2026-06-22 · Week of 2026-06-15 → 2026-06-21

AI Hiring Pulse — Week of 2026-06-15

📊 Pattern of the week — the bank that out-hired the AI companies

Across 669 new AI roles at 183 companies this window, a bank is the story. JPMorgan Chase opened 31 AI roles — more than NVIDIA (15), OpenAI (12), or Apple (8). It now carries 276 active AI roles, up 36% from 203 at the start of June. That puts it ahead of Microsoft (272), Meta (109), and xAI (58) in raw open-role count.

This would be a footnote if JPMorgan were just buying seats on someone else's platform. It isn't. Twenty-four of its 31 new roles sit at the Agent station, and 22 of its active postings have "agent" or "agentic" explicitly in the title. Read them: "Agent Platform Engineer — Chief Data Analytics Office Fusion Platform," "Executive Director, Agentic AI Strategy — Open Banking," "AI/ML Engineer — Agentic Private Bank Engineer," "Lead Software Engineer — Agentic SDLC & AI Engineer." The seniority ladder is complete — from Software Engineer III up through VP and Executive Director. That is a production org, not a pilot.

The comparison that makes this legible is Salesforce. Both companies are overwhelmingly hiring at the Agent layer. Both posted 20+ roles this week. But they are building for opposite audiences.


🪞 Same layer, different logic — JPMorgan vs Salesforce

Salesforce's modal title is "Agentforce Forward Deployed Engineer" — repeated across markets (DACH, BNL, Arabic, French, Hebrew). It is hiring people to deploy agents at customers. The product is the agent platform itself. Salesforce has 109 active Agent-layer roles and 87% of its 125-role AI headcount sits at that single station.

JPMorgan's modal title is "Applied AI ML — Vice President" or "Lead Software Engineer — Agentic AI." It is hiring people to deploy agents on its own data. The teams span the entire bank: Fraud Technology, Quantitative Trading & Research, Card Data & Analytics, Branch Network Modeling, Consumer Business Modeling, LLM Suite Engineering. AI is diffusing into every business line simultaneously — the pattern that distinguishes production-stage adoption from a centralized lab.

The difference shows up in three structural signals:

Seniority. JPMorgan's AI roles are 34% VP/Director/Principal or above — ten times Google's rate (3%), nearly double Capital One's (19%), and more than double Salesforce's (16%). It has exactly two entry-level roles across 276 postings. This is "buy experienced talent to build now" hiring, not pipeline-building. Salesforce, by contrast, runs a large "AI Builder, Emerging Talent" cohort across regions — it is growing a junior bench to deploy a standardized product.

Guardrails and fine-tuning. Among Agent-layer roles, the tech-tag frequencies reveal what the agents are being wrapped in:

Tech signal JPMorgan (180 roles) Salesforce (109 roles) Capital One (64 roles)
Guardrails 28% 14% 44%
RAG 57% 43% 44%
Fine-tuning 30% 17% 39%
Observability 57% 61% 67%

JPMorgan and Capital One mention guardrails at two to three times Salesforce's rate. Fine-tuning at nearly double. The banks are not just plugging in a foundation model — they are wrapping it in compliance infrastructure and customizing it on proprietary data. A bank cannot send transaction histories through a third-party agent runtime. The data that makes the agent valuable is the same data it cannot share.

Applied scientists. JPMorgan has 43 "applied" archetype roles at the Agent layer — data scientists embedded in business units (Card Data & Analytics, Payments, Private Bank). Salesforce has 6. The bank is coupling agent engineering with domain scientists who understand the data the agents will operate on. Salesforce does not need to — its agents operate on the customer's data, not its own.


🚨 The broader picture — net −695

The index contracted this week: 669 opened, 1,364 closed, net −695 across 183 companies. Only four sectors posted net positive — auto (+4), healthcare (+1), and two niche categories. The largest decliners:

Company Opened Closed Net
Amazon 113 195 −82
Google 106 170 −64
JPMorgan Chase 31 80 −49
Microsoft 37 81 −44
Apple 8 40 −32
Capital One 8 33 −25
Monday.com 0 25 −25
NVIDIA 14 39 −25

The "zero-open" cohort is striking: Monday.com (−25), Uber (−23), Spotify (−14), xAI (−8), Netflix (−7), Scale AI (−5), and Intercom (−4) all closed AI roles with zero new openings. Monday.com now has zero active AI roles in the index.

Against that backdrop, JPMorgan's net is also negative (−49) — it is churning, not just expanding. But it is churning forward: 31 new roles, 24 of them explicitly agentic, while the AI-native companies go quiet.


🔮 Prediction watch

Grading last issue's predictions:

  • Compute-ownership split is structural. On a clean weekly window, Meta + NVIDIA + AMD posted 54% of new roles upstream (15 of 28). Salesforce + Adobe posted 12% upstream and 79% at Agent (27 of 34). The split held cleanly. Confirmed.
  • 🟡 Capital One stays the most upstream non-tech incumbent. Capital One's current upstream share is 62% (152 of 245 active roles) — well above the 40% floor. Still above every consumer-facing SaaS company. Three more pulses to go. On track.

New predictions:

  • JPMorgan's agentic build-out is structural, not a one-quarter burst. Prediction: over the next four pulses, JPMorgan's active AI role count stays above 250 and its Agent-layer share stays above 60%. Cashable 2026-07-20. If the agentic titles dry up, this was a seasonal hiring wave; if they persist, it's the new steady state.
  • The "data sovereignty" pattern repeats in pharma. Merck (8 opened), Johnson & Johnson (8 opened), and GE Healthcare (5 opened) all posted this week with roles spanning Data through Ship. Prediction: at least two of these three are net-positive over the next two pulses. Cashable 2026-07-06. If regulated-data incumbents keep hiring when tech contracts, the sovereignty thesis generalizes beyond banking.

📰 Reading list

  1. JPMorgan Chase in our index — read the 22 agentic titles against the Salesforce Agentforce Forward Deployed Engineer list. Same layer, opposite logic.
  2. Salesforce in our index — the AI Builder Emerging Talent cohort across 15+ markets is the clearest "selling the agent layer" signal in the data.
  3. Capital One in our index — still 62% upstream, still the counterexample. Now compare its guardrails and fine-tuning emphasis against Salesforce's.
  4. AI Hiring by Stage — the Agent station is now the single largest destination for new AI roles (378 of 669 this week, 56%). The question is no longer whether companies deploy agents — it's whether they build or buy.

← All issues