The AI hiring wave isn’t just about data scientists and ML engineers anymore. It’s fragmenting into highly specialized roles built around agentic systems, autonomous workflows, and enterprise AI deployment, disciplines that barely had names in 2023.
LinkedIn’s 2026 Labor Market Report shows AI has added more than 1.3 million new roles globally. The growth isn’t slowing. It’s branching into more specific functions every quarter, and the talent pools for most of these roles are still tiny.
For hiring managers and business leaders, the challenge is real: you know you need AI talent, but the job descriptions you’re writing are for roles with no established career paths, no salary benchmarks, and no obvious place to source from.
Here’s the full picture, organized by department, of what’s actually being hired for right now.
Engineering & AI Systems
These are the builders. The roles in this department own the design, deployment, and reliability of AI systems in production. They’re the hardest to hire and the most in demand.
1. AI Agent Engineer
First seen: 2024 | Demand growth (2025): +240%
The fastest-growing role in the entire AI hiring market. An AI Agent Engineer designs and builds autonomous AI systems, software agents that can take multi-step actions, use tools, call APIs, and complete complex tasks without a human approving every decision.
If a standard LLM answers a question, an AI agent completes a workflow. Booking a flight, triaging a support queue, researching a market. The engineer who builds those agents is an AI Agent Engineer.
What they do:
- Design and deploy autonomous agents using frameworks like LangGraph, AutoGen, or CrewAI
- Build the orchestration logic that lets agents chain actions and recover from failures
- Integrate agents with external tools, APIs, and data sources
Per Second Talent’s 2026 New AI Roles Report, AI Agent Engineer postings grew 240% in 2025 alone and accounted for 38% of all AI placements alongside AI Automation Engineers. If you’re trying to hire one, here’s how UltraTalent approaches hard-to-fill AI roles.
2. AI Agent Architect / Agentic Systems Architect
First seen: 2023-2024 | Demand growth (2023-2024): +985%
A step above the Agent Engineer. Where the engineer builds individual agents, the Architect designs the entire multi-agent system: how agents communicate, how they delegate tasks, how failures cascade, and how the system scales.
Agentic AI job listings grew 985% between 2023 and 2024, going from statistically negligible volume in Q3 2023 to over 11,000 postings by Q3 2024, per the AI Talent Report. That growth rate is why sourcing for this role requires a fundamentally different approach than traditional technical hiring – more on that here.
What they do:
- Design multi-agent architectures and define how agents interact and hand off tasks
- Select and evaluate orchestration frameworks for production use cases
- Own the technical blueprint for enterprise agentic deployments
Before this role existed: The work lived inside “ML Engineer,” “Solutions Architect,” or “Platform Engineer” job descriptions. Nobody owned multi-agent system design as a dedicated discipline.
3. Context Engineer
First seen: 2025 | Status: Didn’t exist as a recognized title before 2025
Context Engineer is so new that it didn’t appear as a recognized job title until 2025, when Anthropic started using it internally before it spread to other frontier labs and enterprise AI teams.
The role exists because of a fundamental truth about how LLMs work: the output is only as good as the input. Context Engineers design the information architecture that feeds a model, deciding what context the model receives, how it’s structured, and how to make the most of the model’s context window.
What they do:
- Design retrieval pipelines that surface the right information at the right moment
- Optimize how context is chunked, ranked, and passed into model interactions
- Build memory systems that help agents maintain coherent long-horizon tasks
“Context Engineers determine whether an agent succeeds or fails on multi-step work. They’re the architects of what the model sees.” — Forbes, June 2026
Warren Schilpzand, Regional Director at SambaNova Systems, put it plainly on the UltraTalent GTM Insights podcast: “The secret isn’t re-training — it’s context engineering. You feed agents the right history, the right policy docs, the right prior decisions. Change the context, and you change the behaviour — without another expensive training cycle.”
Senior Context Engineers in the US are pulling $650K+ total compensation at frontier labs, per 2026 market data.
4. LLMOps Engineer
First seen: 2023-2024 | The analogy: DevOps, but for language models
If DevOps brought engineering discipline to software deployment, LLMOps does the same for AI models. The LLMOps Engineer owns the full lifecycle of a language model in production: shipping prompts, running evals, managing model versions, and maintaining CI/CD pipelines built specifically for AI systems.
What they do:
- Build and maintain CI/CD pipelines for prompt and model version management
- Monitor model performance in production and trigger retraining or rollback when needed
- Manage inference infrastructure including latency, cost, and reliability
What separates LLMOps from standard MLOps: Traditional MLOps deals with deterministic models optimizing a metric like accuracy. LLMOps deals with open-ended outputs where “correctness” is subjective, latency is measured in seconds rather than milliseconds, and a single bad response can make the news. The operational discipline is genuinely different.
5. Forward Deployed Engineer
First seen: 2023 (Palantir) | Demand growth (2024-2025): +800%
Palantir invented this role. OpenAI, Anthropic, Google, and EY are all hiring for it now. The Forward Deployed Engineer embeds directly inside a customer’s organization to build and tune AI workflows for their specific tech stack, processes, and use cases.
It’s part engineer, part consultant, part implementation specialist. The FDE doesn’t just hand over software. They live inside the customer’s environment until the AI system is actually working.
What they do:
- Build and customize agentic workflows tailored to a specific client’s infrastructure
- Act as the technical bridge between what the AI product can do and what the customer needs
- Accelerate time-to-value for enterprise AI deployments
Why it’s hard to hire: You need someone with strong engineering credentials who also has client-facing instincts and can operate with minimal structure. Most engineers don’t want to live on-site at client offices. Most consultants can’t write the code. The overlap is small.
Quality, Safety & Governance
As AI agents take on real-world tasks with real consequences, a whole discipline has emerged around making sure those agents don’t cause harm, drift off-target, or violate regulations. These roles didn’t exist in any formal sense before 2024.
6. AI Evals Engineer
First seen: 2024 | Before that: A bullet point inside ML Engineer job descriptions
Every time a model gets updated, its behavior changes. Sometimes better, sometimes worse, often in ways that are subtle and hard to detect without systematic testing. The AI Evals Engineer builds the test suites and evaluation frameworks that catch hallucinations, regressions, and bias before they reach a customer.
This is a permanent discipline, not a one-time QA pass. Models are updated continuously, and each update requires a new round of evals.
What they do:
- Design and maintain evaluation benchmarks for model outputs
- Build automated pipelines that flag regressions across model versions
- Work with safety and product teams to define what “good” actually looks like
The hiring challenge: Most engineers have never worked on probabilistic systems. Evals Engineers need a hybrid of ML knowledge, software engineering rigor, and near-obsessive attention to edge cases.
7. AI Red Team Engineer
First seen: 2024 | Demand growth (2024-2025): +340%
The name comes from military wargaming, where a “red team” simulates the adversary to expose weaknesses before a real enemy does. In AI, the Red Team Engineer stress-tests AI systems by trying to break them, jailbreak them, and expose unsafe behaviors before they reach users.
This role has gone from niche to essential as AI agents gain the ability to take real-world actions. An agent that can send emails, execute code, or make purchases is an agent that can cause serious damage if someone figures out how to manipulate it.
What they do:
- Run adversarial testing to find prompt injection vulnerabilities and jailbreaks
- Simulate misuse scenarios and document failure modes
- Work with safety and engineering teams to patch vulnerabilities before deployment
In regulated industries like finance, healthcare, and government, AI Red Team Engineers are becoming mandatory hires. The EU AI Act and a growing patchwork of US state laws are creating compliance requirements that make systematic adversarial testing a legal necessity, not just a best practice.
8. Agent QA Lead
First seen: 2025-2026 | Status: Newly formalized in 2026
Traditional QA assumed deterministic systems: same input, same output. Agent QA is fundamentally different. The same prompt can produce different outputs across runs, across model versions, and across context states. That requires a completely different testing discipline.
The Agent QA Lead owns regression testing for agent behavior, voice agent rollouts, and LLM-based systems at scale. This is the role that catches the edge cases before a customer does.
What they do:
- Design test suites for non-deterministic agent behavior across scenarios
- Own regression testing when model versions or prompts change
- Specialize in voice agent and multi-modal agent quality assurance
Before this role existed: QA engineers handled this informally, or it fell to the engineering team entirely. Neither approach scales once agents are handling thousands of customer interactions daily.
9. AI Governance Analyst / AI Compliance Specialist
First seen: 2024-2025 | Accelerating in: 2026
As regulation tightens, companies need someone who sits at the intersection of AI capability and legal exposure. The AI Governance Analyst owns the frameworks that ensure AI systems operate within regulatory, ethical, and organizational boundaries.
What they do:
- Build and maintain AI governance frameworks aligned to regulations like the EU AI Act
- Audit AI systems for bias, fairness, and compliance with internal policies
- Work with legal, risk, and engineering teams to document model decisions and accountability chains
Before this role existed: Governance lived in legal, compliance, or risk functions. Nobody owned AI-specific governance as a dedicated discipline, and most companies still don’t have one. That’s changing fast as regulators start asking questions.
Operations & Workflow
Once AI agents are running in production, someone has to manage them. This department covers the roles that own agent operations, workflow design, and the increasingly complex task of supervising autonomous systems at scale.
10. Agent Supervisor / AI Agent Manager
First seen: Late 2025 | Formalized by: Salesforce and others in 2026
Multiple 2026 analyses describe “Agent Supervisor” as a role that “didn’t exist six months ago.” Salesforce formalized it in 2026 as agentic AI moved from experimental to operational across enterprise teams.
The Agent Supervisor oversees a fleet of AI agents the way a team lead oversees a team of people: monitoring output quality, handling escalations, adjusting agent behavior, and ensuring agents are actually completing their tasks correctly.
What they do:
- Monitor agent performance across tasks and flag anomalies or failure patterns
- Handle escalations when agents encounter situations outside their defined scope
- Tune agent behavior and update instructions based on real-world performance data
Before this role existed: Oversight of AI outputs was informal, often described as “AI wrangler” work handled by whoever was nearby. It wasn’t a named position with defined responsibilities. As agent fleets scale into the hundreds, informal oversight breaks down fast.
11. AI Orchestrator / Agentic Workflow Designer
First seen: 2024-2025 | Scaling in: 2026
The AI Orchestrator designs the end-to-end workflows that connect multiple agents, tools, and human touchpoints into a coherent process. Think of them as the process architect for agentic systems: they decide which agent handles which step, how handoffs work, and where guardrails need to sit.
What they do:
- Design multi-agent workflow sequences and define decision logic between steps
- Configure tool use, API integrations, and guardrails within orchestration frameworks
- Optimize workflows for reliability, cost, and latency
Before this role existed: Similar work lived inside “Integration Engineer,” “Automation Engineer,” or “Solutions Architect” titles, but none of those roles were explicitly scoped to agentic orchestration. The complexity of multi-agent systems created enough demand to justify a dedicated function.
Navin Persaud, a RevOps leader featured on the RevOps Lab podcast, described exactly where this role is heading: “The RevOps leader who learns to be a conductor of those systems will outperform the one trying to replace the instruments. Hiring go-to-market engineers who understand both GTM mechanics and where AI fits is the talent move that separates teams that get left behind from those that don’t.”
12. Agentic Workflow Business Analyst
First seen: 2025-2026 | Also titled: Business Analyst (Agentic Workflows), Agentic AI Business Analyst
This is a business-facing role, not a technical one. The Agentic Workflow Business Analyst sits at the intersection of domain expertise and AI delivery. They understand how a department actually works, map where agentic automation creates genuine leverage, and translate business requirements into specifications that AI teams can build against.
What they do:
- Map existing business processes and identify where agentic automation applies
- Define requirements for agent workflows with enough technical precision to be buildable
- Measure the impact of deployed agent workflows against business outcomes
Why this role is new: Companies had “AI Business Analysts” and “Process Analysts” before, but not roles explicitly scoped to multi-step, autonomous workflows. The agentic qualifier matters because the analysis required is fundamentally different: you’re not just automating a task, you’re designing a system that makes decisions.
Product & Strategy
AI is changing what product teams build and how they make decisions. The roles in this department sit at the intersection of business strategy and AI capability, and they require a genuinely different profile from traditional product or strategy hires.
13. AI Product Manager (Agentic)
First seen: 2023 (general AI PM) | Agentic flavor: 2024-2025 | Job posting growth (2024-2025): +89.7%
Every product is becoming an AI product. But the AI PM role has evolved further: the Agentic AI Product Manager specifically owns products built around task-completing agents, not just AI features.
They run probabilistic product roadmaps where features don’t have binary pass/fail states but probability distributions of quality. They own evals as a product metric. They understand the difference between a capability that works 95% of the time and one that works 99.9% of the time, and they know which one is shippable.
What they do:
- Define product requirements for agentic features with a genuine understanding of model capabilities and limitations
- Own evaluation metrics and agent economics (cost-per-task, success rate, latency) as core product KPIs
- Navigate trade-offs between model performance, reliability, and cost
Autodesk’s 2025 AI Jobs Report ranks AI Product Manager as one of the ten fastest-growing AI roles, with nearly 90% year-over-year posting growth. UltraTalent places AI PMs and C-suite AI leaders – see how we work.
What makes this different from a standard PM role: A traditional PM can learn the domain. An Agentic AI PM needs to understand probabilistic outputs, hallucination risk, agent failure modes, and model versioning at a level that drives real technical decisions. That cognitive profile is rare.
14. Chief AI Officer (CAIO)
First seen: 2023 | Job posting growth (2024-2025): +44.4%
A few years ago, “Chief AI Officer” would have sounded like a conference keynote title. In 2026, it’s a real C-suite hire across financial services, healthcare, retail, and professional services.
The CAIO owns enterprise-wide AI strategy and implementation. They sit at the intersection of technology, ethics, risk, and business value, translating what AI can do into decisions about where to invest, which use cases to prioritize, and how to govern AI systems responsibly.
What they do:
- Set the company’s AI strategy and investment priorities
- Own AI governance including risk management, compliance, and ethical guidelines
- Bridge the gap between technical AI teams and the executive leadership team
Why this isn’t just a renamed CTO role: A CTO optimizes existing systems. A CAIO is redesigning how the business operates using AI as the primary lever. The scope, risk profile, and stakeholder management are different. A CAIO regularly presents to the board on model risk and regulatory exposure, not just system uptime.
The talent pool is thin. Most candidates are either strong technologists who lack business credibility or senior executives who lack enough technical depth to make real AI decisions. Finding someone who has both is the hard part.
People, Change & Workforce
AI transformation doesn’t fail at the technology layer. It fails at the people layer. The roles in this department exist because buying AI tools and actually changing how an organization operates are completely different problems.
15. AI Enablement Lead
First seen: 2024 | Also called: AI Adoption Lead, Agent Coach (McKinsey’s term)
This is the role most companies don’t realize they need until they’ve already wasted six months and a significant AI budget on tools that nobody uses.
The AI Enablement Lead runs internal adoption programs. They map where AI can actually change how work gets done, train employees to use the tools effectively, and track whether any of it is sticking. McKinsey calls this the “agent coach,” and it’s one of the roles they’ve flagged as critical to enterprise AI ROI.
What they do:
- Audit workflows to identify where AI automation creates genuine leverage
- Design and run employee training programs for AI tools and agentic systems
- Build measurement frameworks to track adoption and actual productivity impact
Who fills this role: Change management professionals who’ve upskilled into AI, or AI engineers who’ve discovered they’re better at enabling others than building systems. The former is more common; the latter is more effective.
16. Human-Agent Workforce Lead / AI Workforce Planner
First seen: 2025-2026 | Status: Emerging as a formal function in 2026
This is one of the newest roles on this list, and one of the most strategically important. As companies deploy agents at scale, they’re discovering that agent capacity needs to be planned the same way headcount is planned: forecasted, allocated, monitored, and adjusted.
The Human-Agent Workforce Lead owns that planning function. They model how human and agent capacity interact, decide which tasks should be agent-handled versus human-handled, and manage the ongoing rebalancing as both agent capabilities and business needs evolve.
What they do:
- Build workforce models that treat agent capacity as a plannable resource alongside human headcount
- Design human-agent handoff protocols for tasks that require escalation or judgment
- Track agent utilization, failure rates, and cost-per-task as workforce metrics
Before this role existed: Workforce planning was entirely human-focused. Agents were experimental tools, not a capacity class. As companies run hundreds of agents in parallel across business functions, that framing no longer holds.
Sales, Marketing & GTM
AI is hitting revenue teams hard, and not just through automation. It’s creating entirely new role profiles for the people who work in GTM functions. These aren’t technical roles, but they require a fundamentally different way of working.
17. AI-Augmented GTM Specialist
First seen: 2024 | McKinsey’s take: The largest new AI job category by headcount
This one is different from everything else on this list because it isn’t a technical role. McKinsey identifies AI-augmented sales and GTM as the largest single category of new AI-driven employment, and it’s easy to see why.
The AI-Augmented GTM Specialist is a seller, marketer, or customer success professional who has fundamentally changed how they work by integrating AI into every part of their workflow. Research, outreach, qualification, follow-up, forecasting: agents handle the repetitive parts, and the human focuses on the decisions and relationships that actually require a person.
What they do:
- Use AI agents to automate research, personalization, and follow-up at scale
- Interpret AI-generated insights to make better pipeline and territory decisions
- Operate with a much smaller supporting team because AI handles what used to require additional headcount
Why companies are hiring for this specifically: The gap between a seller who has integrated AI into their workflow and one who hasn’t is widening fast. A rep using AI agents for research and outreach can cover territory that used to require three people. Companies that hire for this profile are building a structural cost and productivity advantage over those that don’t.
18. AI Content Strategist / Agentic Content Lead
First seen: 2024-2025 | Status: Rapidly formalizing in 2026
Marketing teams are deploying AI agents to produce content at scale. Someone has to own the strategy, quality standards, and brand consistency for everything those agents produce. That’s the AI Content Strategist.
This isn’t a copywriter who uses AI tools. It’s a strategist who designs the workflows, prompt architecture, and editorial frameworks that govern AI-generated content across channels.
What they do:
- Define content strategy and quality standards for AI-generated output
- Build and maintain prompt libraries and editorial guardrails for content agents
- Measure content performance and tune agent workflows based on results
Before this role existed, content strategy and AI tooling were separate functions. As AI agents handle more of the production workload, the person who governs what those agents produce becomes a critical hire for any marketing team operating at scale.
Data & Analytics
AI is reshaping data and analytics functions too, creating roles that sit between traditional data science and the new agentic stack.
19. AI Data Curator / Training Data Engineer
First seen: 2023-2024 | Why it exists: Models are only as good as what they’re trained on
Fine-tuning and retrieval-augmented generation have created a new discipline around the data that feeds AI systems. The AI Data Curator owns the quality, structure, and relevance of that data, whether it’s training sets for fine-tuning, knowledge bases for RAG pipelines, or ground truth datasets for evals.
What they do:
- Build and maintain high-quality datasets for model fine-tuning and evaluation
- Design knowledge base architecture for retrieval-augmented generation systems
- Audit data for bias, gaps, and quality issues that affect model performance
Before this role existed: Data engineering handled data pipelines for analytics. Nobody owned the specific problem of curating data for AI model performance, which requires a different set of judgments about quality, coverage, and representativeness.
20. AI Insights Analyst / Agentic Analytics Lead
First seen: 2025-2026 | Status: Emerging as agents generate their own data trails
As AI agents handle more business processes, they generate enormous amounts of operational data: task completion rates, failure patterns, cost-per-action, latency distributions. Someone has to make sense of that data and turn it into decisions.
The AI Insights Analyst owns the analytics layer for agentic systems, tracking how agents are performing, where they’re failing, and what the operational data says about how to improve them.
What they do:
- Build dashboards and reporting frameworks for agent operational performance
- Analyze agent behavior data to identify failure patterns and optimization opportunities
- Translate agent performance data into business-level insights for leadership
Why this is new: Traditional analytics roles track human behavior or system performance. Tracking agent behavior requires understanding non-deterministic systems, probabilistic outputs, and the specific failure modes of LLM-based systems. That’s a different analytical discipline, and most analytics teams don’t have it yet.
The Hiring Reality Across All 20 Roles
Here’s where things stand across the full market:
| Department | Role | First Seen | Hiring Difficulty |
|---|---|---|---|
| Engineering | AI Agent Engineer | 2024 | Very high |
| Engineering | AI Agent Architect | 2023-2024 | Very high |
| Engineering | Context Engineer | 2025 | Extremely high |
| Engineering | LLMOps Engineer | 2023-2024 | High |
| Engineering | Forward Deployed Engineer | 2023 | Very high |
| Quality & Safety | AI Evals Engineer | 2024 | High |
| Quality & Safety | AI Red Team Engineer | 2024 | Very high |
| Quality & Safety | Agent QA Lead | 2025-2026 | High |
| Quality & Safety | AI Governance Analyst | 2024-2025 | Moderate-High |
| Operations | Agent Supervisor | Late 2025 | Moderate-High |
| Operations | AI Orchestrator | 2024-2025 | High |
| Operations | Agentic Workflow BA | 2025-2026 | Moderate |
| Product & Strategy | AI Product Manager (Agentic) | 2024-2025 | High |
| Product & Strategy | Chief AI Officer | 2023 | Extremely high |
| People & Workforce | AI Enablement Lead | 2024 | Moderate |
| People & Workforce | Human-Agent Workforce Lead | 2025-2026 | Moderate-High |
| GTM | AI-Augmented GTM Specialist | 2024 | Moderate |
| GTM | AI Content Strategist | 2024-2025 | Moderate |
| Data & Analytics | AI Data Curator | 2023-2024 | High |
| Data & Analytics | AI Insights Analyst | 2025-2026 | Moderate-High |
The common thread across all twenty: none of these roles have established career pipelines yet. There are no university programs producing Context Engineers or Agent Supervisors. Most people holding these titles got there by building skills on the job, often at frontier labs, early-stage AI companies, or inside large enterprises that were early movers.
That’s what makes sourcing for these roles so different from traditional technical hiring. You can’t post on LinkedIn and wait. The people who can do this work are rarely actively looking, and the ones who are get multiple offers fast.
The organizations winning the AI talent race right now are the ones that started sourcing six months before they needed to hire. If you’re already behind on that timeline, the gap closes faster with a recruiter who knows where these people actually are.
If you’re building out your AI team and running into walls, we’d be glad to help. AI talent is one of the fastest-moving areas we work in, and the sourcing approaches that worked 18 months ago no longer work. You can also read more on why AI hiring bottlenecks slow revenue and how we approach executive AI search.
