By ByteBot
Wall Street’s biggest banks are betting billions that AI can replace junior analysts. JPMorgan’s AI creates investment banking decks in 30 seconds—work that took junior teams hours. Goldman Sachs deployed AI assistants to all 46,500 employees. OpenAI hired 100+ ex-bankers at $150/hour to train AI for financial modeling. Meanwhile, AWS CEO Matt Garman calls this “the dumbest thing I’ve ever heard,” warning it kills talent pipelines. Developers are caught in the middle: 84% use AI coding tools, but only 29% trust them. Who’s right?
Wall Street’s Billion-Dollar AI Bet is Working
This isn’t theoretical. JPMorgan’s LLM Suite reached 200,000 users in eight months. Nearly half of JPMorgan employees use generative AI daily. Investment banking decks that took junior teams hours now take 30 seconds. Investment bankers automate 40% of research tasks—summarizing SEC filings, generating valuation models, building presentations.
Goldman Sachs rolled out its GS AI Assistant to every employee. Morgan Stanley followed. These aren’t pilot programs. They’re production deployments with measurable returns: JPMorgan’s AI-attributed benefits grew 30-40% year-over-year.
OpenAI’s Project Mercury makes the strategy explicit: 100+ former bankers from JPMorgan, Goldman, and Morgan Stanley hired at $150/hour to train AI for IPOs, restructurings, and leveraged buyouts. The math is simple: one AI-empowered analyst does the work of five. JPMorgan told investors operations staff would fall 10%+ over five years.
The ROI is real. Junior hiring has already dropped.
Tech CEOs Say This is Corporate Suicide
AWS CEO Matt Garman warned that eliminating junior roles is “the dumbest thing I’ve ever heard.” His reasoning cuts through the hype.
First, economics: junior staff are the least expensive employees and most engaged with AI tools. Cutting them makes no sense when they’re already using AI to multiply output.
Second, talent pipelines: “How’s that going to work when ten years in the future you have no one that has learned anything?” Companies that stop hiring juniors in 2025 won’t have senior talent in 2035.
Third, innovation: “We often find that’s where we get some of the best ideas—from junior people being mentored.” Fresh perspectives disappear when you hire only experienced workers.
Garman also challenged the metrics. Measuring AI value by code percentage is “silly” because “organizations can use AI to write infinitely more lines of code” but it could be bad code.
Wall Street optimizes for quarterly earnings. Tech CEOs think in decades.
Developer Reality: High Use, Low Trust, Real Impact
Stack Overflow’s 2025 Developer Survey captured the contradiction. 84% of developers use or plan to use AI tools, up from 76% in 2024. However, only 29% trust AI accuracy. Moreover, 46% actively distrust it, up from 31% last year. Only 3% “highly trust” AI output.
The frustration is specific: 66% cite “AI solutions that are almost right, but not quite.” Close enough to seem useful, wrong enough to require debugging. Furthermore, 45% find debugging AI code more time-consuming than writing it themselves.
This creates a bottleneck. AI generates a feature in two hours. Code review takes four. Teams wait 4+ days. The 5-10x increase in code output means review capacity, not generation speed, now defines development velocity.
The employment data tells the same story. Employment of software developers aged 22-25 fell nearly 20% between 2022 and 2025. Additionally, unemployment for ages 22-27 sits at 7.4%, nearly double the national average of 4.2%. Entry-level tech hiring dropped 25% year-over-year in 2024. When companies adopt generative AI, junior employment falls 9-10% within six quarters.
Developers use AI because it works. They don’t trust it because it doesn’t work reliably. And junior hiring is collapsing anyway.
Both Sides Are Right (And Wrong)
Wall Street is treating junior developers like Excel macros—automatable, replaceable, commoditized. Tech CEOs know AI is a productivity illusion needing human oversight. The reality is messier.
Wall Street is partially right. Repetitive financial analysis—Excel models, SEC summaries, deck building—are good AI candidates. Cost savings are measurable. One empowered analyst does work that required five. Short-term gains are proven.
Tech CEOs are partially right. AI coding creates new bottlenecks. Code review waits stretch to days. Quality issues are real—66% frustrated by “almost right” output. Talent pipeline risk is genuine. Infinite lines of bad code isn’t a win.
Developer data reveals what both miss: AI is a productivity multiplier, not a replacement. 84% use it because it provides value. 29% trust it because it’s not reliable unsupervised. The junior developer job isn’t dying—it’s transforming from “builder” to “validator and interpreter” of AI output. New skills: AI oversight, prompt engineering, code review at scale.
The role is evolving, not disappearing. But entry points are narrowing.
What This Means for Junior Developer Careers in 2026
Entry-level hiring is down 25%. Junior roles drop 9-10% within six quarters of AI adoption. Nevertheless, companies still need humans to review AI output—the four-day code review bottleneck proves it.
Don’t compete with AI on repetitive tasks. AI wins at boilerplate. Instead, focus on judgment and validation—the “almost right but not quite” problem requires humans. Develop AI fluency: prompt engineering, knowing when to trust AI, code review skills.
Seek mentorship aggressively. Fewer entry points make it harder to get in, but the opportunity exists: AI creates more need for senior oversight, not less. The code review bottleneck means job security for those who validate AI output at scale. Consequently, faster promotion paths await those who master it.
The software development market is still growing 20% annually to $61 billion by 2029. The work isn’t disappearing. The shape is changing. Those who adapt will thrive. Companies that eliminate juniors entirely will pay for it in 2030 when they have no senior talent pipeline.
