Agentic AI spawns new roles

“Agentic AI” is turning from a software feature into a staffing question, as companies sketch a new job: the person who deploys, tunes, and audits autonomous agents across real workflows. (openai.com) In plain terms, an agent is software that does more than answer a prompt: it plans steps, calls tools, keeps state, and completes multi-step work. OpenAI’s Agents software development kit says agents can “plan, call tools, collaborate across specialists, and keep enough state” to finish jobs that span several actions. (developers.openai.com) That shift has moved the work from model demos to production operations. Google says Vertex AI Agent Builder is built to “deploy, manage, and scale AI agents in production,” with Agent Engine handling infrastructure, security, scaling, and monitoring. (cloud.google.com) The emerging role is less “prompt writer” than workflow owner. OpenAI now markets AgentKit as tooling to “build, deploy, and optimize agents,” while its observability docs emphasize traces of model calls, tool calls, handoffs, guardrails, and custom spans for every run. (openai.com, developers.openai.com) Microsoft is framing the same problem from the engineering side. Its Microsoft Agent Framework, released in public preview in October 2025, is an open-source software development kit and runtime for building, deploying, and managing multi-agent systems, with checkpointing and human-in-the-loop controls. (devblogs.microsoft.com, azure.microsoft.com) Amazon Web Services is making a similar pitch to executives. Its agentic artificial intelligence page says these systems are “digital teammates” that “plan, reason, and execute multi-step tasks,” and its 2025 guidance says autonomous agents are moving from chat interfaces to systems that act “in tandem with – or on behalf of humans.” (aws.amazon.com, aws.amazon.com) Consultancies are packaging that into enterprise services. Tech Mahindra launched TechM Orion in July 2025 as an agent development and deployment platform, then added the TechM Orion Marketplace on September 23, 2025, describing it as a global marketplace for “autonomous and action-oriented AI agents” that enterprises can centralize and scale. (pressreleasepoint.com, techmahindra.com) What stays human is becoming more specific. Google’s deployment docs focus on permissions, metrics, and deployment methods for automated pipelines, while Microsoft’s framework highlights middleware, memory, and approval checkpoints, suggesting that people are being pushed toward selecting workflows, setting limits, and reviewing outcomes rather than doing each step by hand. (docs.cloud.google.com, learn.microsoft.com) That is why the “agent deployer and manager” idea is spreading now: the tools are no longer just for prototypes, and the work now includes telemetry, governance, and return-on-investment tracking. The job opening is not for a single universal title yet, but the stack already exists for someone to own agents the way companies once assigned owners for websites, automation, or cloud systems. (docs.cloud.google.com, openai.com, cloud.google.com)