---
title: "Assess Finding With Ai"
method: POST
path: "/companies/{company_id}/review/findings/{finding_identity}/assess-with-ai"
tags: ["statement-review"]
---

# Assess Finding With Ai

`POST /companies/{company_id}/review/findings/{finding_identity}/assess-with-ai`

## Request

- Base URL: `https://app.zeno-works.ch` (the document declares no server; this is the origin it was published from)
- URL: `https://app.zeno-works.ch/companies/{company_id}/review/findings/{finding_identity}/assess-with-ai`
- Auth: none declared

Enqueue the AI-accountant narrative for one finding (accountant).

Brokered like every "Run with AI" flow: one pending ``ai_task`` row the UI
polls; the narrative lands cached on the finding (`ai_assessment`) keyed to
its value fingerprint, so an unchanged finding is never re-billed. The
deterministic verdict stays authoritative — the AI only explains/proposes.

## Path parameters

- `company_id` string, required
- `finding_identity` string, required

## Headers

- `authorization` string, nullable

## Cookies

- `access_token` string, nullable

## Response `202`

Successful Response

- AiTaskOut: One ``ai_task`` row — the poll target for a "Run with AI" click. ``result`` is the executor's structured outcome once ``done`` (a ``failed`` task may carry a partial result too, e.g. payroll blocking findings); ``error`` is the trimmed failure message.
  - `id` string, required
  - `kind` 'payroll_run' | 'ar_draft' | 'period_close' | 'fixed_asset_review' | 'statement_review', required: Which "Run with AI" flow an ``ai_task`` row executes. ``payroll_run`` — advance a draft payroll run calculate→approve→post; ``ar_draft`` — draft + ready an outgoing AR invoice from an instruction; ``period_close`` — run the close checks + the autopost accrual sweep; ``fixed_asset_review`` — apply the recognition settings to a fiscal year's asset inventory, post its due depreciation, accept the year's unambiguous held capitalization proposals, and (under ``auto`` only) finalize the review. ``statement_review`` — produce the AI-accountant narrative assessment for one statement-review finding (doc/pipeline.md §Statement review); the verdict stays deterministic, the task only explains and proposes.
  - `status` 'pending' | 'processing' | 'done' | 'failed', required: Queue lifecycle of an ``ai_task`` row (doc/architecture.md §"Work queue"). ``pending → processing`` on claim (heartbeat ``claimed_at``), then a terminal ``done``/``failed``; the stuck-row sweep returns abandoned ``processing`` rows to ``pending``.
  - `target_id` string, nullable
  - `payload` object, required
  - `result` unknown
  - `error` string, nullable
  - `requested_by` string, nullable
  - `created_at` string, date-time, required
  - `finished_at` string, date-time, nullable

## Other responses

- `422`: Validation Error

---

[API](https://skmtc.dev/zeno-works/apis/zeno-api.md) · [All operations](https://skmtc.dev/zeno-works/apis/zeno-api/llms.txt) · [OpenAPI document](https://skmtc.dev/zeno-works/apis/zeno-api/revisions/1eb93dde146c?raw)