How to Master AI Terminology: Build and Roll Out the Only Glossary Your Firm Will Need This Year

Ambiguity costs time and invites risk—especially when attorneys, paralegals, and staff use different words to describe the same AI concepts. This tutorial shows small and boutique law firms, legal professionals, and operations leaders how to create a firm-wide AI glossary in Microsoft 365 that makes prompting consistent, speeds client onboarding, reduces training time, and strengthens risk controls. You’ll learn how to structure your glossary in Microsoft Lists, govern updates, import a ready-to-use term set, and connect the glossary to daily work in Teams, Word, and Outlook. By the end, you’ll have a living knowledge asset that grounds AI use in clear language, legal context, and documented guardrails.

Table of Contents

• Prerequisites / What You’ll Need
• Step 1: Choose and Create Your Glossary Home in Microsoft 365
• Step 2: Set Governance, Roles, and Review Cadence
• Step 3: Populate the Glossary with the 2026 Starter Pack
• Step 4: Connect the Glossary to Daily Work (Teams, Word, Outlook)
• Step 5: Teach Attorneys to Use the Glossary in Prompts and Workflows
• Step 6: Maintain, Audit, and Measure Adoption
• Troubleshooting: Roadblocks and Solutions
• Success Checklist
• Conclusion & Next Steps

Prerequisites / What You’ll Need

• Microsoft 365 Business Standard/E3/E5 (with SharePoint Online and Teams)
• SharePoint Site Owner permissions for a practice hub or knowledge site
• Optional: Power Automate for approvals/notifications
• Stakeholders: 1 attorney SME per practice area, Operations/Knowledge Manager, IT/Risk
• Approximate time: 2 hours to stand up + 30 minutes/month to maintain

Step 1: Choose and Create Your Glossary Home in Microsoft 365

Your glossary should live where people already work. A SharePoint site with a Microsoft List gives you structured fields, version history, permissions, and quick filtering. We’ll create a list named “AI Glossary” and define fields that make legal context explicit.

1.1 Create the SharePoint list

1. In SharePoint, go to your firm’s knowledge or operations site. Select New > List > Blank list.
2. Name: AI Glossary. Description: “Firm-approved definitions, prompts, and risk notes for AI use.” Save to the current site.

1.2 Add columns for legal context and governance

1. Use “Title” for Term.
2. Add columns:
   • Plain-English Definition (Multiple lines of text)
   • Legal-Specific Notes (Multiple lines of text) — cite privilege, confidentiality, jurisdictional nuance
   • Approved Prompt Snippet (Multiple lines of text)
   • Risk & Controls (Choice): Low, Medium, High
   • Status (Choice): Proposed, In Review, Approved, Retired
   • Synonyms/Variants (Multiple lines of text)
   • Practice Area Tags (Choice, allow multiple): Litigation, Corporate, Employment, IP, Real Estate, Regulatory, Privacy, Other
   • Owner (Person)
   • Review Date (Date)
   • Source/Policy Link (Hyperlink)
3. Turn on Versioning in List Settings > Versioning settings.
4. Create Views:
   • All Approved (Status = Approved)
   • Needs Review (Status = Approved and Review Date in the past, or Status ≠ Approved)
   • High-Risk (Risk & Controls = High)

Pro-Tip: Keep “Term” human-first. If your attorneys say “client confidentiality” more than “PII,” make “Client Confidential Information” the primary term and list “PII” as a synonym when appropriate.

Step 2: Set Governance, Roles, and Review Cadence

Your glossary is a control, not just a dictionary. Treat it like a policy artifact with a light but traceable process.

2.1 Assign roles

• Owner: Knowledge/Operations Manager (keeps cadence, configures list, reports)
• Attorney SMEs: 1 per practice area (drafts definitions, legal notes, prompt snippets)
• IT/Risk Reviewer: sanity-checks security, data handling, and tooling language
• Approver: Managing Partner or delegate for final “Approved” status

2.2 Define the workflow (manual or automated)

1. Submission: Propose a new term in the list (Status = Proposed, Owner = SME).
2. SME Review: Update definition, legal notes, synonyms. Set Status = In Review.
3. Risk/IT Review: Validate “Risk & Controls” guidance (e.g., redaction required before RAG). Add notes.
4. Approval: Finalize and set Status = Approved. Add Review Date = quarter end + 90 days.

2.3 Publish access and change control

• Contributors: SMEs and Owner (edit). Firm-wide: Read.
• Enable alerts for the “All Approved” view so attorneys see updates without noise.
• Quarterly Review: Filter by Review Date in the past and cycle updates within 30 days.

Note: Don’t over-engineer. A simple “Proposed → In Review → Approved” path with date-based reviews is enough for most boutiques. Add automation later if needed.

Step 3: Populate the Glossary with the 2026 Starter Pack

Use Grid view (Quick Edit) to paste terms fast. Below are firm-ready definitions. Adapt language to your jurisdiction, client base, and matter types.

3.1 Core concepts

• Artificial Intelligence (AI): Computer systems that perform tasks requiring human-like reasoning or pattern recognition. Legal note: Use when describing tools in client engagement letters and policies.
• Generative AI (GenAI): Models that create new content (text, images, code) from learned patterns. Legal note: Clarify use restrictions for client data and outputs.
• Large Language Model (LLM): A model trained on vast text data to predict the next token (word piece). Legal note: Treat as an assistive tool; attorney remains accountable.
• Multimodal Model: A model that accepts or produces multiple data types (e.g., text and images). Legal note: Confirm handling for images that may contain PHI/PII.
• Token: A chunk of text the model processes. Legal note: Token counts affect cost and privacy exposure; minimize sensitive content.
• Context Window: The amount of information the model can consider at once. Legal note: May truncate long exhibits; confirm summarization boundaries.
• Embedding: A numeric representation of text used for search and similarity. Legal note: Store embeddings in approved systems only.
• Vector Database: A store optimized for embeddings and similarity search. Legal note: Confirm data residency and encryption.
• Retrieval-Augmented Generation (RAG): Pulls relevant documents via search and feeds them to the model for grounded answers. Legal note: Redact and control source sets.
• Fine-Tuning: Training a model further on a specific dataset. Legal note: Never fine-tune on client-confidential or privileged content without written authorization and approved environment.

3.2 Prompting and quality

• Prompt: The instruction or question given to a model. Legal note: Avoid inserting client names unless necessary and approved.
• System Prompt: Hidden or fixed instructions that set model behavior. Legal note: Maintain a version-controlled library for matter types.
• Prompt Engineering: Techniques for writing prompts to get accurate results. Legal note: Use approved snippets to reduce risk.
• Prompt Injection: A malicious instruction attempting to override rules. Legal note: Treat as a security risk; review links and attachments before use.
• Jailbreak: Attempts to bypass model safeguards. Legal note: Prohibited in firm systems.
• Hallucination: A confident but incorrect output. Legal note: Mandatory human verification before client use.
• Grounding: Citing specific, authoritative sources to support an output. Legal note: Use matter files or public primary law; preserve citations.
• Temperature: Controls randomness in outputs. Legal note: Keep lower for compliance work, higher for brainstorming.
• Top-p (Nucleus Sampling): Limits token selection to the most probable subset. Legal note: Combine with temperature for consistent drafting.
• Function Calling / Tool Use: The model requests structured actions (e.g., look up a case). Legal note: Log calls for audit.
• Agent: A system that plans steps and uses tools to achieve goals. Legal note: Keep agents in sandboxed environments with strict scopes.

3.3 Data, privacy, and risk

• Personally Identifiable Information (PII): Data that can identify a person. Legal note: Redact or mask before external processing.
• Protected Health Information (PHI): Health-related PII. Legal note: Apply stricter controls; confirm client agreements.
• Confidential Information: Nonpublic info protected by contract or ethics. Legal note: Avoid in open models unless explicitly allowed.
• Attorney–Client Privilege: Confidential communications for legal advice. Legal note: Never expose to third-party systems without privilege-preserving agreements.
• Work Product Doctrine: Materials prepared in anticipation of litigation. Legal note: Store only in protected repositories.
• Data Residency: Where data is stored/processed. Legal note: Confirm jurisdictional requirements in engagement letters.
• Data Retention: How long data is kept. Legal note: Align glossary updates with retention policy and legal hold exceptions.
• De-Identification / Anonymization: Removing identifiers to protect privacy. Legal note: Validate risk of re-identification.
• Redaction: Removing sensitive text from documents. Legal note: Apply before RAG ingestion.
• Content Filter / Safety Filter: Blocks unsafe outputs. Legal note: Keep enabled; document overrides.
• Guardrails: Controls that restrict model behavior and data flows. Legal note: Capture guardrails in “Risk & Controls.”

3.4 Operations and Microsoft 365 terms

• Microsoft 365 Copilot: AI features in Microsoft apps to draft, summarize, and analyze. Legal note: Check your tenant’s data access boundaries.
• Teams Message Extension: A way to search and insert glossary items in chats. Legal note: Prefer linking to the approved entry.
• SharePoint: Firm intranet and document management backbone. Legal note: Use site-level permissions aligned to least privilege.
• Microsoft Purview DLP: Data Loss Prevention policies that restrict sharing. Legal note: Tag glossary pages and exports with sensitivity labels.
• Sensitivity Labels: Metadata that enforces encryption and handling rules. Legal note: Mark the glossary list as “Internal.”
• eDiscovery / Legal Hold: Preservation and search for matters. Legal note: Glossary revisions may be discoverable; maintain version history.
• Audit Log: Records of user and admin actions. Legal note: Retain change logs to glossary entries.
• Access Control (Least Privilege): Grant minimum rights needed. Legal note: Read for firm, Edit only for SMEs/Owner.
• Microsoft Entra ID (SSO): Identity and single sign-on. Legal note: Require MFA for editors.
• API / SDK: Interfaces to integrate the glossary with apps. Legal note: Apply app consent reviews.
• Cost per Token: The billing unit for many LLMs. Legal note: Avoid long sensitive prompts to control spend and exposure.
• Latency / Throughput / Rate Limits: Speed, volume, and caps for requests. Legal note: Plan batch jobs for off-peak hours.

Pro-Tip: Add a short, example-friendly “Approved Prompt Snippet” to each term. Attorneys will paste these into real prompts and get more reliable, on-policy outputs.

Step 4: Connect the Glossary to Daily Work (Teams, Word, Outlook)

Visibility drives adoption. Put the glossary where attorneys draft, review, and collaborate.

4.1 Pin the glossary in Teams

1. Open your practice Team. Click the “+” tab and add SharePoint or List.
2. Select the AI Glossary list and pin the All Approved view.
3. Post a channel announcement: “Use this tab when drafting prompts and client communications.”

4.2 Surface in Word and Outlook

1. Create a one-page “Using the AI Glossary” quick reference with links to common terms.
2. Add it to your Word and Outlook templates (e.g., new memo, engagement letter, discovery checklist).
3. Include a footer line in templates: “This document uses firm-approved AI terminology; see AI Glossary for definitions.”

4.3 Use it with Microsoft 365 Copilot and other assistants

• When prompting, include: “Use definitions from the firm’s AI Glossary for ‘RAG,’ ‘client confidential information,’ and ‘attorney–client privilege.’”
• For discovery summarization: “Ground answers in documents tagged ‘Exhibit A–F’ and the glossary definition of ‘work product.’”
• For client onboarding: “Draft a 1-page explanation of our AI usage aligned to the glossary terms ‘Generative AI,’ ‘confidential information,’ and ‘guardrails.’”

Pro-Tip: Add the glossary link to your Teams channel description and pin a message with search instructions like: “Type a keyword, then filter by ‘Synonyms/Variants’.”

Step 5: Teach Attorneys to Use the Glossary in Prompts and Workflows

Training turns vocabulary into results. Run a 45-minute session focused on high-value legal scenarios.

5.1 Workshop agenda (45 minutes)

1. 10 min: Why glossary-first prompting reduces risk and speeds drafting.
2. 20 min: Live demos:
   • Discovery: Summarize 200 PDFs using RAG with glossary-defined “confidential information.”
   • Contract review: Extract obligations; use glossary “hallucination” and “grounding” checks.
   • Client communication: Draft an onboarding note referencing “guardrails” and “privilege.”
3. 10 min: Hands-on: Attendees paste approved prompt snippets and compare results.
4. 5 min: Q&A and commit to one improvement per matter type.

5.2 Example approved prompt snippets

• Discovery summarization: “Using the firm’s AI Glossary definitions of ‘confidential information’ and ‘work product,’ summarize Exhibits B–D. Ground every conclusion with a file name and page reference.”
• Contract analysis: “Apply the glossary definitions of ‘RAG’ and ‘guardrails.’ In your answer, cite the exact clause text and identify missing controls.”
• Client explainer: “Write a plain-English paragraph using our glossary for ‘Generative AI’ and ‘hallucination,’ suitable for a welcome packet.”

Note: Add your jurisdiction and client industry (e.g., healthcare, fintech) to prompts to tighten accuracy and mitigate irrelevant references.

Step 6: Maintain, Audit, and Measure Adoption

Your glossary should evolve with case law, vendor updates, and client expectations. Make updates predictable and measurable.

6.1 Quarterly review cycle

1. Filter “Needs Review” view and assign SMEs by practice area.
2. Update definitions, prompt snippets, and risk notes where client contracts or regulations changed.
3. Set new Review Dates and capture change rationale in the item’s Version Comments.

6.2 Metrics that matter

• Adoption: Teams tab visits, unique viewers, and search queries/week.
• Quality: Percentage of drafts citing sources when required by the glossary.
• Risk: Number of blocked attempts to paste sensitive data into prompts (from DLP).
• Content Health: Count of Approved terms > 40; stale entries < 5% per quarter.

6.3 Backups and retention

• Export CSV quarterly and store in your “Knowledge Assets” library with sensitivity labels.
• Use SharePoint retention policies aligned to your firm’s record categories.
• If you retire a term, keep the record (Status = Retired) to preserve search history and auditability.

Pro-Tip: Create a “What changed this quarter” news post that highlights 5 terms with before/after definitions and why it matters to clients.

Troubleshooting: Roadblocks and Solutions

Success Checklist

• SharePoint “AI Glossary” list created with required columns, versioning, and three views (All Approved, Needs Review, High-Risk).
• Roles assigned: Owner, SMEs, Risk/IT Reviewer, Approver. Permissions scoped to least privilege.
• At least 40 Approved terms with legal notes, risk levels, and prompt snippets.
• Synonyms populated for top 20 terms to reduce search friction.
• Glossary pinned as a Teams tab and linked in Word/Outlook templates.
• Quarterly Review process running with Review Dates set and change logs captured.
• DLP and sensitivity labels applied to the list and CSV exports.
• Training delivered; attorneys can reference the glossary directly in prompts.
• Adoption and quality metrics tracked monthly; “What changed this quarter” post published.

Conclusion & Next Steps

With a governed, searchable AI glossary embedded in Microsoft 365, your firm speaks a common language about models, prompts, and risk. Attorneys draft faster, privilege stays protected, and clients get clearer explanations of how AI supports their matters. Next, expand impact by adding matter-specific sub-terms (e.g., “HIPAA minimum necessary,” “MAE clause” for M&A), integrating the glossary with knowledge bases and contract playbooks, and wiring notifications into onboarding checklists. Revisit definitions quarterly, publish a “term of the week,” and ask every team to contribute one new or improved term per month. Consistent language is leverage—turn it into a durable advantage across discovery, contract review, investigations, and client advisory work.

Ready to explore how you can leverage technology and AI? Reach out to info@legalgpts.com today for expert guidance and tailored strategies.