Glossary of AI Meeting Intelligence & Meeting-to-Deliverable Work

A practical reference for turning meeting conversations into useful work: how capture, tasks, drafts, review, sharing, and handoffs fit together.

A–Z Index

A · B · C · D · E · F · G · H · I · J · L · M · N · O · P · Q · R · S · T · V · W · Z

TL;DR: AI meeting tools split into two categories: notetakers that record what was said, and meeting-to-deliverable tools that ship the work - PRDs, tickets, specs, and updates - before the call ends. This glossary defines the vocabulary of the second category.

A

Acceptance Criteria

Acceptance criteria are the specific, testable conditions a feature must meet before a ticket can be considered done. In agile workflows, they are written into Jira or Linear issues so engineers and QA share a definition of “finished.” AI meeting tools like Earmark draft acceptance criteria directly from what was agreed in conversation, so tickets arrive ready-to-dev instead of as one-line placeholders.

See also: Jira Ticket Generation, Linear Issue Generation, Shippable Work

Action Item

An action item is work identified during a conversation. Record an owner and due date only when they have been agreed. If ownership is unresolved, leave it unassigned and make the gap visible.

See also: Decision Log, Follow-Up, Artifact

ADR (Architecture Decision Record)

An ADR is a short document that captures a significant technical decision, its context, and its consequences, so future engineers understand why a system is built the way it is. ADRs are one of the highest-leverage artifacts to generate from engineering meetings, because the reasoning behind a decision usually lives only in the conversation that produced it.

See also: Decision Log, Engineering Spec, Artifact

Agentic Workflow

An agentic workflow is a process in which AI agents take autonomous, multi-step action toward a goal rather than responding to one prompt at a time. In the meeting context, agentic workflows mean task agents that build summaries, follow-ups, and deliverables while the conversation is still happening — the human sets the intent; the agent does the assembly.

See also: Task Agents, Bring Your Own Agent, Loops

AI Artifact

See: Artifact

AI Chief of Staff

An AI chief of staff is a broad description for a system that helps organize context, priorities, and follow-through. Earmark can supply meeting context and drafts to that kind of workflow. The term is not a promise that Earmark independently manages messages, calendars, or the whole working day.

See also: AI Meeting Assistant, Live Agents, Task Agents

AI Meeting Assistant

An AI meeting assistant is software that uses speech recognition and large language models to capture, transcribe, and act on meeting conversations. The category ranges from simple recorders to real-time systems that generate finished work during the call. The key differentiator is output: a transcript and summary at the low end; PRDs, tickets, and decision logs at the high end.

See also: AI Notetaker, Meeting Intelligence, Meeting-to-Deliverable

AI Notetaker

An AI notetaker is a tool that records meetings and produces transcripts and summaries — examples include Otter.ai, Fireflies.ai, Granola, and Fathom. Notetakers optimize for recall of what was said. Their limitation is that a summary is raw material, not the deliverable: someone still has to turn the notes into the PRD, the tickets, or the stakeholder update. The transcript is raw material, not the product.

See also: Meeting-to-Deliverable, Second Shift, Transcript

Air Pockets

Air pockets are gaps in your context — moments where the information you need exists somewhere in past conversations but isn’t retrievable when you need it. Air pockets are why “we discussed this three weeks ago” so often ends in re-litigating the same decision. Meeting memory systems exist to eliminate them.

See also: Meeting Memory, Recall, Source-backed recall

Artifact

An artifact is the structured output of a task in Earmark. It can be a summary, spec, ticket draft, decision record, or follow-up. Review it before sharing it or using it to direct work.

See also: Task, Cards, Shippable Work, Deliverable

“About the Work” Work

“About the work” work is the administrative layer that surrounds real work: writing up what was discussed, formatting tickets, chasing owners, re-explaining context. It is work that documents work rather than advancing it. Product teams drown in it because meetings historically created this layer instead of completing it.

See also: Second Shift, Productivity Theater, Manual Loop Trap

B

Blank-Page Problem

The blank-page problem is the friction of starting a document, prompt, or ticket from nothing. Even with capable AI, someone must decide what to ask for — and that decision cost is why most meeting follow-up never happens. Templates, pre-seeded tasks, and standing loops exist to remove the blank page entirely.

See also: Template, Pre-Seeded Tasks, Loops

Botless Capture

Botless capture records audio from the user's device without adding a bot to the meeting. On Earmark desktop, this includes microphone and system audio. Participants do not need their own Earmark installation. Capture and consent requirements still apply.

See also: Device-Side Capture, Meeting Bot, System Audio Capture, In-Person Capture

Bring Your Own Agent (BYOA)

Using another AI tool with context you choose to provide. Earmark supports this through local transcript files, copied outputs, and Chat handoffs to Claude Cowork and Codex. Each route has its own access and review step.

See also: Local Markdown, The Handoff, Agentic Workflow

C

Capture Session

A capture session is a single continuous recording window in Earmark, started with one click and running up to two hours per session. Sessions can be tied to a calendar event or started ad hoc via Quick Capture. Longer conversations simply span multiple segments.

See also: Quick Capture, Real-Time Transcription, Live Widget

Cards

Cards are the visual unit of output in Earmark: each artifact renders as a card that can be opened fullscreen, refined in Composer, saved as a template, or shared. Cards turn meeting output from a wall of text into discrete, manipulable pieces of work.

See also: Artifact, Composer, Template

Command Menu

The command menu (Cmd/Ctrl + K) is Earmark’s universal search and navigation surface — one place to find any meeting, task, or artifact across your workspace. It is the front door to meeting memory.

See also: Recall, Meeting Memory, Workspace

Composer

Composer is the panel where you make an artifact your own: you describe what you want on the left, and a live preview re-renders on the right as you type. It is not a chat interface — it is a specification surface. If a card is the answer, Composer is where you write the question.

See also: Cards, Prompt Bar, Template

Context Window

A context window is the amount of information a large language model can consider at once when generating output. Meeting-native tools manage the context window deliberately — grounding every artifact in the actual transcript — which is why their output is specific and quotable where generic chatbot summaries are vague.

See also: LLM, Grounding, Evidence-Grounded Output

Conversation Intelligence

Conversation intelligence is the analysis layer on top of captured speech: extracting decisions, owners, risks, commitments, sentiment, and open questions from what was said. It is the step between transcription (what were the words?) and artifact generation (what work do the words imply?).

See also: Meeting Intelligence, Real-Time Transcription, Decision Log

Cursor-Ready Spec

A Cursor-ready spec is meeting output formatted as a specification and code prompt that can be dropped directly into Cursor (or a similar AI coding IDE) to start building. It closes the gap between “we agreed what to build” and “an agent is building it” — often before the meeting ends.

See also: Bring Your Own Agent, v0-Ready Prompt, Engineering Spec

Custom Templates

Custom templates are reusable prompts you save for yourself or your whole workspace, so a task you’ve refined once can be re-run on any future meeting. They are the mechanism by which one good artifact becomes a repeatable workflow.

See also: Template, Workflow, Slash Commands

D

Decision Log

A decision log is a structured record of what was decided, by whom, when, and why — the single source of “what we decided” for a team. Decisions are the most valuable and most frequently lost output of meetings; a decision log generated from the live conversation prevents the “wait, did we decide that?” cycle.

See also: ADR, Action Item, Artifact

Deliverable

A deliverable is a finished unit of work another person can act on without further translation: a PRD a stakeholder can review, a ticket an engineer can pick up, an update an exec can read. The defining test of an AI meeting tool is whether its output is a deliverable or a description of one. Notes are not the deliverable.

See also: Artifact, Shippable Work, Meeting-to-Deliverable

Device-Side Capture

Device-side capture is audio capture performed on the user’s own machine — combining microphone input and system audio — rather than by a bot participant in the meeting. It captures both sides of a conversation even when you’re wearing headphones, requires nothing from other attendees, and is invisible to them by design.

See also: Botless Capture, System Audio Capture, Privacy-First AI

Diarization (Speaker Diarization)

Speaker diarization is the process of determining who spoke when in an audio stream, so a transcript reads as a dialogue rather than an undifferentiated block of text. Accurate diarization is a prerequisite for artifacts that assign owners and attribute decisions to the right people.

See also: Real-Time Transcription, Transcript, Conversation Intelligence

E

Earmark

Earmark is a botless AI meeting assistant for product and engineering teams. It helps turn live conversations into work people can review and use, and keeps saved context available for later questions and other tools.

See also: Botless Capture, Artifact, Talk → Build, Meeting-to-Deliverable

Engineering Spec

An engineering spec is a technical document describing what will be built, how, and under what constraints — requirements, architecture, risks, dependencies, and acceptance criteria. Specs generated from planning conversations capture the technical context that otherwise evaporates when the call ends, without anyone playing scribe.

See also: Cursor-Ready Spec, ADR, PRD

Evidence-Grounded Output

Grounding means using supplied evidence, such as a meeting transcript, to generate an answer or draft. It can reduce unsupported claims, but does not guarantee correctness. Check consequential details against the source.

See also: Grounding, Hallucination, Transcript

Executive Update

An executive update is a concise, decision-oriented summary written for leadership: what changed, what was decided, what’s blocked, what’s next. Generating exec updates directly from the meetings where the information originated eliminates the Friday-afternoon status-writing ritual.

See also: Stakeholder Recap, SCQA, Artifact

F

First Draft Engine

A first draft engine is an AI system used to produce the 70–90% version of a work product, with a human finishing the last 10–30%. The framing matters: the goal is not to remove humans from the work, but to remove the blank page and the assembly labor. Treating your meeting tool as a first draft engine is the fastest path to real time savings.

See also: Blank-Page Problem, Artifact, Human Review

Follow-Up

A follow-up is the set of communications and work items owed after a meeting: recap emails, tickets, updates, scheduled next steps. Follow-up is where meeting value historically leaked — it happened hours or days later, from memory, if at all. Real-time artifact generation moves follow-up inside the meeting itself.

See also: Second Shift, Action Item, Artifact

G

Grounding

Grounding means using supplied evidence, such as a meeting transcript, to generate an answer or draft. It can reduce unsupported claims, but does not guarantee correctness. Check consequential details against the source.

See also: Evidence-Grounded Output, Hallucination, Context Window

H

Hallucination

A hallucination is an AI output that is fluent but false — an invented decision, a misattributed quote, a fabricated commitment. In meeting AI, hallucinations are uniquely damaging because artifacts travel: a hallucinated decision in an exec update becomes organizational fact. Transcript grounding and human review are the standard mitigations.

See also: Grounding, Evidence-Grounded Output, Human Review

The Handoff

The Handoff is the transfer of meeting context from the conversation to the system that will do the work — your agent, your IDE, your ticket tracker. The thesis: your meeting is the prompt; your agent does the work; the tooling’s job is to make the two talk to each other. Earmark’s Handoff essays describe this as the next phase of meeting AI, beyond transcription and summaries entirely.

See also: Bring Your Own Agent, Local Markdown, Agentic Workflow

Human Review

Human review is the deliberate final pass a person makes over AI-generated work before it ships — checking accuracy, tone, and judgment calls the model can’t make. In mature AI workflows, review replaces authoring as the human’s primary role in documentation.

See also: First Draft Engine, Hallucination, Shippable Work

I

In-Person Capture

In-person capture is recording and transcribing a physical, face-to-face conversation — a conference room, a customer coffee, a hallway decision — using device-side audio. It’s a structural advantage of botless architecture: bots can only attend virtual meetings, but a large share of important decisions happen in rooms.

See also: Botless Capture, Device-Side Capture, Capture Session

Incident Command Document

An incident command document is a living record kept during an outage or incident: timeline, current status, owners, decisions, and communications. Generating it live from the incident bridge call means the postmortem starts with a complete, timestamped record instead of a reconstruction.

See also: Decision Log, Template, Artifact

Integration (Copy-and-Paste by Design)

A transfer in which a person copies reviewed output into another tool. In Earmark, this can mean pasting ticket drafts into Linear or a specification into a coding agent. It does not create a native, automatically synchronized connection.

See also: Local Markdown, Bring Your Own Agent, Privacy-First AI

J

Jira Ticket Generation

Drafting proposed Jira issue content from a conversation, including descriptions and acceptance criteria. In the documented Earmark workflow, a person reviews the output and copies it into the tracker.

See also: Linear Issue Generation, Acceptance Criteria, Shippable Work

L

Linear Issue Generation

Drafting proposed Linear issue content from a conversation. Review the scope, requirements, and ownership before copying the work into Linear.

See also: Jira Ticket Generation, Shippable Work, Artifact

Live Agents

Tasks that generate requested outputs from a live meeting. A person chooses the task or template and can enable live updates. The output still needs review before it directs work.

See also: Task Agents, Personas, Real-Time Transcription

Live Widget

The Live Widget is a small floating control that appears on screen during a desktop recording, putting pin and capture actions one click away without bringing the main app to the front. Invisible AI, visible results.

See also: Pins, Capture Session, Quick Capture

LLM (Large Language Model)

A large language model is an AI system trained on vast text corpora to understand and generate language — the engine underneath transcript analysis and artifact generation. The differentiation between meeting tools is rarely the model itself; it is the context the model is given (the transcript, your role, your templates) and the shape of output it is asked to produce.

See also: Context Window, Grounding, No LLM Training

Local Markdown

A plain-text transcript file saved to your computer. Earmark desktop can save these files automatically when a meeting ends. Local files can be used by other tools you authorize; they do not mean transcription and AI processing happen entirely on your device.

See also: Bring Your Own Agent, Obsidian Workflow, The Handoff

Loops

A recurring task that runs on a schedule or trigger. This glossary uses the term for a workflow concept; check current product documentation for Earmark availability and setup.

See also: Workflow, Pre-Seeded Tasks, Blank-Page Problem

M

Manual Loop Trap

The Manual Loop Trap is the cycle of trading deep work for administrative overhead: back-to-back meetings, scattered notes, generic AI summaries, and faulty memory, followed by hours of cleanup to turn it all into usable work. The trap is structural — no amount of personal discipline fixes a loop where conversation and deliverable are separate systems.

See also: Second Shift, “About the Work” Work, Talk → Build

Meeting Bot

A meeting bot is a virtual participant that joins a call to record it — the visible “AI Notetaker has joined” attendee used by tools like Otter.ai, Fireflies, and Fathom. Bots require attendee tolerance, platform permissions, and IT approval; they can be blocked by hosts, and they cannot attend in-person conversations. Botless capture exists as the architectural alternative.

See also: Botless Capture, Device-Side Capture, AI Notetaker

Meeting Intelligence

Meeting intelligence is the broad category of software that extracts structured value from meetings: transcription, analysis, search, coaching, and output generation. The category is stratifying into recording tools (what was said), intelligence tools (what it means), and completion tools (the work is done).

See also: Conversation Intelligence, AI Meeting Assistant, Meeting-to-Deliverable

Meeting Memory

Saved meeting context that people can retrieve and ask questions about. Its coverage depends on what was captured, retained, and shared with the person asking. It is useful context, not a complete account of everything an organization knows.

See also: Recall, Air Pockets, Source-backed recall, Projects

Meeting-Platform Agnostic

Meeting-platform agnostic describes capture that works identically across Zoom, Google Meet, Microsoft Teams, Webex, phone calls, and in-person conversation — because audio is captured at the device, not through a platform-specific bot or plugin. One tool, every conversation, no IT matrix.

See also: Botless Capture, In-Person Capture, Device-Side Capture

Meeting-to-Deliverable

Meeting-to-deliverable is the category of AI tooling whose output is finished work rather than notes: meeting-to-PRD, meeting-to-tickets, meeting-to-prototype, meeting-to-update. It is defined by a simple test — do you leave the call with the deliverable, or with a doc you still have to turn into one? Every AI meeting tool stops at notes; meeting-to-deliverable tools ship the work.

See also: Artifact, Shippable Work, Talk → Build

N

No LLM Training

“No LLM training” is the commitment that customer meeting content is never used to train AI models — your customer data, strategy, and IP stay out of training corpora. Alongside zero raw-audio storage and retention controls, it is a baseline requirement for security-minded teams adopting meeting AI.

See also: No retained raw audio, Privacy-First AI, Retention Controls

Now / Next / Later Roadmap

A Now/Next/Later roadmap is a lightweight prioritization format that groups work by time horizon instead of hard dates — what’s being built now, what’s next, what’s later. It is a common artifact to synthesize from planning meetings because it communicates priority without committing to fictional deadlines.

See also: PRD, Artifact, Meeting-to-Deliverable

No retained raw audio

Earmark does not retain raw meeting audio after transcription. Standard transcripts may still be saved, local files may exist, and external processing providers have their own retention policies.

See also: No LLM Training, Privacy-First AI, Retention Controls

O

Obsidian Workflow

Using local transcript files in an Obsidian vault. Choose an eligible folder inside the vault as Earmark's transcript destination, or add files yourself. Generated artifacts are a separate output and are not promised as part of the automatic transcript file.

See also: Local Markdown, Bring Your Own Agent, Meeting Memory

P

Personas

Personas are advisory roles the AI adopts during a meeting to provide targeted analysis: Strategic Product Manager, Security Sentinel, Devil’s Advocate, Technical Architect, Data-Driven Analyst, Technical Jargon Translator, Diplomatic Gatekeeper. A persona changes the lens, not just the output format — Devil’s Advocate stress-tests your plan while Security Sentinel scans it for risk.

See also: Live Agents, Template, Task

Pins

Pins are markers you drop at key moments during a live meeting — like digital sticky notes on the timeline — and turn into focused artifacts later. Pins are private to you, available globally via Option/Alt + P, and can be included or excluded from any artifact you generate.

See also: Live Widget, Cards, Capture Session

PRD (Product Requirements Document)

A PRD is the document that defines what a product or feature should do and why: problem, goals, requirements, constraints, and success criteria. PRDs are the canonical product-team artifact — and the canonical second-shift burden, historically assembled from meeting notes over days. Meeting-to-PRD workflows draft them from the planning conversation itself.

See also: Engineering Spec, Acceptance Criteria, Meeting-to-Deliverable

Pre-Seeded Tasks

Pre-seeded tasks are tasks queued before a meeting starts, so artifacts begin generating the moment the conversation does. Pre-seeding turns meeting prep into output configuration: walk in with the PRD, tickets, and recap already assigned, and walk out with them drafted.

See also: Task, Template, Workflow, Loops

Privacy-First AI

Privacy-first AI is an architecture in which the strictest data posture is the default rather than an enterprise add-on: no bots visible to participants, no raw audio stored, no training on customer data, retention controls per meeting or per workspace, and encryption in transit and at rest. Built on zero trust, least privilege, and shift-left security.

See also: No retained raw audio, No LLM Training, Temporary Mode, Zero Trust

Productivity Theater

Productivity theater is activity that performs work rather than advancing it: status meetings about status, decks summarizing decks, updates no one reads. Meetings are where productivity theater concentrates — and automating the documentation layer is how teams convert theatrical time back into shipped work.

See also: “About the Work” Work, Second Shift, Manual Loop Trap

Prompt Archaeology

Prompt archaeology is the ritual of digging through old chats, notes, and transcripts to reconstruct enough context to ask an AI for what you need. It is the hidden tax of general-purpose AI tools: the model is capable, but you spend the session excavating the inputs. Meeting-native tools eliminate it by keeping the context attached to the conversation that produced it.

See also: Air Pockets, Translation Work, Blank-Page Problem

Prompt Bar

The prompt bar is where you type tasks in Earmark — free-form requests or slash commands that pull from the template library. It is the interface between what you want and what gets built.

See also: Slash Commands, Task, Composer

Projects

Projects are shared spaces built around a set of meetings, letting a team chat with a whole cohort of conversations as one body of knowledge. Sharing grants query access, not a backstage pass — your private chats stay yours — and the result is institutional knowledge that outlives people.

See also: Meeting Memory, Workspace, Source-backed recall

Q

Quick Capture

Quick Capture starts an ad-hoc recording immediately, even when the conversation isn’t tied to a calendar event — the hallway decision, the surprise customer call, the “got five minutes?” that turns into a roadmap change. Accessible from the toolbar in one click.

See also: Capture Session, Live Widget, In-Person Capture

R

Real-Time Transcription

Real-time transcription is speech-to-text performed as the conversation happens, rather than after a recording is uploaded. It is the enabling layer for everything live: in-meeting artifacts, live agents, pins, and suggestions all depend on the transcript existing while the meeting is still in progress.

See also: Diarization, Transcript, Live Agents

Recall (Instant Recall)

Recall is the ability to reopen any past meeting and pull its transcript, decisions, and artifacts back into view in seconds — or ask a question across all meetings and get a sourced answer. Recall is what converts a meeting archive from storage into leverage.

See also: Source-backed recall, Meeting Memory, Command Menu

Retention Controls

Settings and policies governing what is retained. Earmark distinguishes saved standard meetings, Temporary meetings, deletion recovery, local transcript files, and provider processing policies. Check each separately when choosing a setup.

See also: Temporary Mode, No retained raw audio, Soft Delete

S

Source-backed recall

Asking questions about available meeting history and checking the supporting sources. Good recall makes its evidence and coverage inspectable; it does not guarantee that every conversation or detail was captured.

See also: Recall, Meeting Memory, Air Pockets

SCQA

SCQA (Situation, Complication, Question, Answer) is a structured communication format that orders information the way executives consume it: context, what changed, the question it raises, and the recommendation. It is a popular artifact format for turning meandering discussions into crisp readouts.

See also: Executive Update, Stakeholder Recap, Template

Second Shift

The second shift is the after-hours work meetings create: writing up notes, drafting tickets, composing updates, and chasing follow-ups once the actual meetings are over. It is the clearest symptom of tooling that records conversations instead of completing them — and eliminating it is the core promise of meeting-to-deliverable software.

See also: Manual Loop Trap, “About the Work” Work, Follow-Up

Shippable Work

Work that has been checked and is ready for its intended next step. An AI-generated draft becomes usable when the responsible person confirms its requirements, evidence, and commitments.

See also: Artifact, Deliverable, Acceptance Criteria

Slash Commands

Slash commands are typed shortcuts (/sprint, /ticket, /smart, /acronym) that invoke any template instantly from the prompt bar. They compress “find the right prompt” into a keystroke.

See also: Template, Prompt Bar, Custom Templates

Soft Delete

Soft delete is a deletion model in which removed meetings enter a 30-day recovery window before being permanently purged, with all associated artifacts cascade-deleted. It balances “I deleted that by mistake” against “when I say delete, I mean gone.”

See also: Retention Controls, Temporary Mode, No retained raw audio

Stakeholder Recap

A stakeholder recap is a tailored summary of a meeting written for people who weren’t in it — framed around what they need to know and do, not around chronology. One conversation typically owes several audiences different recaps; generating them in parallel is a hallmark meeting-to-deliverable workflow.

See also: Executive Update, SCQA, Artifact

System Audio Capture

System audio capture is recording the sound your computer plays — the other side of the call — alongside your microphone, so both sides of a conversation are captured even when you’re wearing headphones. It is the technical mechanism that makes botless capture complete.

See also: Device-Side Capture, Botless Capture, Real-Time Transcription

T

Talk → Build

Preparing a reviewed implementation brief from a conversation. The destination tool produces the implementation or prototype.

See also: Cursor-Ready Spec, v0-Ready Prompt, Meeting-to-Deliverable

Task

A task is a prompt that tells Earmark what you need from a meeting — “draft the PRD,” “generate tickets for what we agreed,” “write the exec update.” The formula is simple: task = what you ask for; artifact = what Earmark creates. Tasks can be assigned live, pre-seeded before the meeting, or run afterward against the transcript.

See also: Artifact, Pre-Seeded Tasks, Template, Task Agents

Task Agents

Tasks that generate requested outputs from a live meeting. A person chooses the task or template and can enable live updates. The output still needs review before it directs work.

See also: Live Agents, Task, Agentic Workflow

Template

A template is a pre-built task that tells the AI how to structure an artifact — Earmark ships with 30+, spanning meeting minutes, PRD outlines, ticket generators, incident command documents, executive summaries, and personas. Templates are how good prompts stop being individual craft and become team infrastructure.

See also: Custom Templates, Slash Commands, Workflow

Temporary Mode

A meeting mode that does not save the transcript, tasks, or artifacts to the Earmark account. They become unavailable when the meeting is closed. This is separate from the 30-day recovery period for deleted standard meetings. Desktop local transcript saving and provider retention are separate.

See also: Retention Controls, Privacy-First AI, Soft Delete

Transcript

A machine-generated text record of captured speech. It can contain transcription and attribution errors. Current desktop capture distinguishes your microphone from other meeting audio, but does not identify every other participant by name.

See also: Real-Time Transcription, Local Markdown, Evidence-Grounded Output

Translation Work

Translation work is the labor of converting information from the form it arrived in to the form it’s needed in — meeting notes into tickets, engineer-speak into exec-speak, discussion into documentation. Product managers are drowning in it; it is the single largest category of work that meeting-native AI eliminates.

See also: “About the Work” Work, Artifact, Second Shift

V

v0-Ready Prompt

A brief describing screens, behavior, and design direction for use in v0. The destination tool produces the implementation or prototype.

See also: Cursor-Ready Spec, Talk → Build, Bring Your Own Agent

W

Workflow

A repeatable sequence: choose the desired output, use a template, generate from the conversation, review the result, and move it into the destination tool. A saved template does not by itself establish an automatic connection to that destination.

See also: Template, Pre-Seeded Tasks, Loops, Custom Templates

Workspace

A workspace is a team’s shared container in Earmark: members, roles, shared templates, workspace-level settings like company vision, and admin controls such as enforced temporary mode and retention policy. Meeting content stays isolated per user; the workspace shares the infrastructure, not your private record.

See also: Projects, Retention Controls, Temporary Mode

Z

Zero Trust

Zero trust is a security model that assumes no user, device, or network is inherently trustworthy — every access is authenticated, authorized, and encrypted, with least-privilege permissions throughout. It is part of the modern security foundation (alongside strong authentication and shift-left security) that meeting AI must be built on, given the sensitivity of conversation data.

See also: Privacy-First AI, No retained raw audio, Retention Controls

Frequently Asked Questions

What is an AI meeting assistant?

An AI meeting assistant is software that captures meeting audio, transcribes it, and uses large language models to produce output from the conversation. Basic assistants produce transcripts and summaries; advanced, real-time assistants like Earmark produce finished deliverables — PRDs, Jira/Linear tickets, specs, and updates — while the meeting is still happening.

What is botless meeting capture?

Botless capture records audio from the user's device without adding a bot to the meeting. On Earmark desktop, this includes microphone and system audio. Participants do not need their own Earmark installation. Capture and consent requirements still apply.

What’s the difference between an AI notetaker and a meeting-to-deliverable tool?

An AI notetaker (Otter.ai, Fireflies, Granola, Fathom) optimizes for the record: transcripts and summaries you still have to turn into work. A meeting-to-deliverable tool optimizes for the outcome: the tickets, PRD, and updates are drafted before the call ends. The test is what you leave the meeting with — a doc about the work, or the work.

How do product teams turn meetings into PRDs and tickets automatically?

By pre-seeding tasks before the meeting (or invoking templates during it), a real-time system generates artifacts from the live transcript: a PRD outline structured from the discussion, and Jira or Linear issues with titles, descriptions, and acceptance criteria drawn from what was agreed. The human’s role shifts from author to reviewer.

Is Earmark’s AI trained on my meeting data?

No. Earmark does not train models on customer data, does not store raw audio, and offers retention controls including a fully ephemeral Temporary Mode — configurable per meeting or enforced workspace-wide.

Does Earmark join meetings as a bot?

Never. Earmark captures audio device-side, so no attendee ever sees it, no participant needs an invite or plugin, and hosts have nothing to admit or block. You’re in complete control — connect or disconnect any time.

About This Glossary

This glossary is maintained by Earmark — a botless AI meeting assistant for product and engineering teams. Earmark turns live meetings into shippable work in real time: PRDs, tickets, specs, decision logs, and updates, generated botlessly from the conversation itself. Definitions reflect the vocabulary used across Earmark’s product guide, blog, and comparison library, and are updated as the category evolves.

Want to see the vocabulary in action? Download Earmark and leave your next meeting with the work already done.