What is the Applied AI Screening Assistant?

In this article:

- What is the AI screening Assistant?
- Role requirements for using the AI screener (and what happens if I don't meet them)
- Question types that can be used with the AI screener
- Which types of questions cannot be used with the AI screener
- How does the AI screener work? 
- What goes into our smart AI screening Assistant?

What is the Applied AI Screening Assistant?

The AI Screening Assistant massively reduces the time it takes to shortlist large numbers of candidate Sift answers for high-volume roles. Our AI model combines years of research on candidate skills and skills-based questions, data points from over a million candidate answers and scores, and the customised and calibrated to score reliably, efficiently, and accurately according to your organisation’s needs, expectations, and requirements. 

Meeting role requirements?

For the best and most accurate results for your role, we require that roles meet the following criteria to enable the AI screening assistant. 

1. You must receive at least 100 applications

We aim to supply your hiring team with the most accurate and fair assessment of each candidate’s answer. The AI model requires a larger subset of candidate applications (100 minimum*) to best apply your personalised AI model to the answers in order to effectively and efficiently score the answers. I

*If your role does not meet the 100 applications minimum despite enabling the AI screening tool, contact your Customer Success Manager to discuss how to handle credits for your organisation.

2. Use Sift questions on the role

Sift questions, or skills-based scenario questions, are still considered one of the most powerful and predictive methods of assessing a candidate’s actual role fit, yet Applied understands that this can also take up a lot of reviewing time from your company. For this reason, our AI model takes a candidate's Sift answers, removes all personal information or identifying data, and then reviews each candidate based on various research-backed criteria and personalised benchmarks/requirements. 

Which types of questions can be used with the AI scorer?

1. Work sample / scenario questions questions accepted

Questions that review how a candidate will deal with a specific tasks or project work well with the AI screener because each candidate's response will be very different and based on their own personal skills and thought process and past experience. 

2. Work sample / skills-based scenario questions accepted

Questions revolving around certain work scenarios and problems or question to specifically assess specific skills work very well with the AI screener. These questions tend to allow canddiates to express their personal thoughts and approaches, which allows the AI model to better identify and assess individual skills and standout answers.

Which types of questions CANNOT be scored by the AI scorer?

1. Motivation questions

2. Past / behavioural questions

3. CV type questions (past experience, university degrees, etc.)

These types of questions will NOT be scored by the AI scorer. These questions tend to inherently show high similarity between answers and are more susceptible to generic or repetitive answers. The AI scorer cannot score or measure levels of interest, passion, or motivations for jobs. This tends to be a question best reviewed by your hiring team directly. CV questions/past behavioural cannot be scored by the AI screener as this involves either human-implicit interactions and codes that the AI model cannot detect/score or lists of requirements, which can be difficult to determine what certain requirements or degrees mean to the role. 

How does the AI screener work? 

Once you've set up your role with skills-based Sift questions (see which types of questions work with the AI screener above), you will set up review guides for each question, clearly outlining what a 1 star, 3 star, and 5 start answer means for your team. These review guides can describe the ways in which you would like candidates to handle certain situations or what types of though processes or behaviours you prefer to see and which would be complicated for the role. 

Once your role goes live, candidates will start applying. When you hit at least 50 applications, you will randomly select and review 20-30 candidates across a few reviewers on your hiring team. This is what your personalised AI model will use to ensure it understands your scoring guidelines, thresholds, patterns, and which answers your team likes and does not like. Then it uses this information to train and recalibrate your AI model so that when the automated scoring takes over, it will be as if your team was scoring all the candidates. 

What goes into our smart AI screening Assistant?

Along with your detailed and personalised review guides and calibration scores from your company's hiring team, the AI model combines years of research and data points across 20,000 live roles and 80 different industries, over 8 million scores across candidates' answers to skills-based questions, and over 1.2 million candidate applications to create a strong and customised AI scoring model. We combine all this data and trends to ensure scoring is accurate, fair, and informed. On top of it all, the reviewing process itself combines the best practices from behavioural science research, providing a structured and ethical background to the scoring and hiring process at Applied.