---
title: 'Check artificial intelligence results and choose input data carefully'
url: 'https://ctg.gmbh/en/wissen/ki-management/ki-ergebnisse-pruefen'
markdown: 'https://ctg.gmbh/en/wissen/ki-management/ki-ergebnisse-pruefen.md'
lang: en
date: '2026-10-16'
description: 'Verify artificial intelligence results against business expectations and choose input data carefully, keeping responsibilities and limits clear.'
taxonomy:
  category:
    - 'Artificial intelligence management'
  tag:
    - 'ISO/IEC 42001'
    - 'Generative KI'
---

![Symbolic illustration, generated with artificial intelligence](https://ctg.gmbh/user/pages/05.wissen/05.ki-management/ctg-ki-management-ai-347cce84e19c4a68b2c1e5a07b3a81f5.webp)

# Check artificial intelligence results and choose input data carefully

A fluent text can appear convincing while containing incorrect facts, invented sources or missing limitations. Professional review and deliberate data selection belong in operational use of generative artificial intelligence (AI).

## Check what could change a decision

The level of review should reflect possible consequences. An internal idea draft needs different checks from customer information about technical specifications. Recalculate figures, actually open cited sources and compare statements with reliable records.

The generative AI guidance from the National Institute of Standards and Technology (NIST) discusses confident false statements and risks to personal and confidential data. Organisations need controls appropriate to their applications.

## Example: preparing quotations

An assistant turns notes into a quotation draft. The responsible employee compares scope, quantities and commitments with approved information. A service promise added by the system is removed. Confidential customer data is entered only if its use in that tool has been checked and permitted.

Removing names does not automatically anonymise a case. Project details can still identify people or customers.

## Your first step

Create a short checklist for a recurring task: what must be correct, which source is authoritative, what data may be used and who decides when in doubt? Trial the process with fictional data. Collect problematic results so that training and instructions address actual errors.

## Sources

- [Official reference: Check artificial intelligence results and choose input data carefully](https://nvlpubs.nist.gov/nistpubs/ai/NIST.AI.600-1.pdf)

Information date: 16 October 2026.

[Back to the topic overview](https://ctg.gmbh/en/wissen/ki-management)

---

## Navigation

- Parent: [Artificial intelligence management](https://ctg.gmbh/en/wissen/ki-management.md)
- Previous: [Who decides how artificial intelligence is used?](https://ctg.gmbh/en/wissen/ki-management/ki-verantwortlichkeiten.md)
