> ## Documentation Index
> Fetch the complete documentation index at: https://docs.enfors.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# Natural-language analysis

> How Enfors turns plain-English questions into safe queries.

Enfors lets you analyze your AI workforce in plain English — no SQL, no dashboard
building. This is delivered through the [MCP server](/mcp/overview) and the
[Ask Enfors](/using/ask-enfors) experience.

## Safe by construction

A core safety principle: **the model never writes SQL.** Instead, an AI client calls
Enfors's typed MCP tools (`list_agents`, `list_decisions`, `search_similar_decisions`,
…), each of which builds a **parameterized query** server-side. Free-text questions are
translated into validated filter objects, not raw SQL strings.

```mermaid theme={null}
flowchart LR
  Q[Plain-English question] --> C[AI client]
  C -->|typed tool calls| M[Enfors MCP]
  M -->|validated filters| DB[(Decision ledger)]
  DB --> M --> C --> R[Rendered report]
```

## What you can analyze

* **Production & cost** — "what did the team produce in the last three days, and what did
  it cost?"
* **Comparisons** — pass rate / spend / subjects across agents or people
  (`list_agents` rollups).
* **Problem-finding** — rank rework or contradiction-heavy sessions
  (`list_agent_sessions(sort_by=rework)`).
* **Semantic recall** — "find decisions like this one"
  (`search_similar_decisions`).
* **Delivery & ROI** — Feature and Epic rollups (`list_features`, `get_feature`).

## Rendering the answer

Results are presented with [`render_ui`](/mcp/render-ui) as a branded, shareable view —
with PII removed. See [Privacy](/concepts/tenancy-privacy-pii).

<Tip>
  Start broad with rollup tools, then drill down. Most "who/what/how much" questions are
  answered without paging individual decisions.
</Tip>