The AIR Index scores how ready your firm is to put AI into legal work without risking confidentiality, privilege, or your standing before the court.
Generative AI is already drafting in your associates' browsers and courts are already sanctioning fabricated citations, so the choice is to govern it on purpose now or explain it under scrutiny later.
The same five pillars of AI readiness, framed in the work, systems, and stakes that law-firm and corporate-legal leaders actually face.
Whether AI is embedded at named steps of how matters get done, with verified workflows and reusable assets, rather than trapped in a few power users.
Whether the firm captures the value AI creates through fixed-fee and productized offerings, or hands it back as shrinking billable hours and realization.
Whether AI competence is broad, owned, and built into roles across attorneys and staff, meeting the duty of technological competence rather than resting on enthusiasts.
Whether real controls protect client confidences, privilege, and output quality under the Model Rules and court orders, instead of hope and informal understanding.
Whether the firm runs a deliberate, owned, data-residency-aware AI stack with visible spend and a review cadence, rather than sprawling shadow subscriptions.
AIR places Legal on a six-tier readiness ladder from 0 to 100, overall and for every pillar. The climb runs from digital but not intelligent, to a compounding, AI-native edge.
AI-native advantage. Compounding intelligence and speed, a durable edge competitors can't copy fast.
Woven through the business. AI shapes the operating model, pricing, and talent, and ROI is proven.
AI in the core, governed. Embedded at named steps with SOPs, policy, and measured gains.
Pockets, not a system. Real use in places, uneven and undocumented.
Experimenting at the edges. Scattered pilots that live in a few people's heads, ungoverned.
Digital, not intelligent. AI is absent or anecdotal, work is hour-priced, the stack sprawls, and no policy exists.
One score tells you that you are behind; five tiers tell you exactly where to start.
The matrix is a 5-by-6 grid: your five pillars of AI readiness scored against the same six tiers, from Legacy to Autonomous. A single overall score tells you roughly where you sit; it hides where you are dangerously behind and where you are quietly ahead. Reading a tier per pillar turns one vague number into five specific, fixable verdicts, so you act on the truth instead of an average.
How Legal's regulatory reality maps onto AIR readiness. Each row is a control your governance pillar has to carry.
| Regime | Pillar | What AI readiness requires |
|---|---|---|
| ABA Model Rule 1.1 (Competence), Comment 8 | P3 | Demonstrate ongoing AI training so lawyers understand the benefits and risks of the tools they use. |
| ABA Model Rule 1.6 (Confidentiality of Information) | P4 | Bar client confidences from consumer AI tools and route privileged matter data only through tenant-isolated, no-training environments. |
| ABA Model Rule 5.3 (Responsibilities Regarding Nonlawyer Assistance) | P4 | Supervise AI as a nonlawyer assistant, with a named owner accountable for its output and conduct. |
| Attorney-Client Privilege and Work-Product Doctrine | P4 | Prove that AI workflows do not waive privilege through third-party disclosure, retention, or model training on matter content. |
| Unauthorized Practice of Law (UPL) limits | P2 | Keep AI-assisted offerings supervised by a licensed lawyer so packaged or productized legal work never crosses into UPL. |
| Court standing orders on generative-AI disclosure and verification | P1 | Embed a mandatory citation-verification checkpoint so no AI-drafted filing reaches a court without confirmed authorities. |
| Data residency and cross-border transfer (GDPR Art. 44-49, state privacy laws) | P5 | Select AI tools whose hosting region and subprocessors satisfy client data-residency and transfer obligations. |
| Outside Counsel Guidelines and client AI-use clauses | P4 | Maintain a consistent, disclosable position on AI use that aligns with each client's outside-counsel guidelines. |
Illustrative mapping for AI-readiness planning, not legal or compliance advice; validate against current rules of professional conduct, court orders, and data regulation with qualified counsel.
An associate files an AI-drafted brief citing cases that do not exist, drawing sanctions, a malpractice exposure, and reputational damage that no efficiency gain offsets.
Confidential matter content pasted into a consumer chatbot is retained or used for training, arguably waiving privilege and breaching the duty of confidentiality under Rule 1.6.
AI compresses the hours a matter takes while fees stay hour-based, so faster work quietly shrinks revenue instead of widening margin or funding new offerings.
Lawyers and staff adopt personal AI accounts with no policy, no logging, and no supervision, leaving the firm unable to answer a client or regulator about how AI touched their matter.
Productized or self-service AI legal work runs without licensed-lawyer review, risking unauthorized practice and confidently wrong advice that the firm cannot defend.
Concrete first moves you can make before the full diagnostic, one per pillar where it matters most.
Issue a written, firm-wide policy naming approved tools, banning client data in consumer chatbots, and requiring lawyer review of all AI output.
Require every AI-assisted filing to pass a documented Shepardize/KeyCite check confirming each authority before it leaves the firm.
Inventory every AI subscription, personal account, and embedded vendor feature in use, then name one owner for the legal-tech stack and its spend.
Deliver a structured training on AI benefits, risks, and confidentiality so competence is demonstrable, not assumed, across attorneys and staff.
Build a governed prompt-and-clause library for a high-volume workflow such as NDA review or discovery summaries, with documented steps the whole team follows.
Convert a commoditizing service into a flat-fee, supervised package so AI speed becomes margin and a new revenue line, not a discount.
The old transformation is finished. The new one is scored.