The templates that rank for this query are interchangeable bullet lists — “set AI vision,” “lead the team,” “ensure ethical use.” None of them name an outcome, a decision boundary, or a 90-day bar. Below is a full template you can adapt, plus the four things it specifies that the generic versions don't: what the role owns, what it explicitly does not own, which requirements actually predict success, and what to pay by stage. Data as of July 2026.
A usable Head of AI job description names the outcome the role is accountable for, states explicit decision rights, lists what the role does not own, and sets a 90/180-day success bar — not a bullet list of AI buzzwords copied from a template site.
Here is the full template. The reasoning behind each non-obvious line follows below.
Head of AI — job description
Role summary
Own the company's AI agent program end to end: prioritize which workflows get automated, decide build vs. buy vs. hire for each one, and get at least one agent from pilot into production against a named, measurable target this year.
Reports to
CEO or CTO (growth-stage). CEO, with a standing coordination line to the CISO and CCO (enterprise).
Owns
Does not own
Requirements
90-day bar
Audit of every AI pilot in flight, one workflow selected with a binary success criterion and an owner-signed budget, automated evaluation coverage in place before any further rollout.
180-day bar
The selected workflow is in production with the metric hit, missed, or the project killed with a documented reason. Decision-rights lines with the CISO and CCO are written down and signed. A second workflow is scoped.
Compensation band
$250k–$400k base at growth-stage; $400k–$1M+ at enterprise (McKelvey). See the pay section below for the full breakdown and sourcing.
The role owns decisions, not tasks: which workflows get automated and in what order, build vs. buy vs. hire, evaluation design, vendor selection within budget, and — the one line that matters most — the pilot-to-production call against a named metric.
The three top-ranking templates all describe this loosely as “overseeing deployment of AI models into production environments” (Digital Waffle) or “lead and manage the AI research and development team” (Lucent Search). Neither names a metric or a deadline. That gap matters: among AI agent pilots that actually reach production, 94% have a named “agent owner” with budget authority and a measurable target outcome (Digital Applied, April 2026). Write the ownership line as a metric and a deadline, not a verb like “oversee.”
It should explicitly exclude AI security and threat modeling (CISO), regulatory and compliance mapping (CCO), and core data infrastructure (data engineering). Naming these exclusions prevents the turf disputes that stall pilots before they ship.
A recent framework for AI accountability splits the work into a triad: the Chief AI Officer “owns AI strategy, oversight, deployment and communication responsibilities,” the CISO “owns the AI security, data governance and threat modeling,” and the CCO “owns regulatory and compliance mapping, audit preparedness and third-party risk” — with the CEO as the final escalation point (Cherry Bekaert, May 2026). None of the ranking job-description templates draw this line. They all fold “ensure ethical and compliant use of AI” into the Head of AI's bullet list without saying who actually signs off when security and shipping speed conflict. Put the exclusion in writing and the first real disagreement has a pre-agreed owner instead of becoming a hiring-manager-versus-CISO standoff three weeks after the hire starts.
Generic years-of-experience and framework-proficiency requirements don't correlate with whether an agent reaches production. What does: prior ownership with budget authority, evaluation discipline, scoping discipline, and comfort with staged human oversight.
Every top-ranking template leans on the same cargo-cult list: “8–12 years in AI, machine learning, or data science,” “proficiency in Python and frameworks like TensorFlow or PyTorch” (Lucent Search), or a 15-year productivity-tools requirement paired with “1–2 years GenAI tools” (Lahanas, LinkedIn). These read like an engineer's resume screen, not a leadership hire — and none of them are backed by any data showing they predict outcomes.
The traits that do correlate come from a study of what separates the roughly 12% of AI agent pilots that reach production from the 88% that don't (Digital Applied, April 2026):
None of that shows up as a requirement on any template we checked. Ask a candidate to walk through one agent they took from pilot to production, what the evaluation setup looked like, and where they deliberately cut scope — that interview finds the signal a years-of-experience line can't.
At growth-stage, the role is hands-on and reports to the CEO or CTO with little or no team. At enterprise, it becomes a governance seat with a multi-team org and a formal coordination line to the CISO and CCO, and pay roughly doubles.
Growth-stage companies pay $250k–$400k base for a builder who sets the roadmap, ships pilots personally, and runs with a small or no team. Enterprises pay $400k–$1M+ for someone who owns governance, compliance coordination, budget, and a multi-team org (McKelvey). The “does not own” section above matters more as scope grows: a three-person startup doesn't need a written triad with a CISO who doesn't exist yet, but a 2,000-person company does, and skipping it is exactly how the role quietly absorbs security and compliance work it was never meant to hold.
If you're not sure which scope you're actually hiring for, that's a separate diagnostic question, not a job-description question — see agent leadership for how the seat is defined across stages before you write the posting.
By day 90: every existing pilot audited, one workflow selected with a binary success criterion, and automated evaluation coverage in place. By day 180: that workflow's metric hit, missed, or the project killed on the record, with decision-rights lines against the CISO and CCO signed.
None of the templates we reviewed mention a timeline at all — they read as static role descriptions with no way to tell, three months in, whether the hire is working. Building the 90/180-day bar into the job description before the search starts does two things: it filters out candidates who can't point to a comparable milestone from a past role, and it gives the hiring manager a pre-agreed date to check the hire against instead of a vague sense that “things are moving.” Given that automated evaluation coverage alone is the difference between a 9% and a 47% rollback rate (Digital Applied, April 2026), “evaluation coverage shipped” is a more honest 90-day milestone than “strategy defined.”
$250k–$540k in US compensation, reconciled from four CAIO-specific sources (VerifyWise, May 2026). The top of that band is a total-comp figure, not base. Roughly $250k–$400k at growth-stage and $400k–$1M+ at enterprise (McKelvey). A fractional equivalent runs $4k–$20k/mo (KORE1, 2026).
That's base salary only — no equity, bonus, or search-fee figures are included, because we don't have sourced ranges for those and won't pad the number. A retained search adds 25–35% of first-year compensation on top, and an average AI engineering search already runs about four months (AY Automate), with leadership searches typically slower. The full comp breakdown, including why the range is this wide, is on the Chief AI Officer salary page. Once the description and comp are set, the sourcing and vetting sequence is covered separately in how to hire a Head of AI Agents.
Is a Head of AI job description the same as a Chief AI Officer job description?
Mostly. Companies use the titles interchangeably for the same mandate at different scope: Head of AI usually maps to $250k–$400k growth-stage scope, Chief AI Officer to $400k–$1M+ enterprise governance scope (McKelvey).
Does a Head of AI job description need coding requirements?
No. Ranking templates list TensorFlow or PyTorch proficiency as a requirement, but Head of AI is an ownership and decision-rights job, not an individual-contributor one. Require hands-on frameworks for an AI engineer hire, not this one.
What is the single biggest mistake in a Head of AI job description?
Omitting a named, measurable outcome. 94% of AI agent pilots that reach production have a named owner with budget authority and a measurable target (Digital Applied, 2026) — a JD without one signals nobody has decided what the role is for.
Who should a Head of AI report to?
CTO or CEO at growth-stage companies. At enterprise scale, CEO with a formal governance triad alongside the CISO and CCO (Cherry Bekaert, May 2026). The reporting line should track decision rights, not org-chart convenience.
How is a Head of AI job description different from an AI engineer job description?
An AI engineer job description lists technical skills for building models or agents. A Head of AI job description specifies what outcome the person owns, what they don't own, and how success is measured — scope, not a tech-stack list.