Sapia.ai buying intent

One of the topics we covered in our recent The Hiring Science Studio webinar was 'Shadow AI' - where AI has already crept into your hiring process through individual employee use. These use cases range from high risk (pasting CVs with PII or PII proxies into a personal Claude or ChatGPT workflow) to seemingly benign (using an LLM as a thought partner for interview design). However, even the benign end can substantially undermine the accuracy of your hiring process without the right guardrails. Here is a real suggested prompt from the OpenAI Academy's HR prompt pack (link in comments): 'Develop behavioural interview questions aligned to our company values for a [role title] opening in [team/department]. We want to assess both technical skills and culture fit. Provide 6 to 8 questions grouped by competency.' Five problems with this, before a single question is generated: 1. No definition of what a behavioural interview question actually is. The LLM will infer. 2. The prompt asks to assess behaviours against values. Behaviours cannot be used to infer values. Values do not predict performance. 3. Technical skills cannot be assessed through behavioural questions. That is a different methodology. 4. Culture fit cannot be assessed through behavioural questions either. And arguably, it shouldn't be assessed at all. Cultural homogeneity is not an asset when organisations are trying to build diversity of thought. 5. '6 to 8 questions grouped by competency' - with no competency framework supplied. The LLM will make one up. AI has a critical role to play in hiring. And not every organisation has the budget or the governance structure to bring in a robust, validated, external AI tool. If you'd like to talk to us about some practical, low-cost steps to bring order to shadow AI in your hiring process and reduce the risk, get in touch.