Glossary

Legal Tech Terms, Explained

The legal tech glossary: 173 legal technology and legal AI terms in plain English, from CLM and eDiscovery to RAG, SOC 2, LEDES and trust accounting.

By The Circle DeskUpdated November 10, 2026

AI and Data

ABA Formal Opinion 512: The American Bar Association's July 2024 ethics opinion on lawyers' use of generative AI, covering competence, confidentiality, communication with clients, supervision and fees.
AI agent: An AI system that carries out a multi-step task on its own, such as gathering documents, reviewing them against a checklist and drafting a summary, rather than answering a single prompt.
AI copilot: An AI assistant built into software people already use, such as Word or a practice management system, that suggests drafts, edits and answers as they work.
AI disclosure: Telling a court, a client or another party that AI was used to prepare a document or deliver a service.
AI governance: The policies, approvals, training and monitoring an organization uses to decide which AI tools people may use, for what, and how the results are checked.
AI hallucination: A confident but false answer from an AI tool, such as a citation to a case that does not exist or a quote no court ever wrote.
AI use policy: A written policy that tells lawyers and staff which AI tools they may use, what information they may put into them and how output must be reviewed.
AI wrapper: A product that adds an interface, prompts and workflows on top of someone else's AI model rather than building its own model.
Boolean search: Keyword search using operators such as AND, OR, NOT and proximity connectors to control exactly which documents match.
Citation checking: Verifying that every case, statute and quote cited in a brief or memo exists, says what it is cited for, and is still good law.
Clause extraction: Using AI to find specific clauses and data points in contracts, such as governing law, renewal dates and liability caps, and turn them into structured fields.
Context window: The amount of text an AI model can consider at once, measured in tokens. Larger windows let a model read longer documents in one pass.
Embeddings: Numeric representations of text that capture meaning, so a computer can tell that "terminate for convenience" and "end the agreement without cause" are related.
Fine-tuning: Further training an existing AI model on a narrower set of examples, such as legal documents, so it performs better on that kind of work.
Foundation model: A large, general-purpose AI model trained on broad data that other products are built on top of. Most LLMs are foundation models.
Generative AI: AI that produces new text, images or code from a prompt. In legal work it drafts, summarizes, translates and answers questions over documents.
Grounding: Tying an AI answer to specific source material, such as a statute, a case or a clause, so the reader can see where each statement came from.
Human in the loop: A process where a person reviews and approves AI output at defined points before it is relied on or sent anywhere.
Legal AI benchmark: A structured test that measures how accurately AI tools perform defined legal tasks, such as answering research questions or reviewing contracts.
LLM (large language model): The type of AI model behind ChatGPT, Claude, Gemini and most legal AI assistants, trained on very large amounts of text to predict and generate language.
Machine learning: Software that learns patterns from examples rather than following hand-written rules. It powered legal tech well before generative AI, for example in technology-assisted review.
NLP (natural language processing): The field of computing that deals with understanding and producing human language, from finding names and dates in a contract to answering questions.
OCR (optical character recognition): Technology that converts scanned images and PDFs into searchable, selectable text.
Prompt engineering: Writing instructions for an AI tool in a way that reliably produces the output you need: the task, the context, the format and the constraints.
Prompt library: A shared, curated set of tested prompts for common legal tasks, such as summarizing a deposition or reviewing an NDA against a playbook.
RAG (retrieval-augmented generation): An AI technique that first looks up relevant documents, such as cases or your contracts, and then has the model answer from what it found.
Semantic search: Search that matches meaning rather than exact keywords, so a question in plain English can find relevant documents that use different wording.
Tokens: The small chunks of text, roughly three quarters of a word on average, that AI models read and write. Usage and limits are measured in tokens.
Training on customer data: Whether a vendor uses the documents and prompts you send to improve its AI models.
Vector database: A database that stores embeddings and finds the closest matches by meaning rather than by exact words.
Zero data retention: A vendor commitment that prompts and documents sent to an AI model are not stored after processing and are not used for training.

Contracts

Approval workflow: Rules that route a contract to the right people for sign-off based on its type, value, risk or deviations from standard terms.
Auto-renewal: A contract term that renews the agreement automatically for another period unless one side gives notice by a set deadline.
Clause library: A curated set of approved contract language that lawyers and AI tools draw on when drafting and negotiating.
CLM (contract lifecycle management): Software that manages a contract from request through drafting, negotiation, approval, signature, storage, obligations and renewal.
Contract analytics: Analysis of a body of contracts to answer questions such as which agreements allow assignment, where liability is uncapped or how terms vary by region.
Contract intake: The process by which the business asks legal for a contract or a review, ideally through one form that captures the details legal needs up front.
Contract playbook: A guide that sets an organization's preferred position on each key clause, the acceptable fallbacks and who must approve anything outside them.
Contract repository: A central, searchable store of an organization's signed contracts and their key data, such as parties, dates, values and renewal terms.
Contract review: Reading a contract against a playbook or standard to flag risky, unusual or missing terms. Increasingly done first by AI and then checked by a lawyer.
Contract template: A pre-approved standard version of a contract that can be reused with only deal-specific details changed.
DPA (data processing agreement): A contract that governs how a vendor processes personal data on a customer's behalf, required in many cases under the GDPR and similar laws.
Due diligence review: Reviewing a target company's contracts, records and risks before a merger, acquisition or investment.
E-signature: An electronic way to sign a document, legally recognized for most commercial contracts in the US, UK, EU and many other jurisdictions.
Fallback position: A pre-approved alternative to the preferred contract term that a negotiator may accept without escalating.
MSA (master services agreement): A framework contract that sets the main legal terms between two parties, so later work can be agreed in shorter statements of work or orders.
NDA (non-disclosure agreement): A contract in which one or both parties agree to keep shared information confidential.
Obligation management: Tracking what each signed contract requires and when: payments, notices, reporting, renewals and deliverables.
Redlining: Marking up proposed changes to a contract or document so the other side can see every edit.
SOW (statement of work): A document describing specific services, deliverables, timelines and fees under a master agreement.
Virtual data room: A secure online space where documents are shared with a limited group, usually for a transaction or investigation.

Litigation and Discovery

Bates numbering: A unique, sequential identifier stamped on each page or document in a production so it can be referenced precisely.
CAL (continuous active learning): A form of technology-assisted review, sometimes called TAR 2.0, in which the model keeps learning from every reviewer decision and continually re-ranks the remaining documents.
Court e-filing: Submitting court documents electronically through a court's filing system, such as PACER/CM-ECF for US federal courts.
Data collection (eDiscovery): Gathering potentially relevant ESI from email systems, devices, cloud apps and other sources in a way that preserves its integrity and metadata.
Deduplication: Removing exact copies of the same document or email so reviewers only see it once.
Deposition summary: A condensed account of deposition testimony organized by topic or page, used to prepare for motions and trial.
Docketing: Tracking court filings, deadlines and events for every matter, often calculated from court rules.
ECA (early case assessment): Analyzing data and facts early in a dispute to estimate its risk, cost and likely outcome before committing to full discovery.
eDiscovery: Collecting, reviewing and producing electronically stored information, such as email, chat and documents, in litigation, investigations and regulatory requests.
EDRM (Electronic Discovery Reference Model): A widely used framework that maps the stages of eDiscovery: information governance, identification, preservation, collection, processing, review, analysis, production and presentation.
Email threading: Grouping emails from the same conversation so reviewers can read only the most complete messages instead of every reply.
ESI (electronically stored information): Any information created, stored or used in digital form that may be relevant in litigation: email, documents, chat messages, databases, audio, video and more.
Legal hold: An instruction to preserve information that may be relevant to a reasonably anticipated or pending legal matter, suspending normal deletion.
Litigation analytics: Data on how judges, courts, opposing counsel and parties have behaved in past cases, such as motion grant rates and time to trial.
Predictive coding: An earlier name for technology-assisted review, where a model trained on attorney-coded samples predicts the relevance of the remaining documents.
Privilege log: A list of documents withheld from production on privilege grounds, with enough description for the other side to assess the claim.
Privilege review: Checking documents in discovery for attorney-client privilege or work product before they are produced.
Processing (eDiscovery): Converting collected data into a reviewable form: extracting text and metadata, removing duplicates, expanding containers and indexing for search.
Production (eDiscovery): Delivering reviewed, responsive documents to the other side or a regulator in an agreed format.
Redaction: Permanently removing privileged, confidential or personal information from a document before it is shared.
Rules-based calendaring: Software that calculates deadlines automatically from court and procedural rules once a triggering event, such as service of a complaint, is entered.
TAR (technology-assisted review): Using machine learning to rank or classify documents for review in eDiscovery, so people review the most likely relevant documents first or not at all.
Transcript analysis: Searching and analyzing deposition, hearing and trial transcripts to find testimony, inconsistencies and themes.

Practice and Firm Management

AFA (alternative fee arrangement): Any way of pricing legal work other than the billable hour, such as flat fees, capped fees, success fees or subscriptions.
Billable hour: The traditional law firm pricing model, in which clients pay for the time lawyers and staff spend on their matters at hourly rates.
Client portal: A secure online space where clients can share documents, message their lawyer, see matter status and pay invoices.
Conflict check: Searching a firm's records for relationships with a prospective client, opposing parties and related entities before taking on a new matter.
Flat fee: A fixed price for a defined piece of legal work, regardless of the time it takes.
Legal billing: Turning recorded time and expenses into invoices, applying rates and client billing rules, and tracking payment.
Legal CRM: Client relationship management software built for law firms, tracking contacts, referrals, pitches and business development activity.
Legal intake: How requests reach a legal team, or how new clients reach a firm, ideally through one front door that captures the details and routes the request.
Legal project management: Applying project management methods, such as scoping, budgeting, staffing and tracking, to legal matters.
Matter management: Tracking each legal matter, its people, documents, deadlines, budget and status, in a law firm or legal department.
Practice management software: The system that runs a law firm day to day: clients and matters, calendar and deadlines, time tracking, billing, trust accounting and a client portal.
Realization rate: The share of the value of time recorded that a firm actually collects, after write-downs, discounts and write-offs.
Three-way reconciliation: Regularly matching the bank statement, the trust account ledger and the individual client ledgers so all three agree.
Time tracking: Recording the time lawyers and staff spend on each matter, usually in six-minute increments, for billing and analysis.
Trust accounting (IOLTA): Holding and tracking client money separately from the firm's own funds, as bar rules require. IOLTA accounts are pooled interest-bearing trust accounts used for this in the US.
Utilization rate: The share of available working hours that a lawyer records as billable time.

Legal Operations and Spend

Accruals (legal spend): Estimates of legal work performed but not yet billed, reported so finance can record the liability in the right period.
ALSP (alternative legal service provider): A company that delivers legal services, such as document review, contract management or eDiscovery, outside the traditional law firm model.
Billing guidelines: The specific invoice rules within outside counsel guidelines, such as rate caps, non-billable tasks, expense limits and timekeeper approvals.
Change management: The work of helping people adopt a new tool or process: communication, training, champions, support and measuring use.
E-billing: Electronic submission and review of law firm invoices by legal departments, usually with billing guidelines applied automatically.
LEDES: The Legal Electronic Data Exchange Standard, a set of file formats for sending law firm invoices and related data electronically.
Legal budget: A legal department's plan for internal costs, outside counsel and technology over a year, often broken down by matter or business unit.
Legal chatbot: A conversational tool that answers routine legal questions or guides users through a process, increasingly powered by generative AI.
Legal KPIs: The metrics a legal team uses to show performance, such as request volume, turnaround time, spend against budget and contract cycle time.
Legal operations (legal ops): The business side of a legal department: budgets and outside counsel spend, vendors, data and metrics, process and technology.
Legal self-service: Tools that let the business handle routine legal needs itself, such as generating standard NDAs or finding answers to common questions, within limits set by legal.
Legal spend management: Software and processes for budgeting, tracking and controlling what a legal department spends on outside counsel and vendors.
Legal tech stack: The set of software a legal team or firm uses and how the pieces connect: for example intake, CLM, matter management, e-billing, document management and AI tools.
Legal vendor management: Selecting, contracting with and reviewing the performance of law firms, ALSPs and technology vendors a legal team relies on.
Managed review: Document review in litigation or investigations run by a provider with its own reviewers, workflow and quality control.
Outside counsel guidelines: A legal department's written rules for the firms it hires, covering staffing, billing, expenses, communication, security and, increasingly, use of AI.
Panel counsel: A preferred list of law firms a legal department uses for particular kinds of work, usually selected through a formal process with agreed rates.
Pilot (proof of concept): A limited trial of a tool with real users and real work, with success measures agreed before it starts.
RFP (request for proposal): A structured document asking vendors to respond to the same requirements, questions and pricing format so they can be compared fairly.
ROI (return on investment): The value a technology delivers, such as time saved, spend avoided or risk reduced, compared with its total cost.
Total cost of ownership: The full cost of a system over its life: licenses, implementation, integration, training, administration, support and eventual migration.
UTBMS codes: Standard task and activity codes law firms use on invoices so legal departments can compare and analyze spend by type of work.

Security, Privacy and Compliance

BYOK (bring your own key): An arrangement where the customer controls the encryption keys protecting its data in a vendor's cloud, and can revoke them.
CCPA: The California Consumer Privacy Act, as amended by the CPRA, which gives California residents rights over personal information businesses collect about them.
Data residency: Where a vendor physically stores and processes your data, such as the US, the UK or the EU.
Data retention policy: Rules for how long different types of records are kept and when they are deleted.
DPIA (data protection impact assessment): An assessment of the privacy risks of a new system or process, and how they will be reduced, required under the GDPR for higher-risk processing.
DSAR (data subject access request): A request from an individual to see, correct or delete the personal data an organization holds about them.
Encryption at rest and in transit: Protecting data with encryption while it is stored (at rest) and while it moves between systems (in transit), so it cannot be read if intercepted.
Ethical walls: Restrictions that block certain lawyers and staff from accessing information about a matter, to manage a conflict of interest.
EU AI Act: The European Union's law regulating AI systems by risk level, which entered into force in 2024 with obligations phasing in over the following years.
GDPR: The EU General Data Protection Regulation, which governs how personal data about people in the EU is collected, used, transferred and protected. The UK has its own version.
GRC (governance, risk and compliance): Software that manages policies, risks, controls, audits and compliance obligations across an organization.
Information governance: Managing information across its whole life, including creation, classification, storage, access, retention and disposal, to meet legal and business needs.
Internal investigations software: Tools for managing workplace and compliance investigations: intake, evidence, interviews, findings and reporting.
ISO 27001: An international standard for information security management. Certification means an accredited auditor confirmed the vendor runs a formal security program.
ISO 42001: An international standard, published in 2023, for managing AI systems responsibly, covering risk assessment, data, transparency and oversight.
MFA (multi-factor authentication): Requiring a second proof of identity, such as an app code or security key, in addition to a password.
Penetration test: An authorized simulated attack on a system by security specialists to find weaknesses before real attackers do.
PII (personally identifiable information): Information that can identify a specific person, such as names, addresses, ID numbers, financial details or health data.
Regulatory change management: Tracking new and changing laws and regulations, assessing their impact and assigning the work needed to comply.
Role-based access control: Granting access to data and features based on a person's role, so people see only what they need.
SOC 2 Type II: An independent audit report showing that a vendor's security controls were designed well and worked over a period of months, usually six to twelve.
SSO (single sign-on): Logging in once with an organization's identity provider, such as Microsoft Entra ID or Okta, to access many applications.
Vendor security questionnaire: A standard set of questions a buyer sends a software vendor about security, privacy, data handling and AI before signing.
Whistleblower hotline: A confidential channel, often anonymous, for employees and others to report suspected misconduct.

Knowledge and Documents

Citator: A tool that shows whether a case or statute is still good law and how later decisions have treated it.
DMS (document management system): The system a firm or legal department uses to store, version, secure and search its documents and email, such as iManage or NetDocuments.
Document automation: Turning templates into guided questionnaires that assemble a finished draft from the answers.
Document comparison: Software that compares two versions of a document and shows every insertion, deletion and move.
Enterprise search: Search across many internal systems at once, such as the DMS, email, intranet and matter systems, from one search box.
Knowledge management: Capturing and reusing a firm's or legal team's best work and expertise: precedents, templates, research, know-how and who knows what.
Legal design: Applying design thinking to legal documents and services so they are easier for people to understand and use.
Legal email management: Filing email into the right matter in the document management system, often with predictive suggestions, so it is stored securely with the rest of the file.
Legal research platform: A database of cases, statutes, regulations and secondary sources with search and, increasingly, AI-generated answers with citations.
Matter-centric filing: Organizing documents and email by client and matter rather than by individual, so everything about a matter is in one place.
Metadata: Data about a document or email, such as author, dates, recipients and edit history, stored alongside its content.
Metadata scrubbing: Removing hidden information such as comments, tracked changes and author data from documents before they leave the firm.
Precedent bank: A curated collection of high-quality example documents a firm or team reuses as starting points for new work.
Version control: Keeping a record of every version of a document and who changed it, so earlier versions can be compared or restored.

Intellectual Property

AI patent drafting: Using AI tools to produce first drafts of patent claims, specifications and office action responses.
IP docketing: Recording and tracking deadlines for patent and trademark applications and registrations across patent and trademark offices worldwide.
IP management software: Software that tracks an organization's patents, trademarks and other IP assets, their deadlines, costs and ownership.
Patent analytics: Analyzing patent data to understand competitors, technology trends, portfolio strength and licensing opportunities.
Patent annuities: Periodic fees paid to keep patents and applications in force, due at different times in each country.
Prior art search: Searching patents, publications and products to find earlier disclosures relevant to whether an invention is new and non-obvious.
Trademark watch: Monitoring trademark filings and the market for marks that may conflict with your own.

Platforms and the Market

Access to justice technology: Technology that helps people without lawyers understand and resolve legal problems, such as guided forms, online dispute resolution and legal information tools.
API: An application programming interface: the way two software systems exchange data, for example a CLM sending signed contracts to a document management system.
Data migration: Moving existing data, such as matters, documents, contracts or billing history, from an old system into a new one.
Funding round: A stage of investment in a startup, such as seed, Series A or Series B, in which investors buy equity to fund growth.
Implementation: The work of setting up a new system: configuration, data migration, integrations, testing and training.
Integration: A built connection between two systems so data moves between them automatically, such as between practice management and accounting.
Legal engineer: A professional, often legally trained, who designs and builds technology solutions for legal work, such as automations, workflows and AI tools.
Legal innovation team: A group in a law firm or legal department responsible for finding, testing and rolling out new technology and ways of working.
Legal technology (legal tech): Software and services built for legal work: running a practice, drafting and reviewing documents, managing contracts, researching the law, handling discovery and running a legal department.
Multi-tenant vs single-tenant: In multi-tenant software many customers share the same application and infrastructure with their data kept separate; in single-tenant software each customer gets its own instance.
No-code: Tools that let people build forms, workflows and automations with visual builders instead of writing code.
ODR (online dispute resolution): Resolving disputes through online platforms, from negotiation and mediation tools to fully online courts and tribunals.
On-premises: Software installed on an organization's own servers rather than hosted by the vendor in the cloud.
SaaS (software as a service): Software you use in a browser on a subscription, hosted and updated by the vendor.
Venture debt: Loans to venture-backed companies, usually alongside equity rounds, to extend runway without selling more shares.
Workflow automation: Automating the steps of a process, such as routing a request, collecting approvals or sending reminders, so they happen without manual chasing.

General information, not legal advice. Written by The Circle Desk.