Resources / Employers
Legal AI & Automation Job Description Template (2026)
A complete, copyable JD for the role that selects AI tools, designs prompts and evaluation harnesses, and operates governance with the GC, CISO, and DPO — with platform-agnostic qualifications and directional comp anchored to Robert Half 2026.
Why most Legal AI & Automation job descriptions fail
The most common failure is reading like a job posting for an ML engineer. Requirements list PyTorch, model training, and a CS or ML degree. That filters to a candidate pool with $50K-higher comp expectations and almost no workflow exposure — and the role does not train models. It selects vendor tools, designs prompts, builds evaluation harnesses, and operates governance. The required-qualifications section should lead with workflow judgment and tool fluency, not framework names.
The second failure is naming “AI” in the title without naming the actual scope. Strong candidates need to see workflow design, attorney-in-the-loop gate design, governance partnership with the GC and CISO, and named tools. A JD that lists “evaluate emerging AI technologies” as the primary responsibility attracts people who want to research and filters out the operators who would actually ship workflows.
A third failure: silence on governance. Real legal-AI work lives or dies on the AI usage policy, data classification per workflow, vendor security review, and DPO partnership. JDs that omit governance attract candidates who treat governance as compliance overhead — the wrong frame — and miss the candidates who have operated under real policy and built their workflow design around it.
Legal AI & Automation job description template
Copy this template and adapt the bracketed fields. The structure leads with the workflow-design layer, follows with tool fluency, then governance, then change management. Compensation reads from the Robert Half 2026 Salary Guide for senior legal-operations roles plus an AI-domain premium.
Job Description Template — Legal AI & Automation
Job Title
[Legal AI & Automation Lead / Senior Legal AI & Automation Lead / Manager, Legal AI & Automation]
Reports To
Legal Operations Manager [or General Counsel for pre-legal-ops-team companies] — [City, State / Remote / Hybrid: X days in-office, City]. Dotted lines to the CISO for vendor security review and the DPO for data-handling decisions.
Role Summary
[Company Name] is hiring a Legal AI & Automation Lead to select AI tools, design workflows that use them, build prompt-evaluation harnesses, and operate the governance that lets the legal team trust the outputs. You will partner with attorneys to scope workflows where AI beats the human baseline, partner with the General Counsel on the AI usage policy, partner with the CISO and DPO on vendor security and data-handling, and own the measured outcomes of every deployed workflow — including the decision to retire workflows that stop earning their keep.
Key Responsibilities
- Identify workflow candidates for AI assistance by partnering with attorneys to baseline the current human cost, output quality, and failure modes
- Select AI tools across categories: legal-domain LLM platforms ([Harvey, Thomson Reuters CoCounsel, Lexis+ AI]), document-grounded research ([Hebbia, Robin AI]), contract draft and review ([Spellbook, Ironclad AI]), litigation-specific ([EvenUp]), and general-purpose LLM APIs ([OpenAI via Azure, Anthropic via Bedrock or Vertex])
- Design prompts, evaluation harnesses, and regression tests for each deployed workflow; maintain reference outputs and metric-based scoring
- Design attorney-in-the-loop gates per workflow: where the gate sits, what the attorney reviews, escalation path for ambiguous outputs
- Co-author the AI usage policy with the General Counsel; partner with the CISO on vendor security review and the DPO on data classification (PII, privileged, MNPI) per workflow
- Build training and enablement for attorneys and legal-team users; measure adoption with completion rate, time saved per task, and explicit user-feedback cadence
- Maintain workflow inventory: deployed workflows, ownership, evaluation harness, attorney-in-the-loop gate, measured outcome, retirement criteria
- Lead vendor evaluations: reference checks, security review, pilot scoping, total-cost-of-ownership analysis including change-management cost
- Respond to AI-related incidents: hallucinations or retrieval failures that reach production, privacy or privilege issues surfaced by deployed workflows, vendor outages
- Retire workflows that no longer beat the human baseline or whose maintenance burden exceeds the value delivered
Required Qualifications
- 4–8 years of operations, legal-tech, or workflow-design experience with at least 2 years deploying production AI or automation workflows in a regulated environment
- Demonstrated portfolio of deployed AI workflows: tool selected, prompt designed, evaluation harness built, measured outcome documented, attorney partnership maintained
- Hands-on prompt design and evaluation depth: reference outputs, regression tests, structured-output schema design, few-shot vs zero-shot judgment
- Working fluency with at least three of: legal-domain LLM platforms (Harvey, Thomson Reuters CoCounsel, Lexis+ AI), document-grounded research (Hebbia, Robin AI), contract draft and review (Spellbook, Ironclad AI), general-purpose LLM APIs (OpenAI via Azure, Anthropic via Bedrock or Vertex)
- Track record of partnering with the General Counsel on AI usage policy and with the CISO and DPO on vendor security review and data classification
- Comfortable distinguishing hallucination from retrieval failure with named mitigations; comfortable making the call when an AI workflow should not be deployed
- Strong written communication for memos to executives: GC, CISO, DPO, and CFO audiences
Preferred Qualifications
- Microsoft AI-900 (AI Fundamentals) or AI-102 (AI Engineer) certification; AWS AI Practitioner or Google Cloud Generative AI Leader equivalents accepted
- Prosci ADKAR or comparable change-management certification
- JD, paralegal certificate, or contract-manager background (IACCM, NCMA, CCCM)
- Hands-on experience with at least one general-purpose LLM API (OpenAI, Anthropic, Vertex) including structured outputs, function calling, and tool use
- Experience with a prompt-evaluation framework (LangSmith, Braintrust, Helicone, Weights & Biases, or comparable)
- Familiarity with privacy and AI-governance frameworks: NIST AI Risk Management Framework, EU AI Act categorization, SOC 2 / ISO 27001 for vendor security
- SQL fluency for ad-hoc workflow-outcome analysis
Compensation and Benefits
Base salary $[X]–$[Y] depending on experience and scope; [10–20]% annual bonus target; equity at market rate for stage. Directional band: $130,000–$180,000 base for an IC senior role and $160,000–$210,000 base for a manager-tier role with one or more reports — benchmarked from the Robert Half 2026 Salary Guide for senior legal-operations roles plus a 10–20% AI-domain premium for the prompt-design and evaluation depth. HCOL metros (NYC, SF, Boston) trend toward the upper end. Full benefits including [health, dental, vision, 401(k) with match]. Professional development budget for one AI certification and one legal-tech or AI conference. We publish our compensation bands and do not ask for prior salary history.
Equal Opportunity
[Company Name] is an equal opportunity employer. We are committed to building a diverse team and will consider all qualified applicants without regard to race, color, religion, sex, national origin, disability, veteran status, or any other legally protected characteristic.
Comp band above is directional (Robert Half 2026 Salary Guide, senior legal-operations roles, plus a 10–20% AI-domain premium). Public benchmarks for “AI engineer” or “ML engineer” overshoot the role — the work is operations-tier with AI depth, not model-training engineering.
How to adapt the template by function shape
The required-qualifications section should be tightened to match the legal function shape and AI maturity of your company.
Contracts-heavy function (high-volume CLM, NDA review, redlining)
- Required: Deployed at least one CLM-adjacent AI workflow (Ironclad AI, Spellbook, Robin AI, Hebbia for contract Q&A) with measured outcome and attorney-in-the-loop design.
- Preferred: Hands-on with structured-output schemas for clause extraction; experience integrating LLM workflows with a CLM platform.
Litigation-heavy function (eDiscovery, document review, pleading drafting)
- Required: Deployed at least one litigation-adjacent AI workflow (Lexis+ AI, Thomson Reuters CoCounsel, Hebbia, EvenUp for PI work) with attorney partnership.
- Preferred: Familiarity with eDiscovery platforms (Relativity, DISCO, Everlaw) and the AI features inside them; understanding of work-product privilege as it applies to AI workflows.
Compliance and privacy-heavy function (privacy program, regulatory tracking)
- Required: Deployed at least one compliance-adjacent AI workflow (regulatory-change tracking, policy review, control mapping) with measured outcome.
- Preferred: Familiarity with NIST AI RMF and EU AI Act categorization; experience partnering with the DPO on data-classification decisions for AI workflows.
Early-stage function (first AI hire, no policy in place)
- Required: Co-authored at least one AI usage policy in a prior role; comfortable with the workflow-selection-before-tool-selection sequence; willing to operate in a function with no prior AI deployment baseline.
- Preferred: Experience standing up an AI program from zero at a prior company; Prosci ADKAR or comparable change-management certification.
What good looks like — evaluation rubric
Use this rubric to evaluate candidates against the JD above. Each criterion should produce a clear pass / fail signal.
Has deployed a real AI workflow with measured outcome
Ask for the most recent end-to-end deployment: workflow selected, baseline measured, tool chosen, prompt designed, attorney-in-the-loop gate, rollout cadence, measured result. Strong candidates have at least one. Candidates who only narrate evaluations have not operated the role.
Has a rejection story
Ask about an AI tool they evaluated and decided NOT to deploy. Strong candidates have a story about the rejection — vendor claims that did not survive a pilot, integration friction that broke the workflow, governance gates that could not be met. Candidates who think every tool is worth deploying have not operated under real constraints.
Distinguishes hallucination from retrieval failure
Ask them to describe the difference and how they would mitigate each. Strong candidates name retrieval as a grounding-and-source-design problem and hallucination as a prompt-and-evaluation problem — with specific mitigations for each. Weak candidates treat them as the same vendor problem.
Has co-authored an AI usage policy
Ask for the policy they have worked under or written. Strong candidates can name the data-classification language, the attorney-in-the-loop requirements per workflow category, the vendor-approval gates, and the exception process. Candidates who have never operated under written policy will create governance debt in your function.
Has retired a workflow
Ask for a workflow they deployed and later retired. Strong candidates have at least one — the value declined, the maintenance burden grew, or a better tool replaced it. Candidates who have only deployed workflows and never retired any are missing half the job; AI workflows degrade if no one is willing to turn them off.
Where to post the job description
Legal AI & Automation candidates are concentrated in three channels. Post the JD directly to HireLegalOps first — the niche board has a Legal AI & Automation family filter. Then post to the CLOC community job board, where the legal-ops AI subcommittee monitors postings, and to ILTA, which concentrates legal-tech practitioners with AI exposure. LinkedIn Boolean searches against “legal AI” combined with named tools (Harvey, CoCounsel, Spellbook) surface adjacent candidates from law firms and corporate legal departments.
Job description questions answered
Should this role report into Legal Ops, the GC, or IT?
Legal Operations with a dotted line to the GC and working partnership with the CISO and DPO. Reporting solid into IT pulls the role away from workflow design; reporting solid into GC without a Legal Ops home leaves it isolated from change management.
Technical role or legal role?
Workflow role with technical depth and legal-domain instinct. Strong candidates come from either legal-first or tech-first paths; both work. What does not work is pure ML researcher or pure attorney with no prompt-evaluation depth.
What tools should we name?
Name categories with one preferred tool each, not all tools. Legal-domain LLM (Harvey, CoCounsel, Lexis+ AI); document-grounded (Hebbia); contract draft (Spellbook); litigation (EvenUp); general LLM API (OpenAI on Azure, Anthropic on Bedrock). Requiring all of them filters to fewer than 200 candidates nationally.
Require ML or data-science credentials?
No. The role does not train models — it selects vendor tools, designs prompts, builds evaluation harnesses, and operates governance. Microsoft AI-900 or AI-102 are reasonable preferred; Prosci ADKAR is more load-bearing than any technical credential.
Name governance scope?
Yes. AI usage policy co-authorship; data classification per workflow; CISO and DPO partnership; SOX-relevant audit evidence for finance-adjacent workflows; attorney-in-the-loop gate design. Naming this attracts candidates who have operated under real policy.
How to set comp without a clear benchmark?
Anchor against Robert Half 2026 senior legal-ops roles plus 10–20% AI-domain premium. Directional range $130K–$180K base IC senior; $160K–$210K manager-tier. Public “AI engineer” benchmarks overshoot.
Hire vendor or build internal?
Both, in sequence. First hire is internal — selects vendors, designs workflows, owns governance. Outsourcing the workflow-design work means outsourcing the strategic AI decision, which no consultancy should own. Vendors implement once your internal owner has scoped.
What should we NOT include in the JD?
Five inclusions that tank the pool: model-training or PyTorch requirements; a CS or ML degree as required (preferred is fine); vague “evaluate emerging AI” as primary responsibility; silence on governance; and “Director” in the title without a real director scope. Each shifts the role away from the workflow-and-governance work it should select for.
Ready to post the role? Browse Legal AI & Automation interview questions, review the Legal Operations Tools & Tech Stack 2026 for context on the broader platform layer, or post directly on HireLegalOps to reach AI & automation practitioners across the nine legal-ops role families.
Post a Legal AI & Automation job