This page explains how the Durability Score is built — the
components, the evidence behind each one, and the named sources. For who
this work fits and what a career path through it looks like, see the
Deep Read. For your personalized match, take
the free quiz.
Where the 40 comes from.
Three components — Automation Resistance, Structural Moat, and Demand — add up to the 40.
FJP Durability Score
40/100
Automation Resistance
15/40
AI reaches research, drafting, summaries, citation checks, and document review directly. Attorney supervision and procedure preserve a human floor, but that floor is thinner for paralegals than for licensed legal practice, especially in first-pass document work.
Sub-components
Substitution Resistance
10/30
The exposed work includes research, document review, citation checks, routine drafting, records summaries, contract review, and case-file organization. The resistant work is procedure, deadline control, client facts, evidence handling, local filing rules, and supervised verification. AI can produce a first pass; a paralegal's value depends on knowing what must be checked.
Sources feeding this sub-component
BLS Occupational Outlook Handbook → Federal projection authority explicitly names AI as the demand limiter for paralegals: "advances in technology, including artificial intelligence... are expected to make paralegals and legal assistants more efficient at tasks such as conducting research and preparing documents, which may reduce demand for these workers."
Clio Legal Trends Report 2024 → Industry survey: 69% of paralegal-typical hourly billable tasks could be automated with AI; ~$27,000 estimated annual hourly billing reduction per lawyer attributable to AI automation.
Massenkoff & McCrory "Labor market impacts of AI" → Legal major-group ~15% observed-exposure (paralegal not in top-10 most-exposed at occupation level — top 10 ranges Computer Programmers 75% to Data Entry Keyers 67%); 14% job-finding-rate decline for ages 22–25 entering high-exposure occupations.
Thomson Reuters CoCounsel → Enterprise-deployed agentic legal-AI in production at 20,000+ firms and corporate legal departments; majority of Am Law 100; serves 100% of Fortune 100. Vendor-disclosed adoption metric.
MIT Iceberg Index + Tufts AAJRI → Skills models flag legal research, document summaries, citation lookup, and contract analysis as highly exposed.
Stanford RegLab + Yale legal-hallucination research → Dahl et al. 2024 (general-purpose AI like ChatGPT hallucinates 58–88% on specific verifiable legal questions); Magesh et al. 2025 (legal-specific RAG tools hallucinate 17–33%). Even purpose-built tools still need human verification.
ABA Model Rule 5.3 + Formal Opinion 512 → Supervising-lawyer verification requirement for AI-assisted nonlawyer work. Enforced through court sanctions (see the Charlotin database under Regulatory Moat).
Augmentation Leverage
5/10
Legal AI can save time on summaries, research, drafting, contract review, and document triage. That helps firms produce work faster, but it may reduce paralegal hours when demand does not grow enough to absorb the productivity gain. The worker benefits most when the tools support higher-value litigation support or e-discovery work.
Sources feeding this sub-component
Forrester Total Economic Impact study for LexisNexis (2025) → Modeled in-house team at $10B-revenue company: AI tools reduce outside-counsel volume 13%; reduce internal legal-inquiry time 25%; paralegals achieved a 50% time savings on administrative tasks. Vendor-commissioned but methodologically transparent.
Clio Legal Trends Report 2025 → 79% of legal professionals using AI (up from 19%); 82% expect to use more in next 12 months; 59% of firms now using flat-fee or subscription pricing; 69% of wide-AI-adopters report positive revenue impact.
Anthropic Economic Index → Real-usage data on legal-task AI conversation patterns. Supplementary materials note that legal-secretary/paralegal task profiles resemble lawyer-typical tasks (~17.7 years required-education equivalent) — strong qualitative evidence that the task profile is heavily LLM-feasible.
Vals Legal AI Benchmark 2025 → Independent benchmark across Harvey, CoCounsel, Spellbook, Lexis+ AI; CoCounsel highest avg 79.5% across four tasks; Harvey 94.8% on document Q&A.
Structural Moat
14/35
The moat is procedure, confidentiality, attorney supervision, and voluntary credentials rather than broad licensure. Some settings add practical protection through deadlines and evidence systems, but most entry work lacks a legal practice gate. That makes setting unusually important.
Sub-components
Physical & Environmental
1/10
Paralegal work is mostly desk-based, with some courthouse, client, deposition, or trial-support activity in litigation settings. Physical conditions add little protection. The job's real demands are concentration, confidentiality, deadlines, attention to version control, and the pressure of legal consequences.
Sources feeding this sub-component
BLS Occupational Outlook Handbook → Task profile: legal research, document organization, drafting, intake interviews, citation lookup. Most paralegals work in offices.
Regulatory Moat
4/12
Most states do not license paralegals. Voluntary credentials can help, and attorney supervision rules require lawyers to oversee nonlawyer work, including AI-assisted work. That supervision protects legal responsibility and quality control, but it does not give paralegals independent practice authority or a strong entry gate.
Charlotin AI Hallucination Cases Database (HEC Paris) → Live database, updated daily; about 1,370 cases globally as of early May 2026; party-type split (Lawyer 38% / Pro Se 59% / Paralegal 0.15%); Sixth Circuit Whiting v. City of Athens $30K sanction (March 2026); Special Master Hon. Michael R. Wilner (ret.) $31,100 sanction in Lacey v. State Farm (C.D. Cal., May 6, 2025); multiple recent bar-disciplinary actions.
Robotics Resistance
8/8
Robotics does not affect the occupation. Paralegal substitution pressure comes through legal AI, document systems, search, drafting, and workflow automation rather than physical machines.
Typical entry is an associate degree, bachelor's degree, or paralegal certificate, with voluntary credentials available. The preparation depth is real but modest compared with licensed legal practice. The deeper skill comes from procedure, filings, discovery systems, contracts, and supervised case experience.
Sources feeding this sub-component
BLS Occupational Outlook Handbook → Typical entry is an associate degree, certificate, or bachelor's degree; no license is required.
Demand is soft because growth is essentially flat and the entry workflow is exposed. Openings still exist, but research, drafting, discovery, document review, and support summaries face direct legal-AI pressure, especially for beginners in the first-draft layer.
Sub-components
Volume
5/10
The direct occupation has about 376,200 jobs, roughly 39,300 annual openings, and about 0.2% projected growth. The base is sizable, but it is not expanding enough to offset the automation pressure cleanly.
Sources feeding this sub-component
BLS Employment Projections → Paralegals and Legal Assistants (23-2011): 376.2K jobs, 0.2% growth, and 39.3K annual openings.
Source Quality
4/8
Demand quality is mixed. Law firms, government offices, courts, and corporate legal departments still need support, but much entry work is exactly the kind of text and document processing legal AI can accelerate.
Resilience is limited because legal research, drafting, document review, discovery support, and summaries are exposed. Attorney supervision and court procedure preserve some human work, but they do not guarantee the same number of paralegal seats.
Sources feeding this sub-component
BLS wage tables, May 2015 and May 2025 → May 2015 median $48,810 equals $66,291 in 2025 dollars using CPI-U 237.017 to 321.943; May 2025 median $62,890 is about -5.1% real. The wage-pressure reduction is included in Resilience.
Agentic legal-AI matures past the need for human verification.
The case weakens if legal AI becomes reliable enough that attorneys trust it for common workflows with less paralegal verification. The threshold is fewer supervised support hours, not a perfect replacement for lawyers. Watch whether firms cut junior review hours after adoption.
Direction
Down, meaningful
Components affected
Substitution Resistance, Demand
Scenario 2
Lawyer-side AI adoption rebalances supervision burden.
The case becomes ambiguous if AI changes lawyer staffing itself. Paralegal demand could fall if law firms shrink support teams, or rise if AI output creates more review, filing, client, and coordination bottlenecks. The variable is whether supervision creates work or removes it.
Direction
Ambiguous
Components affected
Demand, Substitution Resistance
Scenario 3
A meaningful state moves to license paralegals.
The case improves if a large state creates meaningful paralegal licensing or a defined AI-supervision role. A voluntary certificate is not enough; the trigger is enforceable scope, practice rights, or required oversight responsibility. Watch scope-of-practice rules and employer requirements, not badge language.
