Goldman, JPMorgan, Citi, and Barclays are shrinking incoming analyst classes by as much as two-thirds while quietly redeploying survivors into AI-adjacent work on the same deals. The resume that worked in 2023 (deal count, model complexity, hours) is being read by an ATS that now scores whether you directed an AI agent on live diligence. Almost nobody in the current junior banker pool claims that they did.
The 2-in-3 cut is a redeployment, not a wipeout
Banks are firing juniors and rehiring the survivors into AI-supervisor roles inside the same institution. Fortune's June 2026 reporting shows bulge brackets shrinking analyst cohorts by up to two-thirds while sourcing roughly 62% of their new AI talent from those same cohorts. A Wall Street Oasis thread in May 2026 pinned summer class cuts near 30% with workload unchanged. The mechanism is McKinsey QuantumBlack's "analyst multiplier": one first-year supervising AI now produces what took three analysts.
Nobody at the top of a bank believes a junior analyst is being deleted. They believe two of every three are being converted into overhead for an LLM. John Waldron called Goldman staff a "human assembly line." Jamie Dimon said AI has let JPMorgan cut jobs by as much as 40% in some units. Jane Fraser said certain roles "will no longer be required." Bill Winters, at Standard Chartered, called it "replacing lower-value human capital."
The resume implication is not defensive. It is that the redeployment lane exists and almost no one is claiming a seat in it.
Only 16 of 11,134 investment banking analysts and associates in Refolk's index reference AI anywhere in their public history.
The 16-of-11,134 scarcity, in one table
In Refolk's index of US professionals currently holding an Investment Banking Analyst or Associate title, exactly 16 people out of 11,134 mention AI anywhere in their public headline or history. Python appears on 244 profiles. A credible AI-supervisor bullet, written specifically, puts a candidate in the top tenth of one percent of the visible pool.
| Slice | Count | % of total | Note |
|---|---|---|---|
| Total US IB Analysts + Associates | 11,134 | 100% | Baseline, Refolk's index |
| Listing Python as a skill | 244 | 2.2% | 244 / 11,134 |
| Mentioning "AI" in profile | 16 | 0.14% | Scarcity signal |
| Python-skilled to AI-mentioning | 15.3x | - | 244 / 16 |
| Junior finance hiring, 0-2 yrs, since early 2024 | -24% | - | Randstad, June 2026 |
| Analyst class cuts, top bulge bracket | up to -66% | - | Fortune, June 2026 |
The 15.3x gap between Python listers and AI mentioners is the tell. Python has been a resume line since 2019 and is now table stakes for 2.2% of the population. AI supervision, in June 2026, is being written on 16 resumes. That window closes fast. Within two cycles, "used Copilot" will look like "used Excel."
Fourteen of the top 25 Python-skilled IB analysts in Refolk's index sit in New York. Top employers of that group: Bank of America (8), Wells Fargo (5), then RBC Capital Markets, J.P. Morgan, William Blair, Perella Weinberg, UBS, and Centerview Partners. If you are targeting elite boutiques, that list is your comp set, not a hiring pipeline.
Why the AI-supervisor bullet is the new deal-count line
The single most valuable line on an investment banking analyst resume in 2026 is a specific, quantified instance of directing an AI agent on live deal work. Generic "AI-savvy" gets filtered as noise. Specific, measured, and dated gets flagged as the redeployment candidate.
Banks have already bought the tools and now need proof that a junior can operate them without burning a partner's Saturday. JPMorgan is spending roughly $18B on technology this year and has its in-house LLM Suite live for 200,000+ employees. Goldman rolled its GS AI Assistant firm-wide. Barclays has summarized more than 8 million customer calls with AI since launching in October. Kimi Work reportedly completes in 20 minutes what took junior analysts three days.
Compare the old and new versions of the same bullet:
| Old bullet (2023) | New bullet (2026) |
|---|---|
| Worked on 12 M&A transactions across TMT | Supervised Claude and Copilot on Q3 diligence for a $420M carve-out; cut model build from 14 hours to 4 |
| Built 3-statement models in Excel | Directed LLM agent to auto-populate comparable-company tearsheets across 47 targets; QA'd outputs, flagged 6 mispriced adds |
| Prepared pitchbooks for MD reviews | Ran pitchbook first drafts through GS AI Assistant; reduced turn cycles from 5 to 2 before MD sign-off |
| Fluent in Excel, PowerPoint, Bloomberg | LLM Suite (JPMorgan), Copilot, Claude, Python (pandas, numpy), Bloomberg, Capital IQ |
Notice what the new bullet does. It names the model. It names the deal type. It names the size. It names the time saved. It preserves the analyst as the decision-maker (supervised, directed, QA'd), not the tool user. That framing matters because Debasish Patnaik at McKinsey QuantumBlack put the surviving apprenticeship logic bluntly: senior judgment cannot be manufactured laterally. The bullet has to signal judgment plus throughput.
If you have not written a resume against a specific posting since 2024, this is where Refolk earns its keep: paste the JD, get your own history back rewritten to match its keyword field, with the AI-supervisor bullets moved to the top of the section the ATS scans first.
The window to claim an AI-supervisor bullet closes the moment 244 becomes the new 16.
What the ATS is actually scoring in 2026
Every bulge bracket and most elite boutiques now filter IB resumes through an AI-driven ATS before a human ever opens them. JPMorgan Chase IB applications flow through Oracle Cloud HCM. Goldman, Morgan Stanley, BofA, Citi, Evercore, Centerview, PJT, Moelis, and Lazard run a mix of Workday, Oracle Taleo, and Eightfold AI. The scoring is semantic, not keyword-count, but the semantic model was trained on job descriptions that now include AI-tool language.
What the ATS is looking for on a finance analyst resume AI tools section, in rough order of weight:
- Named LLMs (Copilot, Claude, ChatGPT, Gemini, LLM Suite, GS AI Assistant)
- The word "agent" or "agentic" attached to a verb you own (supervised, directed, deployed, QA'd)
- Python (pandas, numpy, and at least one of: SQL, VBA replacement, notebook workflows)
- Deal-adjacent verbs paired with AI nouns (diligence, tearsheets, model build, comparable analysis, memo drafting)
- Time-saved or throughput deltas with real numbers (hours, days, headcount equivalent)
- Bank-specific tool names when applying to that bank (LLM Suite for JPM, GS AI Assistant for GS)
Two failure modes are common. First, candidates list "AI/ML" as a bland skill line and score below someone who wrote a single specific bullet. Second, candidates write great AI bullets but bury them under a 2023-style deal list, so the ATS front-half of the resume looks identical to everyone else's.
The redeployment lane, and how to write for it
The highest-yield target for a cut junior banker is not another bulge bracket. It is the AI or data-ops team inside the same bank you just left, or the one that just called you back. Fortune's number, 62% of AI hires sourced internally from cut cohorts, is the entire thesis of the banking analyst job search in the second half of 2026.
Retraining a known analyst who understands deal fluency costs less than hiring an outside ML engineer who has never sat through a Sunday tearsheet cycle. Banks care about the fluency, not the ML depth. A first-year who can explain a DCF and prompt an agent to run 40 of them is worth more, to a bank, than a Kaggle grandmaster who cannot.
The resume version:
- Retitle the objective (if you use one) as "Investment Banking Analyst, AI-augmented deal execution"
- Move Python and named LLM tools above deal experience, not below
- Convert every "worked on" verb into "supervised," "directed," or "deployed"
- Add one line quantifying pod-equivalent throughput (for example, "output equivalent to a 3-analyst pod on Q2 diligence")
- Name the specific bank tool if you used it (LLM Suite, GS AI Assistant); a lot of ex-bank candidates leave this off out of habit
What junior finance hiring actually looks like right now
Junior finance hiring at 0-2 years is down 24% from early 2024 per Randstad, but deal activity is at record highs and the workload has not moved. The gap is being closed by tooling, and by the surviving juniors absorbing what the cut cohort used to do. That is the market you are applying into.
Randstad's June 2026 read. Deal volume is at record highs over the same window.
Named cuts to calibrate against:
- Morgan Stanley: roughly 2,500 jobs after 2025 revenue of $70.6B
- Citigroup: committed to cutting approximately 20,000 roles
- JPMorgan: Dimon says AI has enabled cuts of up to 40% in some units
- Goldman: linked 1,000+ role reductions to AI productivity
- Citigroup research: approximately 54% of US banking jobs could be automated
The counter-signal, and the reason this is not a doom piece: Patnaik's apprenticeship point. Banks will over-correct back inside 24 to 36 months because they cannot manufacture MDs laterally. The junior banker AI layoffs of 2026 create unusually thin senior competition by 2028-2029 for the cohort that survives with AI-multiplier proof. That is a two-year window to build the resume that matters.
The five-line rewrite, in order
The fastest useful edit to an investment banking analyst resume in 2026 is five lines, in this order:
- Headline swap. Replace "Investment Banking Analyst" with "Investment Banking Analyst, AI-augmented deal execution." You are now one of the 16.
- Tools line, above deals. "LLM Suite (JPMorgan), GS AI Assistant, Copilot, Claude, Python (pandas, numpy), SQL, Bloomberg, Capital IQ." Name the ones you used, not the ones you did not.
- One supervisor bullet per deal. For each transaction, add one line naming the model, the task, the deal size, and the time saved. If you cannot honestly write one, do not fake it: pick the deal where you came closest.
- Throughput metric. One line stating pod-equivalent output. Example: "Handled diligence workload historically staffed by 3 analysts, using Claude-driven doc review and Python-based comp screens."
- Redeployment target. If you are applying internally, add "Interested in AI-ops and data-augmentation roles within [Bank]" as the last line of your summary. It routes the ATS.
Do these five and the resume clears the 0.14% bar. Getting it done in an afternoon, tailored to every posting you send, is what Refolk was built for.
FAQ
Should I even stay in investment banking if analyst classes are being cut two-thirds?
Yes, if you can credibly claim AI-supervisor work on live deals. The 2-in-3 cut is a compression, not a shutdown: deal volume is at record highs and the surviving analysts are being paid to run more throughput per head. The candidates in trouble are the ones whose resumes still read like 2023. Patnaik's apprenticeship point suggests a rebound inside 36 months, which favors juniors who make it through this window.
What if I have not actually directed an AI agent on a deal?
Write the closest true version and stop there. If you used Copilot to accelerate a comps build, say that, with the hours saved. If you QA'd an LLM-drafted memo, say that. Do not invent an agentic workflow you did not run: bulge-bracket recruiters are already interviewing on this and will catch it in ten minutes. The 16-of-11,134 scarcity means even a modest, honest AI bullet outperforms silence.
Which banks are the best redeployment targets in 2026?
JPMorgan (LLM Suite deployed to 200,000+ employees), Goldman (GS AI Assistant firm-wide), and Barclays (8 million customer calls summarized) have the most mature internal AI stacks and therefore the most internal AI-ops roles. Elite boutiques hiring selectively per the Refolk index: Centerview, Perella Weinberg, William Blair, Evercore. Bank of America and Wells Fargo lead the Python-skilled IB analyst count.
How do I get through the Workday and Eightfold filters?
Match the JD's exact tool names in your skills line, put the AI-supervisor bullet in the first 200 words of the resume, and include Python with specific libraries. Semantic ATS models weight named entities heavily, so "Claude" beats "large language models" and "LLM Suite" beats "internal AI tools." Tailoring per posting matters more than a single polished master resume.