Glossary of Common Tech Terms

Agentic AI: Agentic AI systems can complete a series of related tasks without needing a person to prompt each step. Because they act more independently than other GenAI tools, they require strong oversight and clear guardrails. That oversight may occur at defined checkpoints or through ongoing monitoring rather than reviewing every individual action.

Example: AI agents are used in health care to help manage patient follow-up by monitoring information from a patient’s record or connected at-home medical equipment (e.g., blood pressure or glucose monitoring devices), identifying when follow-up may be needed, sending reminders or check-ins to the patient, and alerting care providers when human review or intervention is needed.

Application Programming Interfaces (APIs): Digital connections that allow different software systems to exchange information. APIs let tools share data automatically instead of requiring staff to enter the same information in multiple places.

Example: APIs can connect a legal aid case management system with other tools, such as an online intake form or a referral partner’s case management system, so client data can be shared automatically instead of entered manually.

Automated Systems: Tools or processes that complete tasks automatically based on predefined rules or instructions. Unlike GenAI, these systems typically do not create original content in response to open-ended prompts; instead, they follow logic set in advance by people.

Example: Legal aid organizations can use an automated texting tool to send clients appointment reminders, deadline reminders, surveys, or follow-up messages based on rules set by staff.

Black Boxes: Systems or tools where users cannot readily understand or examine how an output, recommendation, or decision was produced. Concerns about black boxes commonly arise with predictive or automated decision systems, but can also apply to GenAI products when vendors do not disclose relevant information about models, data use, testing, or system dependencies. Limited transparency can make it difficult to evaluate accuracy, explain results, identify bias, or assign accountability.

Example: An automated triage system may steer a person toward or away from legal aid services without clearly showing staff or the affected person which information, rules, or model behavior produced that result. A well-designed system should include documentation, testing, oversight, and explanations proportionate to the risks associated with its use.

Cloud Collaboration Tools: Online tools that allow people to create, share, and edit documents and communicate in real time.

Examples: Microsoft 365, Google Workspace

Data Flows: The ways information moves among people, systems, vendors, and storage locations. Understanding data flows includes identifying what information is collected, where it is sent and stored, who can access it, and how it is retained or deleted.

Example: A legal aid organization may map how information moves from an online intake form to its case management system, document storage, GenAI tools, reporting systems, and referral partners to identify where client data are accessed, shared, or stored.

Ethical Treatment (Data): Collecting, using, sharing, storing, and deleting sensitive or bias-relevant data in ways that protect privacy, respect the people represented by the data, and avoid unfair or discriminatory outcomes.

Example: Before using sensitive or bias-relevant data—such as immigration status, disability, race, primary language, or ZIP code—in an AI-supported intake system, a legal aid organization documents why the data are needed, limits who can access the data and how they may be used, evaluates whether their use could produce unfair outcomes, and establishes when they should be deleted.

Foundational Systems/Processes: The core tools and workflows an organization depends on every day, such as email, document storage, case management, and security practices. Strong foundations are necessary before adopting more advanced tools like GenAI.

Examples: Foundational systems may include a reliable case management system, secure document storage, staff email, device management, and clear processes for storing, finding, and reporting on client information.

Generative Artificial Intelligence (GenAI): A type of artificial intelligence that generates content, such as text, images, or code, in response to user inputs. Other types of AI may classify information, predict outcomes, or automate decisions.

Examples: GenAI includes large language models (LLMs), such as OpenAI’s GPT models and Anthropic’s Claude models, and text-to-image models, such as DALL·E.

Hallucinations: When a GenAI system produces information that seems accurate but is incorrect, made up, or unsupported.

Example: A GenAI tool citing a legal case that does not actually exist.

Human in the Loop (HITL): A process where a person reviews or approves GenAI-generated outputs before they are used.

Example: Staff reviewing a GenAI-drafted client letter, intake summary, or legal research output before it is used or shared.

Identity Management: Tools and policies that control who can access systems and data within an organization.

Example: A legal aid organization might use identity management to give intake staff access to intake records, case handlers access to their assigned matters, and pro bono volunteers limited access to relevant case files—while automatically removing access when someone leaves.

IT Modernization: Updating outdated technology, systems, and processes to improve security, reliability, and compatibility with newer tools, including GenAI.

Examples: IT modernization might include moving from paper files and desktop-installed software to secure cloud-based systems, standardized devices, multifactor authentication, and tools that can integrate with one another.

Large Language Models (LLMs): GenAI models trained on large amounts of text that can generate, summarize, and analyze language. LLMs often power GenAI assistants and applications that create new content, such as text, images, or code, in response to user inputs.

Examples: OpenAI GPT-based models and Anthropic Claude models are examples of LLMs. ChatGPT, Claude, and Microsoft Copilot are GenAI assistants or applications powered by LLMs or other GenAI models.

Legacy IT Systems/Policies: Older systems or rules that no longer meet current needs and can make it harder to adopt new tools or integrate with other systems.

Examples: A legacy system might be an older case management platform that cannot easily connect with online intake forms or reporting dashboards. A legacy policy might require staff to store client documents on a local shared drive instead of a secure cloud-based document system.

Life Cycle Risks: Risks that can arise at different stages of a technology system’s lifecycle, including its selection or design, testing, deployment, ongoing use, modification, and retirement. Risks may change over time as the system, its users, its data, or the surrounding environment changes.

Example: A legal aid organization may test a GenAI intake tool before deployment, but new risks may arise if the vendor later changes the underlying model, the organization expands the tool’s use, or the tool's performance changes over time.

Managed Service Provider (MSP): An outside company that manages and supports some or all of an organization’s technology systems on an ongoing basis. Services may include technical support, cybersecurity, cloud and network management, device management, software updates, data backups, and incident response.

Example: A legal aid organization without internal IT staff might hire an MSP to manage employee devices and accounts, maintain cybersecurity protections, and provide day-to-day technical support.

Modular Systems: Technology systems built so that individual parts can be added, replaced, or updated without rebuilding everything.

Example: A legal aid organization could connect its case management system with a separate document assembly tool, such as Docassemble, so staff can generate documents from case data while still utilizing other functions of the case management system.

Multifactor Authentication (MFA): A security method that requires more than one type of verification—such as a password plus a onetime code, authentication app, or security key—before someone can access an account or system.

Example: A staff member enters their password to log into the case management system and then approves the login through an authentication app or enters a one-time code sent to their phone.

Prompt Injection: An attempt to manipulate a GenAI system by placing instructions in a prompt, document, webpage, or other information that the system processes. These instructions may cause the system to disregard its intended rules, reveal protected information, or take unintended actions.

Example: A document uploaded for summarization contains hidden instructions telling the GenAI tool to disregard its safeguards and reveal information from other files it can access.

Retrieval-Augmented Generation (RAG): A method that allows a GenAI system to search or retrieve information from designated sources—such as organizational policies, legal materials, or a document repository—and use that information when producing a response. RAG can make outputs more relevant and traceable, but it does not by itself guarantee accuracy, completeness, or appropriate access to information.

Example: A legal aid organization could use a RAG-based internal knowledge assistant to answer staff or volunteer questions about organizational policies, intake procedures, or legal resources. The knowledge assistant would search through an approved collection of documents and generate a response based on those materials.

Secure Infrastructure: The technology foundation that protects an organization’s systems, data, users, and operations from security risks. This may include secure networks, devices, cloud platforms, access controls, monitoring, policies, and regular backups. Backups are saved copies of important data or systems that can be restored if information is lost, damaged, deleted, or locked by a cyberattack.

Examples: A legal aid organization may use multifactor authentication, encrypted systems, regular backups, and security monitoring to help prevent unauthorized access to client data and recover critical information if a system is disrupted.

System Dependencies: When one system, tool, or workflow relies on another to function. Some dependencies are necessary and expected, but too many dependencies—or poorly managed dependencies—can make systems harder to update, replace, troubleshoot, or govern.

Example: A GenAI tool may become embedded across intake, case notes, reporting, and document drafting workflows. If the tool becomes unreliable, too expensive, or unavailable, the organization may have difficulty replacing it without disrupting daily operations.

Technology Sprawl: When an organization uses too many tools that overlap or do not connect well, leading to inefficiencies and increased risks.

Example: A legal aid organization may use separate tools for intake, client texting, document drafting, reporting, and GenAI-assisted research that do not connect with one another. Staff may have to enter the same information in multiple places, manage different access rules, and track work across disconnected systems.

Tool Rationalization: A structured process for reviewing an organization’s tools to identify overlapping functions, unmet needs, costs, risks, and system dependencies. The process helps determine which tools should be retained, consolidated, replaced, or retired.

Example: A legal aid organization may review its GenAI subscriptions and consolidate overlapping tools based on staff needs, security and privacy protections, accessibility, system integration, and cost.

Vendor Lock-In: When an organization becomes dependent on a specific vendor, making it difficult or costly to switch to another option.

Example: A legal aid organization may build intake, document drafting, reporting, and GenAI features around one vendor’s platform. If the vendor raises prices, changes key features, discontinues a product, or ceases operating, switching to another system may require major cost, staff retraining, data migration, and workflow redesign.

Webhooks: Automated system-to-system updates sent when a specific event occurs. APIs create the connection between systems; webhooks use that connection to notify another system that something has happened, without requiring staff to manually check for updates.

Example: When a potential client submits an online intake form, the intake platform can send a webhook to the case management system to create a new intake record, notify staff, or trigger a follow-up workflow.