​
Core concepts

Everything in Tacit is built from a small set of nouns. Each gets a paragraph here, with a link to its full page.

​
Scenario

A scenario is a simulated situation drawn from your domain: a claim to assess, a customer to talk to, a case that unfolds over weeks. Scenarios come in three shapes of increasing richness: decision tasks, conversation scenarios, and journey simulations. Any shape can be completed by a person or by an artificial intelligence (AI) agent. See scenario types.

​
Persona

In conversation scenarios and journeys, the operator interacts with a persona, a simulated person who holds the facts of the situation. A persona is built from two reusable parts: an identity (who they are: name, age, occupation, background) and a personality (how they behave: how talkative, how direct, how guarded). See personas.

​
Operator

The operator is whoever completes a scenario: a person on your team or an AI agent you have configured. Shape and operator are separate axes. The same decision task, conversation, or journey can be run by either, and is scored the same way for both.

​
Session

A session is one completion of a scenario by one operator. It is the unit everything else attaches to: the transcript, the outputs the operator produced, and the score. Run the same scenario with ten operators and you get ten sessions you can compare.

​
Known state and hidden state

A persona’s information is split in two. Known state is what the persona knows about themselves, and each item carries a reveal trigger: volunteered freely, disclosed only when asked directly, or shared only after the operator builds rapport. Hidden state is what the persona genuinely cannot know, such as test results or records they have never seen, which the operator must uncover through their own process. See state.

​
Briefing

The briefing is what the operator sees before a session starts: the situation, the setting, who the operator is in this scenario, and what they know going in. A good briefing gives enough context to begin and nothing that gives the answer away. See briefing.

​
Outputs

Outputs are what the operator is expected to produce during a session. They come in two kinds: artifacts, which are documents or records like a summary note or a referral letter, and decisions, which are choices like approve or deny. See outputs.

​
Scoring criteria

Scoring criteria define what good looks like for a scenario: success metrics the operator should achieve, failure metrics they must avoid, rubrics that grade quality across levels, scope boundaries that mark what is in and out of bounds, and terminology the operator should use correctly. Every session is scored against these criteria, and the report shows which were met and why. See defining good.

​
Agent

An agent is a configured AI model that can take the operator seat. You give it instructions, assign it scenarios, and its sessions are scored against the same criteria as everyone else’s. See running your AI.

​
Cohort

A cohort is a batch of scenario assignments sent to a group of operators. Use cohorts to run your whole team through the same scenarios and compare the resulting sessions. See running your people.

​
Library

The library holds your organization’s reusable definitions: identities, personalities, rubrics, artifact and decision definitions, and terminology. Define something once in the library and link it into any scenario that needs it. See library.