How to Pass HireVue Interviews - I Sat Too Many

[PotentiallyABanker1]

Maybe you’ve just applied for an internship or graduate program, or you are thinking of doing so and have heard of these “HireVue” video interviews. I had many questions when I first heard of them, and you might too.
They are used by firms including Goldman Sachs, J.P. Morgan, UBS, Bank of America, BlackRock and more, and now form a key part of the hiring process.

Having done a countless number of them over the past couple of years, I’m going to break down a little about what they are, and what I’ve learned about them.
(Note: this is all information I’ve gathered from specifically UK-based roles, I’m unsure how much of this applies to the US or other countries)

What is a HireVue/Video Interview?

It’s essentially a pre-recorded job interview that you complete in your own time, on your own laptop, answering a set of pre-defined questions. Rather than speaking with a live interviewer, questions will pop up on-screen for you to answer, with your response being recorded using your webcam.
Yes, you talk at your computer screen in 1-2 minute chunks, while sitting in your bedroom. Feels odd, right?

As unnatural as it might feel, a large majority of the large investment banks now include a video interview as part of the initial hiring process, particularly on university-targeted programmes. It also serves as one of the earliest and easiest (for the bank) filters to weed out candidates, so it is important that you ace it!

HireVue specifically is the name of one of the most commonly used video interview providers by banks. There are others (such as Sonru etc.), however, they all function similarly.

Why are they used?

They are a low-resource and low-cost method for a firm to quickly gather responses from thousands of candidates and filter down the candidate pool quickly. It saves wasting time during 1st round interviews, and the results are often used in conjunction with psychometric test results to decide who to progress to the interview stage.

How is it judged, if there is no human being involved?

It isn’t entirely removed from human judgement. The hiring managers at each firm typically review all of the video interviews manually, which means you can forget your worries about a computer misunderstanding your voice.

However, some companies turn on HireVue’s AI-based scoring system. This also analyses your facial expressions, eye contact, tone of voice and other data from the video stream, which is all used to issue you with an overall “score”. Although not used exclusively to make decisions, hiring managers can see this score, and thus it is additionally important to conduct yourself as you would in a real interview.

It’s therefore important first and foremost to act naturally, as you would in a real interview, even if the whole process seems unnatural.

What should I wear?

The interview is still viewed by a member of the recruiting team, so it is important to still wear smart business attire, as you would for any ordinary interview.

What might be I asked?

Quite frankly, anything. I’ve had questions ranging from “Talk about something that’s happening in the markets currently” to “What is 144 divided by 90?”. In general, though, it’ll be broad motivational and behavioural questions and ones that test your basic finance, markets and current affairs knowledge.

Expect questions such as:

• “Why us (this company)/why does this role interest you?”
• “Name a time you’ve worked in a team”
• “What is your proudest non-academic achievement?”
• “What industry or product would you invest in, and why?”.

This is by no means an extensive list, so prepare for them as you would a standard first-round interview.

Should I prepare answers?

Many companies use identical questions for all candidates, which you might be able to figure out ahead of time. Preparing answers word-for-word/in script form though is definitely not advisable. It comes off as unnatural in the video, and obvious that you are reading from a script, which will be picked up on.

Instead, I’d suggest making a few bullet points and conducting research (about the firm and current affairs) beforehand, so you are poised well to answer said questions but do so in a natural and “flowy” way. The last thing you want to do is to be caught reading answers from your laptop screen!

My Personal Tips

The key is to remain calm and unflustered, and treat it like an actual interview.
I’d definitely avoid making detailed notes to read from to keep the responses conversation-like and as natural as possible.
Dress smart, and ideally have good lighting, a good webcam, a plain background and no background noise.

If you have a friend who applied to the same role, it is useful to find out the questions they were asked. However, don’t count on this! I did this once (for Bank of America), and found out the hard way that they serve each candidate a different set of questions, even for the same program. This got me flustered, as I was relying on my carefully curated pre-prepared answers, and ultimately led to a pretty bad HireVue.
And finally, make notes of the questions you were asked. They might be the same next year if you happen to be applying again, or you might be able to help a friend out!

Feel free to reply if you have any questions, or DM me privately if preferred.

This is all based on my personal experiences, but it would be great to hear other people’s tips and tricks for approaching these!
I hope that was useful, and good luck!

 

[magellan1]

Thank you for the comprehensive explanation. HireVue interviews have a few advantages over traditional first-round phone interviews. When I was applying for spring weeks (>8 years ago), the interviewers of many banks did not inform me in advance about the day/time of the interviews and several times I found myself in weird situations. For example, once Bank of America called me 10 minutes after I woke up and I had a class starting in 20 minutes. Obviously, the interview didn't go well. With HireVue you can do the interview when it is most convenient for you. Another advantage is that a lot of the questions are the same in different interviews and you can improve over time. In the past, I have completed HireVue interviews for Blackrock and GIC. I don't remember the exact percentage but something like 30% of the questions was the same. A few years ago there was some scandal with AI scoring systems related to discrimination against black people but I guess that the algorithms have improved significantly since then.

 

[AugustusSeptimius1]

Yeah thanks for this. Seems these video interviews are becoming a permanent part of all undergrad interviews so definitely applicable advice. That's crazy that there is an AI facial expression assessor, didn't know that would be a thing.

 

[PotentiallyABanker1]

Thank you for the comprehensive explanation. HireVue interviews have a few advantages over traditional first-round phone interviews. When I was applying for spring weeks (>8 years ago), the interviewers of many banks did not inform me in advance about the day/time of the interviews and several times I found myself in weird situations. For example, once Bank of America called me 10 minutes after I woke up and I had a class starting in 20 minutes. Obviously, the interview didn't go well. With HireVue you can do the interview when it is most convenient for you. Another advantage is that a lot of the questions are the same in different interviews and you can improve over time. In the past, I have completed HireVue interviews for Blackrock and GIC. I don't remember the exact percentage but something like 30% of the questions was the same. A few years ago there was some scandal with AI scoring systems related to discrimination against black people but I guess that the algorithms have improved significantly since then.

I've seen a few articles regarding discrimination and bias that can essentially end up "baked" into the AI algorithms, like penalising non-native English speakers for example (because their tone of voice might not be as clear or enthusiastic as a native speaker, as judged by the AI), so I guess in that respect it is definitely a little controversial.

Actually being able to do the interview at any time that's convenient for you though, as you said, is a lifesaver. Definitely saves a lot of hassle!

 

[JustAnotherHuman1]

There is nothing crazy about AI tech, or using AI-assisted technology to aid and speed up decision making itself. It utilises machine learning under the hood and is NOT magic but application of statistical models on computers, at scale. Is basically enabled by fast processing of large data sets, with something like 1000 computers processing the same data set, distributed and each with like 100 CPU cores. This will hopefully uncover patterns and valuable insights, that were previously impossible to do with only pen and paper.

Recent digitalisation of all industries has caused an exponential growth in data that is available to be crunched and analysed. Fun fact: there was 64 zettabytes (64 trillion of gigabytes) of data available globally in 2020. What’s even more, this processing power is readily available at anyone’s fingertips thanks to data platforms offered by the likes of Amazon Web Services, Microsoft Azure and Google Cloud Platform. Going off on a tangent here, but I’m sure some of you will find this useful.

So how does HireVue rank candidates exactly? Their algorithm is obviously non-public, as it then could be manipulated (although the AI bot is most definitely supervised by a human as well - an HR rep), and their exact solution to this problem is a trade secret so that competitors can’t copy their product. I’m sure it’s quite sophisticated, with some complex and interesting metrics/weights that might even be a bit wild but still work, the input is really only: a resume and the words on it, a video and audio file of the interview, and whatever that can be extrapolated based on these variables. Then, there are only 2 groups of candidates: ones who get an interview, and ones who don’t. Then based on historical data the AI can probably make a reasonable bet on whether a candidate is “worthy” or “not worthy”. For instance if the AI finds with the help of an emotion recognition library that most people who have the words “improved sales”, “Duke of Edinburgh award” and are also generally positive and happy throughout the interview, probably belong to the “worthy” category and so they will get an interview. This is a simplification but you get the idea.

Is HireVue more advanced than a WordPress website selling t-shirts? Absolutely. Is it perfect, with a 100% accuracy and no false-positives? It’s not.

 

[PotentiallyABanker1]

There is nothing crazy about AI tech, or using AI-assisted technology to aid and speed up decision making itself. It utilises machine learning under the hood and is NOT magic but application of statistical models on computers, at scale. Is basically enabled by fast processing of large data sets, with something like 1000 computers processing the same data set, distributed and each with like 100 CPU cores. This will hopefully uncover patterns and valuable insights, that were previously impossible to do with only pen and paper.

Recent digitalisation of all industries has caused an exponential growth in data that is available to be crunched and analysed. Fun fact: there was 64 zettabytes (64 trillion of gigabytes) of data available globally in 2020. What’s even more, this processing power is readily available at anyone’s fingertips thanks to data platforms offered by the likes of Amazon Web Services, Microsoft Azure and Google Cloud Platform. Going off on a tangent here, but I’m sure some of you will find this useful.

So how does HireVue rank candidates exactly? Their algorithm is obviously non-public, as it then could be manipulated (although the AI bot is most definitely supervised by a human as well - an HR rep), and their exact solution to this problem is a trade secret so that competitors can’t copy their product. I’m sure it’s quite sophisticated, with some complex and interesting metrics/weights that might even be a bit wild but still work, the input is really only: a resume and the words on it, a video and audio file of the interview, and whatever that can be extrapolated based on these variables. Then, there are only 2 groups of candidates: ones who get an interview, and ones who don’t. Then based on historical data the AI can probably make a reasonable bet on whether a candidate is “worthy” or “not worthy”. For instance if the AI finds with the help of an emotion recognition library that most people who have the words “improved sales”, “Duke of Edinburgh award” and are also generally positive and happy throughout the interview, probably belong to the “worthy” category and so they will get an interview. This is a simplification but you get the idea.

Is HireVue more advanced than a WordPress website selling t-shirts? Absolutely. Is it perfect, with a 100% accuracy and no false-positives? It’s not.

The controversy around using AI isn't necessarily about the fact machine learning/statistical methods are being applied as part of the decision-making process, but more about the biases that might be introduced into a model through training, or through a self-fulfilling loop as the model is continuously retrained.

For example, it might be the case that a particular age, gender or ethnicity does better/worse than the average candidate, and a typical predictive machine learning model would place a positive weight on this. In short, this means that it'll start giving higher/lower scores based on these characteristics.
In the UK, these are protected characteristics on which it is illegal to discriminate on (regardless of whether you believe they are predictors of a successful candidate), which opens up many questions. Does the model have built-in preventative measures to stop it from building these biases, and is it regularly checked and screened to ensure it doesn't?

If it doesn't/isn't, then it is in a murky area, both ethically and legally.

The fact that the model is proprietary for HireVue for example is particularly problematic for them, and places them under increased scrutiny, as nobody can analyse how it works and thus must take their word for the fact it's fair and free from bias.

Naturally, all humans have some element of bias too, so I'm not saying that necessarily an AI generating these scores is objectively a negative thing, but without regulatory oversight, it is easy for these sorts of products to get out of hand and defeat their original purpose, in the interest of cost savings and efficiency.

 

[JustAnotherHuman1]

The controversy around using AI isn't necessarily about the fact machine learning/statistical methods are being applied as part of the decision-making process, but more about the biases that might be introduced into a model through training, or through a self-fulfilling loop as the model is continuously retrained.

For example, it might be the case that a particular age, gender or ethnicity does better/worse than the average candidate, and a typical predictive machine learning model would place a positive weight on this. In short, this means that it'll start giving higher/lower scores based on these characteristics.
In the UK, these are protected characteristics on which it is illegal to discriminate on (regardless of whether you believe they are predictors of a successful candidate), which opens up many questions. Does the model have built-in preventative measures to stop it from building these biases, and is it regularly checked and screened to ensure it doesn't?

If it doesn't/isn't, then it is in a murky area, both ethically and legally.

The fact that the model is proprietary for HireVue for example is particularly problematic for them, and places them under increased scrutiny, as nobody can analyse how it works and thus must take their word for the fact it's fair and free from bias.

Naturally, all humans have some element of bias too, so I'm not saying that necessarily an AI generating these scores is objectively a negative thing, but without regulatory oversight, it is easy for these sorts of products to get out of hand and defeat their original purpose, in the interest of cost savings and efficiency.

Thanks for your reply; I think the solution is quite simple and that is to punish white candidates for their “head start” or even better, have quotas to promote diversity. That way it’s a fixed number that needs to be selected, eliminating the problem of not having access to their black box. This would help to level the playing field and promote diversity in intern classes, if the goal is to promote equal opportunity and not to hire the best of the best. There is a bias in the real world that favours white privately educated students, as they will simply have more impressive backgrounds than their less fortunate counterparts. So naturally this will be reflected in the algorithm as well, not necessarily because the AI is inherently racist, but because this industry can afford to hire the best of the best, and those people tend to come from a specific background. This issue however has nothing to do with the AI, really, and discrimination happened long before it came along in recruitment.

 

[AugustusSeptimius1]

Thanks for your reply; I think the solution is quite simple and that is to punish white candidates for their “head start” or even better, have quotas to promote diversity. That way it’s a fixed number that needs to be selected, eliminating the problem of not having access to their black box. This would help to level the playing field and promote diversity in intern classes, if the goal is to promote equal opportunity and not to hire the best of the best. There is a bias in the real world that favours white privately educated students, as they will simply have more impressive backgrounds than their less fortunate counterparts. So naturally this will be reflected in the algorithm as well, not necessarily because the AI is inherently racist, but because this industry can afford to hire the best of the best, and those people tend to come from a specific background. This issue however has nothing to do with the AI, really, and discrimination happened long before it came along in recruitment.

I must say, I think that is a terrible solution. It’s one thing encouraging individuals from different backgrounds to look into and/or apply to certain industries or roles to make sure that industry/position is more reflective of the overall society and encompasses a greater diversity of thought (an undoubtedly good thing) and even taking the necessary steps to bridge any blatantly unfair gaps that may preclude certain groups from accessing particularly lines of work; however, it is another thing altogether to selectively “punish” and exclude a specific group of people on the basis of their race, an act which is dictionary-definition racist and discriminatory and one which you suggest as your solution to fixing discrimination.

 

[JustAnotherHuman1]

I must say, I think that is a terrible solution. It’s one thing encouraging individuals from different backgrounds to look into and/or apply to certain industries or roles to make sure that industry/position is more reflective of the overall society and encompasses a greater diversity of thought (an undoubtedly good thing) and even taking the necessary steps to bridge any blatantly unfair gaps that may preclude certain groups from accessing particularly lines of work; however, it is another thing altogether to selectively “punish” and exclude a specific group of people on the basis of their race, an act which is dictionary-definition racist and discriminatory and one which you suggest as your solution to fixing discrimination.

I meant to write "penalize" and not "punish", as in reduce the AI-assigned numerical score to candidates (being white could reduce your score by let's say 10 points). I didn't mean actually limiting opportunities to anyone group - sorry for the confusion, just making the system work fairer.

 

[WLaidlaw1]

I meant to write "penalize" and not "punish", as in reduce the AI-assigned numerical score to candidates (being white could reduce your score by let's say 10 points). I didn't mean actually limiting opportunities to anyone group - sorry for the confusion, just making the system work fairer.

In my opinion, there may be other stages in the hiring process that disadvantage individuals from over represented groups, particularly those who are white. For example, during the initial phase of collecting personal information, there may be a diversity page that considers this factor. Although AI may still require further development to avoid penalizing marginalized groups, these procedures are usually scrutinized by diversity teams to ensure fairness. However, it's worth noting that some sixth forms and universities may offer coaching for these types of interviews, which can create an additional form of bias that is not related to race.

This process was something I wasn't aware of when I began applying and my first interview absolutely bombed. It really is just about practice and trying to just enjoy giving your answer. A tactic I came up with was cutting a hole in a picture of my friend and, after practicing talking through my answers with them, giving them to the picture which I stuck to the camera so it felt like I was not just speaking to myself.