and toward something closer to a supervisory function. “Just like how people function, team managers can parse out a problem and delegate activities, and teams can run for a lot longer to do more complex things,” Waldron explained .

This is not a linear improvement. It is a phase change. A model that can maintain coherence for hours can shift from being a “doer” to being a **“manager.”** It can break down complex problems, delegate subtasks to specialized sub-agents, and synthesize the results – all without human intervention.

### The Enabling Technologies

Three specific advances have made this possible :

| Technology | What It Enables |

| :--- | :--- |

| **Code Generation** | Agents can write, test, and deploy software autonomously |

| **Browser Navigation** | Agents can interact with web-based tools and data sources |

| **Direct Desktop Interaction** | Agents can operate existing enterprise software directly |

These capabilities mean that an agent is no longer confined to a single application or a single data source. It can operate across the entire digital workplace, coordinating workflows that span multiple software environments .

### The “Guardrail” Question

None of this would be possible without solving the security and governance puzzle. For years, large organizations have cited safety concerns as the primary barrier to deploying autonomous agents .

JPMorgan’s willingness to deploy long-running agents in 2026 suggests that these hurdles are finally being cleared. “We will have those in 2026,” Waldron said .

However, the bank is not moving recklessly. “It’s the regulations that constrain us, because we have to be very careful with what we do,” said Adolfo Lopez, senior vice president of corporate technology at JPMorgan . The bank is balancing innovation with its fiduciary duty to protect client assets and data.

**The Human Touch:** For the JPMorgan employee, the arrival of long-running agents is a double-edged sword. The agents will handle the “grunt work” – sifting through data, drafting reports, routing tasks. But they will also change the nature of the job. The human’s role will shift from “doer” to “supervisor.” That requires new skills, new training, and a new mindset.

## Part 2: The $20 Billion Engine – Why JPMorgan Can Lead Where Others Follow

JPMorgan is not the only bank experimenting with AI. But it is the one with the deepest pockets.

### The Technology Budget

JPMorgan spends nearly **$20 billion annually** on technology . That is more than the GDP of many small countries. It is a war chest that allows the bank to invest in AI projects that would be unaffordable for competitors.

### The “Build vs. Buy” Shift

One of the most significant – and least reported – implications of JPMorgan’s AI push is its effect on the software industry. Waldron said that internal development has become a more attractive option, with the bank scrutinizing more carefully whether vendor solutions are truly necessary .

“The moat around certain types of software companies is most certainly diminished versus where it was in the past,” Waldron said .

This is a threat to traditional software vendors. If JPMorgan – and other large enterprises – can build their own AI agents to handle tasks previously outsourced to third-party software, the market for those vendors could shrink.

### The Revenue Impact

The private banking example is the most concrete evidence of AI’s value. The bank’s AI systems now analyze overnight market data, client holdings, and research reports, allowing bankers to focus on client relationships .

The result: a **20% rise in gross sales** .

This is not a cost-saving measure. It is a revenue-generating measure. And that changes the calculus entirely. “The key to succeeding with AI isn’t about cutting the most jobs – it’s about building sustainable competitive advantage,” Waldron said .

| Metric | Current Impact | Projected Impact |

| :--- | :--- | :--- |

| **Annual Tech Budget** | $20 billion | Stable |

| **Private Banking Gross Sales** | +20% | Ongoing |

| **Client Coverage per Banker** | Baseline | +50% (expected) |

*Sources: *

**The Human Touch:** For the software vendor, the “build vs. buy” shift is an existential threat. For the JPMorgan shareholder, it is a reason for optimism. The bank is reducing its dependency on third-party suppliers, which improves margins and reduces risk.

## Part 3: The Timeline – From 2026 to “Weeks”

The deployment of long-running agents is not the end of the road. It is the beginning.

### The 2026 Milestone

JPMorgan plans to deploy long-running agents capable of operating for **one to two hours** later this year .

These agents will be used internally, not for customer-facing applications. The bank will start with use cases that have clear governance guardrails and limited downside risk.

### The Progression

Waldron is explicit about the trajectory. Over time, agents will be capable of running coherently for “multiple hours, then days, then weeks” .

That is a roadmap. Each stage requires advances in reasoning, memory, and error recovery. But the direction is clear. The bank is not stopping at two-hour agents.

### The External Threat

Waldron noted that long-running agents have already emerged in the wild over the past year. Examples include Anthropic’s Claude Code (an autonomous coding agent) and OpenClaw (a general-purpose agent framework) .

These external examples demonstrate that the technology is mature enough for widespread adoption. JPMorgan’s internal deployment is the validation that security and governance concerns have been addressed.

**The Human Touch:** For the technology executive watching from a competitor, the message is clear: the time to act is now. If JPMorgan can deploy long-running agents in 2026, so can you. The question is whether you will be a leader or a follower.

## Part 4: The “Agentic” Ecosystem – 80 Services and Counting

JPMorgan is not starting from scratch. AI agents are already embedded across the bank.

### The Existing Footprint

According to IT Brew, AI agents are currently being used in the workflow of **80 services** across JPMorgan . These range from fraud detection to risk management to customer service.

### The “Assistant” Phase

At present, most of these agents operate in an “assistant or delegative” mode, according to Adolfo Lopez, senior vice president of corporate technology . They are not fully autonomous. They require human oversight and approval for critical decisions.

The move to “long-running autonomous agents” is a step change. It moves agents from the role of “assistant” to the role of “employee.”

### The Citigroup Comparison

JPMorgan is not the only bank pursuing agentic AI. Citibank has rolled out an internal AI platform called **Arc** that lets employees create their own AI agents . Wells Fargo partnered with Google in 2025 to build up its agentic AI capabilities .

But JPMorgan’s scale – and its $20 billion technology budget – gives it a significant advantage.

**The Human Touch:** For the JPMorgan employee, the expansion of AI agents is not a future threat. It is a present reality. The agents are already in their workflow. The question is whether they will be seen as a help or a hindrance.

## Part 5: The “Competitive Advantage” Thesis – Why This Isn’t Just About Layoffs

The most common narrative about AI is that it will replace jobs. Waldron argues that this is the wrong framework.

### The Growth Thesis

“For enterprises to win with AI, it’s not about cutting the maximum number of jobs,” Waldron said. “It’s all about trying to create a sustainable competitive advantage” .

The 20% increase in private banking sales is evidence of this thesis. The AI is not just reducing costs. It is generating revenue.

### The Client Coverage Multiplier

Waldron expects that AI could eventually enable each banker to serve a client base roughly **50% larger** than what is currently manageable .

That is not a layoff. It is a force multiplier. The banker does the same work – or more – with the same number of human hours. The AI handles the data gathering, the research, and the routine analysis. The banker focuses on the relationship.

### The “American Idol” and “Salesforce” Analogy

Two other analogies have emerged in the AI discourse. One compares AI to **American Idol** – a selection mechanism that identifies the best human performers . The other compares AI to **Salesforce** – a platform that augments rather than replaces human sales efforts.

Waldron’s thesis aligns with the latter. AI is a tool. It is not a replacement. But it is a tool that changes the nature of work.

### The Dimon Acknowledgment

CEO Jamie Dimon has been transparent about the likely impact. He has said that AI will eliminate certain roles . But he has also indicated that the bank intends to offer retraining and redeployment pathways for workers whose positions are affected .

| AI Impact | Waldron’s View | Dimon’s View |

| :--- | :--- | :--- |

| **Job Elimination** | Not the primary goal | Acknowledged as likely |

| **Growth Driver** | Primary goal (20% sales increase) | Supportive |

| **Employee Impact** | Force multiplier (50% larger client loads) | Retraining and redeployment pathways |

*Sources: *

**The Human Touch:** For the JPMorgan employee worried about their job, Waldron’s message is both reassuring and unsettling. The bank is not trying to cut jobs – but jobs will change. The banker who can use AI as a force multiplier will thrive. The banker who cannot will fall behind.

## Frequently Asked Questions (FAQ)

**Q: When will JPMorgan deploy long-running AI agents?**

A: JPMorgan plans to deploy long-running AI agents capable of operating autonomously for one to two hours **later in 2026** .

**Q: How much does JPMorgan spend on technology annually?**

A: JPMorgan spends nearly **$20 billion annually** on technology, making it one of the largest corporate tech spenders in the world .

**Q: What is “intellectual coherence”?**

A: “Intellectual coherence” is a term coined by JPMorgan’s Derek Waldron to describe whether an AI model can sustain productive, independent operation over an extended period without losing the thread of its task .

**Q: What is the difference between current AI agents and long-running agents?**

A: Current AI agents typically complete single tasks in two or three minutes. Long-running agents can operate for one to two hours, coordinating complex workflows across multiple software environments .

**Q: Has JPMorgan already seen results from AI?**

A: Yes. In private banking, AI systems that analyze overnight market data, client holdings, and research reports have driven a **20% increase in gross sales** .

**Q: Will these AI agents replace human workers?**

A: JPMorgan CEO Jamie Dimon has acknowledged that AI will eliminate certain roles. However, the bank’s stated strategy is to use AI as a growth driver, not a cost-cutting tool. The bank also intends to offer retraining and redeployment pathways for affected workers .

**Q: What is the “build vs. buy” shift?**

A: JPMorgan is increasingly building its own AI solutions rather than buying from third-party software vendors. This could pressure traditional software companies as enterprises internalize capabilities that were previously outsourced .

## Conclusion: The Era of the “Digital Employee”

We started this article with a number: two hours. That is how long JPMorgan’s new AI agents will be able to run autonomously.

We end with a different number: **20%** . That is how much private banking sales have increased as a result of AI tools.

The era of the “digital employee” is not coming. It is here. JPMorgan is deploying AI agents that can work for hours without human intervention. They will handle tasks that once required armies of analysts. They will allow bankers to serve 50% more clients. And they will change the nature of work for everyone in the organization.

**For the Investor:**

JPMorgan’s AI push is a reason to be optimistic. The bank is using technology to drive revenue, not just cut costs. The $20 billion annual tech budget is a moat that competitors cannot easily cross.

**For the Employee:**

The arrival of long-running agents is a signal to adapt. The jobs that remain will require skills that AI cannot easily replicate: judgment, creativity, and relationship management. The banker who masters AI will thrive. The banker who resists will not.

**For the Competitor:**

JPMorgan is moving fast. If you are not already planning your own AI agent deployment, you are falling behind. The technology is ready. The governance hurdles are being cleared. 2026 is the year to act.

**The Bottom Line:**

JPMorgan Chase is deploying long-running AI agents that can work for hours without human intervention. The bank has the budget, the data, and the talent to lead. And it is betting that AI will be a growth driver, not a cost cutter.

The era of the digital employee has begun. And JPMorgan is writing the playbook.

---

**#JPMorgan #AI #AgenticAI #LongRunningAgents #Banking #TechBudget #DigitalWorkforce**

---

*Disclaimer: This article is for informational purposes only. It does not constitute financial advice. Deployment timelines and features are subject to change.*