WRITER Glossary

Terms by category

AI Basics

Large Language Model (LLM)

Think of this as: The core engine or "brain" of the AI.

What it is: A type of artificial intelligence trained on vast amounts of text data to understand, generate, translate, and reason with human language. WRITER uses its own family of state-of-the-art LLMs called Palmyra, which are purpose-built for enterprise work.

Prompt

Think of this as: The instructions or questions you give to the AI.

What it is: The text input you write to tell the AI what you want it to do. A good prompt is specific, clear, and describes the desired outcome.

Hallucination

Think of this as: A confident mistake or "made-up" information.

What it is: A phenomenon where an AI model generates incorrect, fabricated, or unrealistic information that sounds highly convincing. We prevent hallucinations by grounding our models in your company's actual data using the Knowledge Graph. Additionally, you can ask WRITER Agent to cite it's sources or instruct WRITER Agent to check it's work to make sure it didn't make anything up.

Session

Think of this as: A single, continuous conversation or workspace with the AI.

What it is: An individual interaction thread with the WRITER Agent. Each session maintains its own context, files, and history so you can work on a specific task without starting over each time.

Deliverable

Think of this as: The final product or output of your work.

What it is: The completed asset, report, document, or presentation generated by the AI that you can save, share, or copy into other tools.

Agentic AI / AI Agents

Think of this as: AI that can act, plan, and use tools on its own.

What it is: AI systems designed to act as autonomous or semi-autonomous teammates. Instead of just answering questions, they can set goals, plan a sequence of actions, use external tools, and adapt based on feedback to complete complex tasks end-to-end.

The core experience

WRITER Agent

Think of this as: Your primary AI teammate.

What it is: A single, intelligent interface where you collaborate with AI to get work done.

Chat Mode vs. Agent Mode

Think of this as: Quick tasks vs. end-to-end automated workflows.

What it is:

  • Chat mode is optimized for quick, conversational tasks that don't require a complex setup; like summarizing a short text or drafting a quick email.
  • Agent mode is fully autonomous, executing multi-step workflows from start to finish by utilizing specialized skills and playbooks to achieve a specific outcome.

My Work

Think of this as: Your personal, persistent library of AI deliverables.

What it is: A centralized space where the outputs generated by your agents live and persist across sessions. You can use it to organize past deliverables, share completed work with your team, or pull previous outputs into a brand-new session to maintain project continuity.

Projects

Think of this as: Dedicated workspaces for specific campaigns or initiatives.

What it is: A way to organize your WRITER Agent sessions. When you create a project, you can apply custom instructions, upload reference files, and set a specific voice profile so that every session within that project produces consistent, tailored outputs.

Building Blocks of AI Workflows

Playbooks

Think of this as: Saved, repeatable recipes for complex tasks.

What it is: A set of custom instructions that tells the WRITER Agent exactly what steps to take, and in what sequence, to complete an entire workflow. Playbooks help you automate and scale unstructured, repetitive tasksā€”like generating weekly reports, analyzing sales data, or onboarding new clients.

Skills

Think of this as: The specialized "expertise" or micro-credentials your agent can use.

What it is: Reusable units of execution that capture your team's unique way of doing a specific task. A skill encapsulates specific logic, such as how to analyze a dataset, format a particular document, or apply compliance rules, ensuring the task is done correctly every single time.

Triggers

Think of this as: An automated "start button" or event that kicks off a workflow.

What it is: Rules that define when WRITER Agent should start executing a playbook. Triggers can be time-based (e.g., "run this report every Monday at 9:00 AM") or event-driven (e.g., "run this workflow whenever a new email arrives"), allowing automation to happen in the background without manual intervention.

Integrations and data grounding

Connectors

Think of this as: Secure bridges between WRITER and your other work apps.

What it is: Prebuilt integrations that give the WRITER Agent a safe, standardized way to interact with external tools (like Jira, Asana, or PitchBook) and data platforms (like Snowflake or Databricks). Connectors allow the agent to retrieve real-time data or update systems directly through natural language.

Knowledge Graph

Think of this as: Your company's shared brain and source of truth.

What it is: A system that provides factual grounding for everything your agent does. It securely connects your company's internal documents, structured data, and context so that the WRITER Agent retrieves accurate, reliable, and audit-ready information without hallucinating.

Voice Profiles

Think of this as: Your brand's unique writing style and personality settings.

What it is: Custom configurations that define the tone, style, and vocabulary WRITER Agent should use when generating text. Voice profiles ensure that all outputs sound like they were written by your company, maintaining consistency across all communications.

Custom Instructions

Think of this as: Your personal preferences or rules for how the AI should behave.

What it is: Specific guidelines or rules you apply to a session or project to tailor WRITER Agent's behavior and outputs to your exact needs, such as "always write in short paragraphs" or "format outputs as a table".

Terms

Think of this as: Your company's approved dictionary and banned-word list. W

hat it is: A feature in AI Studio that allows organizations to manage, enforce, and share company-specific terminology.

Snippets

Think of this as: Reusable text shortcuts for common phrases.

What it is: Reusable blocks of text that you can create for frequently used messaging, allowing you to insert standardized language quickly and accurately into your documents or chats without typing them out every time.

Administration, governance, and aecurity

AI Studio

Think of this as: The control tower for your company's AI ecosystem.

What it is: An enterprise-grade hub where administrators observe, connect, and control every aspect of their WRITER deployment, from managing user permissions and models to monitoring security, compliance, and agent usage.

Palmyra LLMs

Think of this as: The state-of-the-art engine powering WRITER.

What it is: WRITER's core family of large language models (LLMs) purpose-built for enterprise work.

Model-Agnostic / Model Choice

Think of this as: The freedom to use the best model for the job.

What it is: While Palmyra handles the majority of your tasks, WRITER is model-agnostic. This means your organization can bring and use third-party models from providers like Amazon Bedrock or NVIDIA NIMs directly within the same governed platform.

AI Guardrails

Think of this as: Safety nets and filters that keep AI outputs safe and compliant.

What it is: Rules, filters, and real-time checkers built into the platform that prevent the AI from generating non-compliant, unsafe, or off-brand content.

Role-Based Access Control (RBAC)

Think of this as: Permissions based on your specific role.

What it is: A security method that restricts system access to authorized users based on their role within the organization (e.g., Org Admin, Team Admin, or Member). WRITER comes with the ability to customize roles ensuring your users have access to the right AI tools.

Model Context Protocol (MCP)

Think of this as: A universal connection between AI models to tools and data.

What it is: An open standard protocol that allows AI models to securely and uniformly connect to external data sources, APIs, and tools across different platforms.

Did this answer your question? Thanks for the feedback There was a problem submitting your feedback. Please try again later.