Agent hiring surges: 65% of AI jobs now agentic roles

Get Ready For Agents, a research and index platform, has published live job, governance, and agent-tool tracking data as of July 2026. The headline finding is a wholesale shift of AI hiring toward agent-focused roles. The Jobs Index identifies 2,071 open agentic AI positions across 216 companies in a tracked field of 261—representing 65% of all monitored AI job postings. This is not a niche trend: it is the dominant category of AI hiring.

Databricks is the largest hirer, with 134 open agentic roles, followed by Salesforce (119), OpenAI (97), Palantir (68), Adobe (63), Citi (61), LangChain (57), and Okta (52). Across all these roles, Python dominates the skill demand (1,025 mentions), followed by cloud infrastructure (752), RAG (568), MCP—Model Context Protocol, a standard for AI tool integration (470), Kubernetes (395), LangChain (388), multi-agent systems (365), TypeScript (329), prompt engineering (290), and LangGraph (277). At frontier AI labs—OpenAI, Anthropic, and similar—68% of all AI roles are agentic; in enterprise hiring, the figure is 61%. Australia shows an even higher skew: 69% of its 55 tracked AI roles are agentic, across 28 companies. Only 9% of agentic roles are listed as remote; agent teams are overwhelmingly being built in-office.

Yet production reality diverges sharply from hiring velocity. Industry surveys cited in the report find that 62% of organisations are at least experimenting with AI agents, but only 23% are actively scaling them. Of those deploying agents, just 39% report measurable impact on EBIT. The forecast is blunt: over 40% of agentic AI projects will be scrapped by end of 2027, driven by escalating costs, unclear business value, and inadequate risk controls. The governance picture is stark. Only one in five organisations has a mature governance model for autonomous AI agents. The Get Ready For Agents Governance Index tracks 35 AI governance frameworks in force across 16 countries and 4 regions. Among teams building agents, 89% have implemented observability, but only 52% have implemented evaluations—leaving a 37-point gap between watching and testing.

Where agents have succeeded at scale, the numbers are large. Klarna's AI assistant handled 2.3 million customer conversations in its first month, cutting average resolution time from 11 minutes to under 2. JPMorgan's COiN platform reviews 12,000 commercial credit agreements in seconds—work that previously consumed roughly 360,000 lawyer-hours annually. Salesforce's Agentforce resolves around 85% of incoming customer service requests autonomously on Salesforce's own help portal. Morgan Stanley's DevGen.AI reviewed 9 million lines of legacy code, saving an estimated 280,000 developer-hours. These are self-reported figures by the companies named. They illustrate the upside, but remain exceptions in an industry where most projects are still in experimentation or are forecast to fail. The data is published by Get Ready For Agents under CC BY 4.0 and updated automatically from live job postings and official regulator and standards-body sources, making it citable and verifiable.

The data reveals a hiring surge in agent-focused AI roles that outpaces the industry's ability to govern, evaluate, and scale them. As of July 2026, agent roles account for 65% of all tracked AI hiring—a dramatic shift toward autonomous systems. Yet the same surveys that report near-universal experimentation (62% of organisations) show a steep drop-off in actual production deployment and measurable value: only 23% are actively scaling, and just 39% report quantifiable impact on profitability. This adoption-to-value gap is accompanied by a stark governance deficit. Of 3,164 tracked AI job postings, only 40 are for governance or security roles—a 79-to-1 imbalance in favour of builders over guardrails. Even among teams already running agents in production, the control gap is stark: 89% have implemented observability (monitoring what agents do), but only 52% have implemented evaluations (testing whether they work correctly)—a 37-point gap that suggests many teams can see what their agents are doing but lack confidence in whether they are doing it right.

The forecast that over 40% of agentic projects will be scrapped by end of 2027 is a direct consequence of this mismatch. Escalating costs, unclear business value, and inadequate risk controls are not abstract concerns—they are predictable failures when hiring, building, and deployment vastly outrun governance, testing, and business accountability. The few named production successes (Klarna's 2.3 million conversations in one month, JPMorgan's 360,000 lawyer-hours saved, Morgan Stanley's 280,000 developer-hours preserved) demonstrate the upside when agents are well-scoped and controlled, but they remain exceptions. For most organisations, the gap between excitement and execution—and between building and controlling—remains the central constraint.

No comments yet. Be the first to share your thoughts!