Glossary

Version 1.0 · Last updated 2026-06-30

Entries are alphabetical. Where a legacy term was renamed, the entry uses the new term and notes the old one.

---

A

Account

Customer organization profile in Unless used to configure agents, manage users, and oversee interactions across channels. One account can hold multiple workspaces.

Accountability

One of the four pillars of Trust and compliance. Covers the audit trail every Customer Agent decision leaves, the documented incident response plan, and the controls that let you prove what the agent did and why.

Acquisition

Moment on your marketing site where the Customer Agent qualifies, converts, and educates 24/7. One of the four moments.

Action button

Interactive UI element in a component that triggers a configured Procedure or flow.

AI constellation

EU-hosted set of foundation models the platform runs on, from providers like Anthropic, Mistral, OpenAI, Google, Cohere, and Amazon Titan. Unless picks the right model per job, balances load, and switches in real time if regulation or vendor terms change. No single-model lock-in. See also Bring-your-own-LLM.

AI maturity

Analyze dashboard that shows where your Unless rollout sits on the maturity curve and which actions move you forward.

Allowed language

Languages enabled for Customer Agent interaction in an account. Defaults to all languages, but can be limited to the set your organization can support.

Analyze

One of the four phases of the Engine. Covers Performance, Business impact, and AI maturity dashboards.

Answer

Response provided by the Customer Agent or Team Assistant to a question or query.

Audiences

Customer or visitor segments defined by behavioral, demographic, contextual, or engagement data. Used for personalization, access, and authorization. Replaces the legacy term “segments.”

B

Base language

Primary reference language for the Sources that feed Living Knowledge. Usually the language most of your content is written in.

Browser extension

Technology that distributes the Team Assistant into the browsers used by your team.

Bring-your-own-LLM

Option to run the Customer Agent on a model you already license, alongside or instead of the models in the AI constellation. Keeps model choice in your hands.

Business impact

Analyze dashboard that translates Customer Agent activity into revenue, cost savings, and CSAT outcomes.

C

Close

Action to dismiss or exit a component, panel, or conversation.

Compliance controls

Built-in safeguards aligned to GDPR, EU AI Act, DORA, ISO 42001, ISO 27001, and other frameworks. Enforced at platform level.

Component

Interface module deployed into a website, product, or portal. Component types include starter buttons, conversational UI, popups, sideboxes, inline, popover, and multistep. Components may be opened by the Customer Agent when relevant.

Consent management

Process for obtaining, storing, and managing customer permissions for data handling. Required for GDPR-relevant deployments.

Content library

Section in Train where structured content lives, ready for the Customer Agent to use.

Control answer

Predefined response used as a benchmark when validating Customer Agent accuracy in Quality reports.

Control group

Baseline Audience used for A/B testing and performance comparison.

Control question

Question used to verify the accuracy and consistency of Customer Agent responses, paired with a control answer.

Conversation

Structured exchange between a customer and the Customer Agent, tracked for analytics and improvement.

Cookie whitelist

Authorized set of custom cookies passed to Unless for use in Audience targeting.

CRM connector

Integration that syncs CRM data into Living Context so the Customer Agent can personalize based on lifecycle stage, account history, and ownership.

Custom rule

User-defined condition that shapes how the Customer Agent behaves under specific circumstances.

Customer Agent

Customer-facing AI agent that runs across your marketing site, product, portal, help center, and helpdesk. One agent per moment. Replaces legacy terms like “AI assistant,” “chatbot,” and “bot.”

D

Dashboard

Interface where Customer Agent managers configure agents, monitor performance, and act on Inbox Messages.

Data access request

Formal request from an end user to view their stored personal data. Tracked in the dashboard.

Data correction request

Process to update or amend incorrect personal data records for an end user.

Data erasure request

Process to delete all personal data for an end user, in compliance with GDPR and similar regulations.

Deploy

One of the four phases of the Engine. Covers Customer Agent setup, Team Assistant setup, and the four moments.

Downvote

Customer action signaling dissatisfaction or inaccuracy in a Customer Agent response.

DPO

Data Protection Officer. Role assigned in Unless to oversee data protection strategy and compliance.

Draft version

Unpublished iteration of content, Procedures, or configuration still under review.

E

End user

Customer or visitor who interacts with a Customer Agent. Different from the dashboard user, who configures the platform.

Engine

Back-of-house platform behind the agents. Holds the three senses (Living Knowledge, Living Memory, Living Context) and runs the four phases (Train, Test, Deploy, Analyze).

Escalation

Routing a conversation from the Customer Agent to a human - either through live chat, a ticket, or a Procedure that books a meeting.

Expansion

Moment where the Customer Agent catches upsell signals the moment they appear and routes them to the right account owner. One of the four moments. Sometimes written Sales expansion in the dashboard.

F

Fallback language

Default language used when the customer’s preferred language cannot be confidently detected.

Fixed

One of two subscription plans. €1,999 per month by invoice, 12-month minimum, with 5,000 outcomes included, 50 team seats, unlimited audiences, all Flex modules bundled, an SLA, dedicated support, a named account manager, and a Slack channel. Includes a free four-week guided onboarding pilot when you start, plus one month free if you pay annually up front. Best for regulated buyers, finance teams, and procurement-led organizations. See also Flex, Outcome, Pilot.

Flex

One of two subscription plans. €0.99 per outcome with a 100-outcome monthly minimum, 10 team seats, 5 audiences, and email support. Flex modules are à la carte. Best for variable demand and teams testing the waters. See also Fixed, Outcome, Flex modules.

Flex modules

Productized add-ons priced à la carte on Flex and bundled into Fixed. Cover Onboarding, Premium support, Live chat integration, CSAT, SLA, and premium integrations. See also Subscription plans.

Flywheel

Collective term for the four phases of the Engine. The agent sharpens with each pass. Also known as the loop.

Follow-up question

Supplementary question prompted by the Customer Agent to clarify or expand on a previous answer.

G

Generated answer

Customer Agent response based on Living Knowledge, Living Context, and live inputs.

H

Hand-off

Passing a conversation from the Customer Agent to a human with full context. Counts as an outcome and is billed as one, the same way a support member handling and routing a ticket carries a cost. Also called a hand-over. See also Escalation, Resolution.

Help Center

Auto-generated, customer-facing help center surface that Unless publishes from your Living Knowledge. Distinct from this help center (lowercase), which is the knowledge base for managing the platform.

I

Inbox

Dashboard section where the team handles Messages from the Team Assistant and the Engine.

Indexed item

Content cataloged by the Engine’s retrieval index for fast, relevant use by the Customer Agent.

Integration

Connection between Unless and another system (CRM, helpdesk, CMS, MCP server) to unify data flows and execute actions.

K

Knowledge base

Section in Train where Living Knowledge is configured. Includes Sources, Content library, and topics.

L

Limited agency

Safeguard that prevents LLM-based systems from taking unintended actions. Unless enforces user approval for new action categories.

Limited reliance on LLMs

Safeguard that prevents LLMs from making critical decisions without human review. Unless enforces disclaimers and limits the functional purpose of LLMs.

Living Context

One of the three senses. Where the Customer Agent is acting. Connects to your CRM, billing, and product data so the agent can see the situation around each conversation.

Living Knowledge

One of the three senses. What the Customer Agent knows. Self-maintaining library built from your Sources.

Living Memory

One of the three senses. Who the Customer Agent is talking to. Memory layer that remembers customers across sessions.

Loop

See Flywheel.

M

MCP server

Model Context Protocol server. Acts as a bridge between the Customer Agent and any product API. The agent exposes MCP tools as Procedures that the customer can choose to execute.

Messages

Work items in the Inbox. Come from Team Feedback, Engine-detected gaps, and Living Knowledge change proposals. Replaces the legacy term “tasks” in the Inbox sense.

Model Denial of Service (DoS)

Security threat where attackers cause resource-heavy LLM operations to degrade service or inflate cost. Unless mitigates with context window limits and API rate limits.

Moments

Four customer lifecycle moments the Customer Agent serves: Acquisition, Retention, Expansion, and Support. One agent per moment.

O

Optimizer

Engine tool that refines Sources using gap detection, deduplication, and rewriting. Improves Quality scores without changing the original source.

Outcome

The single billing unit across both plans. An outcome is either a resolution, when the agent answers the customer to their satisfaction, or a clean hand-off, when the agent passes the conversation to a human with full context. You pay for work a person would otherwise have done, never for misfires. See also Resolution, Hand-off.

Outcome-based pricing

Pricing model where you pay for results the Customer Agent and Team Assistant produce, not seats or conversations.

P

Performance

Analyze dashboard showing conversations, responses, feedback, common questions, top Sources, and component activity.

Persona

Predefined profile that shapes the Customer Agent’s tone of voice and focus (support, sales, informative, or custom).

PII

Personally Identifiable Information. Data such as names, emails, and identifiers that can uniquely identify a person.

PII filtering

Automatic detection and removal of personally identifiable data by the Privacy Vault before it reaches any model.

PII filtering whitelist

Approved list of data points exempt from filtering for specific operational needs.

PII obfuscation

Masking of identifiable personal data during processing.

Pilot

Four-week guided onboarding run on your own data. Part of the Fixed plan and free when you start there. The pilot configures the agent across the moment you choose and reports the numbers back at the end, finishing with a go or no-go decision. Stop there and you owe nothing; continue and the 12-month term starts from go-live. See also Fixed.

Preview

Test dashboard where you sit on the customer’s side of a conversation and try the Customer Agent end to end before customers see it.

Primary click

Customer’s main action click recorded inside a component.

Privacy

Protection and respectful handling of customer data. One of the four pillars of Trust and compliance.

Privacy Vault

Twelve numbered privacy measures, including PII filtering, tokenization, data minimization, prompt anonymization, and EU data residency.

Procedures

Agentic actions the Customer Agent and Team Assistant can take on behalf of the customer. Includes follow-up questions, redirects, components, switch topic, custom JavaScript, live chat, and API calls. Replaces the legacy terms “tasks” (in the agentic sense), “agent actions,” and “AI actions.”

Prompt injection

Security threat where attackers craft inputs to manipulate the LLM. Unless mitigates with privilege control, input validation, and cleansing of untrusted input.

Published version

Live iteration of content, Procedures, or configuration visible to customers.

Q

Quality reports

Test dashboard that runs control questions against the Customer Agent and scores each answer.

Query

Structured request for information sent to the Customer Agent. Often equivalent to a question.

Question

Inquiry submitted by a customer to the Customer Agent or through the search interface.

R

Report score

Numerical rating in a Quality report showing how close the Customer Agent’s answer was to the control answer.

Resolution

The agent answering a customer to their satisfaction, with no human needed. One of the two ways to deliver an outcome. See also Hand-off.

Response

Single unit of output provided by the Customer Agent - an answer, recommendation, or content piece - triggered by a customer action or query.

Retention

Moment inside the product where the Customer Agent sees churn coming and acts before it happens. One of the four moments.

S

Sales expansion

See Expansion.

Search

Customer query that returns a list of relevant Sources or answers.

Secondary click

Alternative customer click inside a component, often for additional options or context menus.

Secure output handling

Safeguard that prevents insecure LLM outputs from causing downstream vulnerabilities like XSS or privilege escalation. Unless enforces output validation and sanitization.

Secure plugin design

Safeguard that prevents LLM plugins from exposing remote code execution risks. Unless enforces parameterized inputs and plugin authentication.

Security

Protective measures guarding platform systems, integrations, and customer data. One of the four pillars of Trust and compliance.

Sensitive information protection

Safeguard that prevents the LLM from leaking confidential information. Unless applies sanitization filters and limits external data access for the LLM.

Session

Continuous period of customer interaction, tracked end to end.

Simulations

Test dashboard that runs scripted conversations against the Customer Agent at scale to catch regressions and validate edge cases.

Source

Origin of content ingested into Living Knowledge. Includes website pages, FAQs, files, and integration-pulled content.

Source connector

Integration that retrieves data from an external system to be used as a Source.

Subscription plans

Two plans under outcome-based pricing: Flex (€0.99 per outcome, à la carte) and Fixed (€1,999 per month, full platform, free pilot). Both include the full platform; the difference is commitment, scale, and service level, not basic versus premium. See also Flex, Fixed, Flex modules.

Suggested control question

Accuracy-check question recommended by the Engine to ensure proper Customer Agent calibration.

Supply chain security

Safeguard against risks from third-party model or dataset dependencies. Unless applies a Content Security Policy and validates third-party components.

Support

Moment where the Customer Agent resolves customer issues and the Team Assistant co-pilots the team. One of the four moments.

T

Targeting condition

Rule that specifies criteria for delivering a tailored experience or Audience membership.

Team Assistant

AI-powered tool that helps your team inside the helpdesk they already use. Drafts replies, surfaces context, runs Procedures. Replaces the legacy term “browser assistant.”

Team Feedback

Surface where the team submits content suggestions and reviews automated changes. Feeds the Inbox.

Test

One of the four phases of the Engine. Covers Preview, Simulations, and Quality reports.

Test mode

A/B testing tactic applied to a component to compare a variation against a control.

Three senses

Collective term for Living Knowledge, Living Memory, and Living Context. Together they tell the Customer Agent what it knows, who it is talking to, and where it is acting.

Topic

Thematic category that organizes Sources, conversations, and the slice of Living Knowledge the Customer Agent uses for an answer. Topics help prevent cross-product mix-ups.

Tracking

Monitoring of customer interactions and behavior across sessions.

Tracking whitelist

Approved list of pages where tracking is allowed.

Train

One of the four phases of the Engine. Covers Living Knowledge, Living Memory, Living Context, Sources, Audiences, and Procedures.

Training data poisoning

Security threat where malicious actors insert harmful or biased data into LLM training. Unless mitigates with anomaly detection on Sources.

Translation corpus

Stored dataset used to train translation for multilingual experiences. Used for terms like product names that should not be translated.

Transparency

Clear and open disclosure of how the Customer Agent makes decisions, personalizes, and uses data. One of the four pillars of Trust and compliance.

Trust and compliance

Help center section covering Transparency, Privacy, Security, and Accountability. Aligned to GDPR, EU AI Act, DORA, ISO 42001, ISO 27001, and entities supervised by BaFin and AFM.

U

Unless Studio

Hourly professional services arm. Offers development, custom connectors, legacy-product enablement, use-case discovery, EU AI Act readiness, and AI maturity roadmaps. Billed by the hour, never bundled, available on either plan.

Upvote

Customer action signaling approval or usefulness of a Customer Agent response.

Use case

Defined scenario where Customer Agent capabilities are applied to a business objective. Usually a combination of a domain (website, product) and a target audience (anonymous or identified).

User

Person with authenticated access to the Unless dashboard, with defined roles and permissions to manage content, Procedures, or analytics.

V

Variables

Data points the Customer Agent needs to complete a Procedure. Collected from the customer during the conversation if not already known.

View

Event recorded when a component appears in the customer’s interface.

W

Workspace

Environment that holds all configuration and data for a single customer journey, covering one or more moments. Most companies use one production workspace. Multiple workspaces apply when you run separate products, brands, or markets that need fully independent setups.