AI safety software

Financial Advice Detection for Guardrails AI

Name: Data Annotation Service
Brand: Label Your Data
Rating: 4.9 (36 reviews)

Location:

USA

Services:

NLP Annotation

Overview

Guardrails AI needed labeled data to train and test safety filters for enterprise AI apps. These filters help block unapproved financial advice and incorrect factual responses.

1000 sentences labeled

5764 facts extracted

7 annotators

3.5 weeks total

Client

AI safety software company providing real-time filters for GenAI apps like chatbots and LLM agents.

Challenges

Public datasets missed edge cases and internal teams had no time for manual annotation.

Solutions

Collected sentences, labeled financial advice, and extracted fact phrases from supplied texts using lightweight tools and pilot review.

Results

Delivered 1000 labeled lines and 5764 facts in 3.5 weeks; data used to ship new open-source validators with ready-to-use exports.

Client

Guardrails AI is an AI safety software provider offering real-time filters that help enterprises control chatbot and LLM responses for compliance and accuracy.

Group photo of the Guardrails AI team wearing branded sweatshirts, representing the company’s focus on AI safety and compliance solutions

Woman annotator working on a laptop to produce task-specific labeled data for safety filter improvement

Challenges

They needed task-specific labeled data to improve safety filters.

Public datasets missed real-world edge cases

Internal team lacked time for manual collection and labeling

Project required fast, accurate delivery with minimal oversight

Solution

We managed end-to-end annotation using a lightweight, tool-agnostic workflow that included both Google Sheets and the client’s internal platform.

Collected and annotated 1000 sentences for financial advice detection

Extracted 5764 factual claims from 580 client-supplied texts

Delivered clean, JSON-ready files for direct use in training

Screenshot of the client’s annotation interface showing prompt, response, and extracted factual claims.

Training

Project completed in 3.5 weeks with zero rework

Data used to train open-source safety filters for hallucination and advice detection

Results

Project completed in 3.5 weeks with zero rework

Data used to train open-source safety filters for hallucination and advice detection

Labels fit directly into the client’s model pipeline

Accelerated release of new validators, now live in Guardrails Hub

Screenshot of a Google Sheets document containing labeled financial advice sentences with columns for text, label, and comments.

Start Free Pilot

fill up this form to send your pilot request

Thank you for contacting us!

We'll get back to you shortly

Our collaboration with Label Your Data was smooth and easy. The measurable impact was that we shipped state-of-the-art validators and guardrails using open source models, which people now use to detect hallucinations.

Safeer Mohiuddin

Co-Founder of Guardrails AI

Trusted by ML Professionals

Why Projects Choose Label Your Data

No Forced Commitment

Check our performance based on a free trial

Flexible Pricing

Pay per labeled object or per annotation hour

Tool-Agnostic

Working with every labeling tool, even your custom tools

Quality Backed by SLAs

We commit to accuracy and deadlines – or you don’t pay