Job interview analysis platform Sapia launches generative AI chatbot to explain its hiring decisions

The next time you go for a job interview at Woolies or Qantas, you may end up with generative AI not only telling you how you did, but also offering advice on how to perform better next time.

Melbourne recruitment startup Sapia.ai has launched a generative AI model it’s called SAIGE to give candidate feedback. It uses the platform’s large language model (LLM) capabilities developed over six years, to evaluate candidates using a blind chat interview, then deliver comprehensive scoring along with a set of job-related competencies, including explanations for its assessments, with the ability coach candidates as to how they can improve, or say what it was looking for to give a higher score for a particular skill or trait.

Sapia, which launched in 2018 and raised $17 million in a Series A from Woolworths and Macquarie Capital in late 2022, has used artificial intelligence (AI) to build a hiring analysis data set from more than 3.5 million job interviews spanning 15 million answers and 2 billion words.

The platform combines natural language processing (NLP) and machine learning (ML) to build a job candidate’s HEXACO (for Honesty-Humility, Emotionality, Extraversion, Agreeableness, Conscientiousness, and Openness to Experience) personality profile, alongside analysing their role-based skills, behaviours, experience and more, also sharing those findings with the person involved.

Companies, including Woolworths, Qantas, Suncorp, Starbucks Australia, Holland & Barrett, and BT Openreach, use Sapia as part of the initial screening process for recruitment.

Sapia.ai founder and CEO Barb Hyman said SAIGE has been trained through the use of Reinforcement Learning from Human Feedback (RLHF) to ensure candidates are assessed accurately, fairly, and in a way that mitigates human biases.

“We have been on a mission to humanise the way people are evaluated, and SAIGE is the next step in this journey. As much as ChatGPT was a revolution in the way we interact with AI, SAIGE is a quantum leap in the way people will be assessed and coached,” she said.

“Structured interviews have been proven to lead to better hiring outcomes and reduce bias in the selection process. Now with SAIGE organisations have a way to consistently interview and evaluate at scale using a rubric set by experts, and also generate detailed interview notes. What we are really excited about is the potential SAIGE has on providing coaching to hiring managers and new employees on setting them up for success.”

Sapia.ai chief data scientist of Dr Buddhi Jayatilleke said that as well as detecting and assessing the skills and traits employers require for various roles, the platform can adapt over time and be trained to look for particular requirements for each role and organisation.

“A unique power of these LLMs is their ability to explain themselves,” he said.

“SAIGE is bringing this level of transparency that is imperative for the responsible use of AI and especially in people decisions, such as who to hire and promote, helping to reduce the unconscious bias we see in human-led decision making.”

Every job candidate receives an individual insights report detailing their strengths, their work style, and personalised coaching. Hyman said the low-pressure chat used by Sapia gives potential hires a chance to shine and the subsequent feedback means unsuccessful candidates feel seen.