Purpose
Smaya ships pre-built agents, not an empty builder. Each agent is a specialized worker accountable for one job, already designed, tested and governed. You configure them and combine them into AI Employees.Why pre-built. An agent that makes judgments about people needs an
evidence standard, a fairness posture and an approval model behind it. Those
cannot be assembled from a blank canvas by a customer in an afternoon. We
build the agent; you set its scope, thresholds and authority.
The library today, and where it is going
26 — Exterview
A specialized workforce operating inside the talent application. See Exterview Agents.
10 — Gradientflo
A specialized workforce for everything after the hire. See Gradientflo
Agents.
100 — the platform
The library grows with each workload, on one governance model. See Roadmap to 100.
See the library in the product
Upcoming. An Agent Catalog in Settings → Setup → Operations that lists the agents in the library and what each one does, and Agent Lineage, built from live handoffs, that shows how agents hand work to each other.What every agent declares
Before an agent may run, five things are declared and enforced. This is the same contract in every application.
Not every agent uses a language model. Several are pure rules and connections, and are labelled as such — a deliberate design choice, because a calculation should never be a guess.
What a customer can change
Configuration is real, and its limits are deliberate.You control
Which agents run in your hiring pipeline · their thresholds · tone and language · the criteria they evaluate against · the agent’s role for each job. Upcoming: composing agents into your own AI Employees.
The platform controls
The structure of what an agent returns · its permission ceiling · the two-outcome decision model · terminal decisions reserved to humans · the quality and fairness thresholds an agent must pass.
One agent is deliberately read-only
Upcoming. Every agent’s output will be independently reviewed by a Quality & Fairness Agent. That reviewer cannot be reconfigured by anyone, including us, on a per-customer basis — a tenant able to soften it could pass a biased outcome through a clean scorecard. It stays fully visible and readable: the transparency is the point, the immutability is the control. Today, automated checks run on every scored output and their results show in Scoring Logs. See AI Observability.FAQs
Are all 26 Exterview agents available today?
Are all 26 Exterview agents available today?
No. Exterview Agents labels each one Available or Upcoming. The agent workforce is 26 roles either way, and a buyer asking “what does your AI do?” deserves the whole map, not only the finished half.
Can we request an agent that does not exist?
Can we request an agent that does not exist?
Yes. Customer demand is the main input to the roadmap toward 100. Custom agent
development is available as a services engagement.
Can an agent be given more permission than the person using it?
Can an agent be given more permission than the person using it?
No, in any configuration. An agent’s effective authority is the intersection
of its own declared scope and the permissions of the person who invoked it.
How do we know an agent still performs after we change its settings?
How do we know an agent still performs after we change its settings?
Job criteria are locked once candidates are in flight, and automated checks run on every scored output. Upcoming: evaluating every configuration change before it takes effect, and blocking one that degrades quality or moves a fairness measure. See Agent Studio.
Related
- Exterview Agents. The 26, by stage.
- Gradientflo Agents. The ten, all Upcoming.
- AI Employees. How agents become a deployed role.