Checkatrade CTO Gorden Pretorius: Making AI Transparent
Checkatrade CTO Gorden Pretorius explains how proprietary data and strict guardrails keep AI transparent, accurate and fair for physical trades
TL;DR
- AI's 'black box' nature makes it difficult to manage, explain, and regulate, yet it's increasingly used in high-stakes industries.
- Gorden Pretorius, CTO at Checkatrade, advocates for responsible and transparent AI use.
- Physical trades present unique challenges for AI due to high variability compared to simpler input/output systems like taxi pricing.
- Checkatrade uses proprietary data and numerous input factors to provide transparent pricing breakdowns for consumers.
- The company enriches LLM outputs with its own data to validate pricing and ensure fairness, using a 'confidence score' for areas with potential gaps.
- Deep, native integrations with LLM providers like Anthropic and ChatGPT are employed to minimize 'hallucinations'.
- AI implementation requires explicit governance, high observability, and a tailored risk appetite for different business functions.
- Checkatrade applies strict guardrails, often with a human in the loop, for sensitive areas like vetting.
- The company launched TradeMore, an AI-powered job management platform, in 2025.
- Pretorius sees AI evolving from information discovery to problem resolution, envisioning a system that can autonomously handle tasks like booking a tradesperson for a repair.