Getting started
Key concepts
Version 1.0 · Last updated 2026-05-27
Unless uses a small set of concepts that recur across every section. Once you know them, the rest of the help center clicks into place quickly.
The agent side
Customer Agent
The customer-facing AI agent. Runs on your marketing site, product, portal, help center, and any helpdesk channel. One agent serves all four moments of the customer lifecycle.
Team Assistant
An AI assistant for your internal team. Lives inside the helpdesk your team already uses. Drafts replies, surfaces context, and runs Procedures on behalf of agents.
The moments
Four customer-lifecycle moments where the agent works: Acquisition, Retention, Sales expansion, and Support. One agent covers them all, but each moment has its own Signals, Personality, Procedures, and Notifications.
The engine side
Engine
The back-of-house platform behind both agents. Two structural ideas live here: the three senses and the loop.
The three senses
- Living Knowledge is what the agent knows. Built from your Sources, structured into the Content library, published as the Help Center.
- Living Memory is who the agent is talking to. Per-customer preferences, history, sentiment, and goals. Scoped per agent, kept private.
- Living Context is where the agent is acting. Connections to your CRM, billing, ERP, helpdesk, and custom APIs so the agent acts inside your systems of record.
The loop
Four phases the Engine cycles through: Train, Test, Deploy, Analyze. Each pass sharpens the agent. Train builds the knowledge and procedures. Test validates before customers see anything. Deploy puts the agent to work. Analyze measures what happened and feeds the next cycle.
Trust and governance
Trust
Governance layer that sits alongside the loop. Four pillars: Transparency, Privacy, Security, Accountability. Built into the platform at the architecture level, not added after the fact.
Agent boundary
The agent does real work across the lifecycle, but never takes an irreversible action without a human approving the boundary first. Customers get help. Your auditor gets the receipts.
Model constellation
The Engine runs on a constellation of EU-hosted foundation models from Anthropic, Mistral, OpenAI, Google, Cohere, and Amazon Titan. The constellation lets Unless pick the right model for the job and swap providers in real time if regulation or vendor terms change. Customer data is never used to train public models.
Frequently asked questions
How do I tell Customer Agent from Team Assistant?
Customer Agent talks to your customers. Team Assistant helps your team. Same Engine behind both, different surface.
What is the difference between Living Knowledge, Memory, and Context?
Knowledge is what the agent knows from your Sources. Memory is what it remembers about a specific customer. Context is what it sees right now from your CRM, billing, and product data.
Why is Sales expansion sometimes written as just Expansion?
On unless.com the moment is labeled Expansion. The help center uses Sales expansion to be unambiguous. They mean the same thing.
Where do Procedures fit?
You set up Procedures in Train. You attach them to a specific Moment in Deploy.