### **Key Takeaways**

1. 28% of high scores on CodeSignal's objective skills-based assessments came from schools not in the US News and World Report's top 50 engineering schools — meaning traditional proxies like school pedigree are creating significant noise that obscures the best candidates.
2. Placing a validated skills assessment at the top of the funnel — before recruiter screens — drives 70% candidate completion rates, 20–50% improvement in on-site-to-offer rate, and an average of 10 hours saved per candidate.
3. Cheating and fraud have increased 4x over an 18-month period, with the highest rates in early career roles. The four types: copy-paste from online forums, hiring a ringer, AI-generated answers, and deepfakes.
4. Each type of cheating requires a different mitigation: assessment variation and browser controls (copy-paste), ID verification and proctoring (ringers), behavioral pattern detection and AI flagging (AI-generated answers), and training interviewers to spot deepfake tells (deepfakes).
5. Before deploying assessments, your organization must define its integrity stance: is AI use considered cheating for the roles you're hiring for, or is it a core skill you want to measure? Communicate that policy clearly to candidates — many say they don't want to cheat but are afraid of accidentally doing something that's considered cheating.
6. CodeSignal's Cosmo AI agent can match a role description to the right assessment from a library of 3,000+ assessments across virtually any role done at a computer, and surface a sample report so you can see the output before deploying.
7. Assessment scores, skill proficiency breakdowns, and fraud flags all surface directly in Ashby — so recruiters never need to leave the system to evaluate candidates or make decisions.

### **Transcript**

*Note: transcript is auto-generated and may contain minor inaccuracies.*

\[00:00\]  **Katie Fairbank**

I gotta be honest, I wasn't quite sure what the vibe was going to be at 4:45 right before happy hour, but I appreciate you all being here. Quick poll before we start. How many of you have ever said, I would like to prioritize skills-based hiring? Yes. And how many of you feel like you totally have it figured out? Yep, that looks about right. Skills-based hiring, of course, is nothing new. And I don't need to tell all of you how hard it is, but I think we're all feeling the pressure now to move to skills-based hiring. AI is changing everything, we have more applications than ever. And the ways that we've been hiring before just aren't working like they used to. A really good example of this is university recruiting. Historically, we target certain schools because they have a great program and do some filtering as part of the admissions process. It's a solid strategy. But at CodeSignal, we wanted to see if top-ranked schools were actually producing the most skilled students. So we used engineering as an example and compared the US News and World Report's top engineering schools to the objective skills data that we have on engineering students.

**Katie Fairbank**

What we found was pretty interesting. 28% of high scores on our objective skills-based assessment were from schools that were not part of the US News and World Report's top 50. And in fact, 12 of the top 50 schools in our report weren't on the US News top 50 at all. So what this means is that our traditional proxies and ways of filtering candidates are creating a lot of noise that we need to cut through in order to find who is best for the role. So you're probably wondering, okay, great, but what am I supposed to do about this? What I'm about to show you really isn't revolutionary, but it's what our top customers are doing and having great success doing. They are rethinking their process to identify skill as the primary signal at the very top of the funnel. So when a candidate applies, the first thing they do is send out a skills assessment. From there, the team reviews the skills assessment right in Ashby, and decides who they want to move forward. From there, recruiters can connect, focused on building a relationship. The hiring manager then has a different conversation — more focused on team fit and building relationships rather than identifying if this person has the foundational chops for the role.

**Katie Fairbank**

You might be saying, okay, that sounds great in theory, but is the juice really worth the squeeze? Here's what we see: 70% of candidates complete the assessment. Candidates are used to being ghosted these days. They apply and it's a black box, and they just want a chance to show what they can do. So we see quite high assessment completion scores. From there, we see a range of 20 to 50% improvement in on-site-to-offer rate, and an average of 10 hours saved per candidate. So this changes your process in terms of speed and in terms of quality.

**Katie Fairbank**

Now, we can't talk about hiring today without talking about cheating and fraud. If you're feeling like things are exploding, yes, you are correct. We looked at all of our assessments over an 18-month period to see how many flags there were for cheating and fraud, and we saw a 4x increase. This was highest in early career roles. So this isn't a blip — this is a structural change that we need to be thinking about at all stages of our hiring process. At CodeSignal, we've identified really four types of cheating. The first is the classic copy and paste — going on Reddit or some other online forum to copy someone else's solution and paste it. The way you mitigate that is through assessment design: you need tons of variations of the assessment that are all calibrated to be equally difficult, so even if a question does leak, it really doesn't matter. You might also consider looking at browser activity and even disabling copy-and-paste functionality entirely.

**Katie Fairbank**

The second form we see a lot is hiring a ringer to take the assessment — asking a friend who's really good at this to take it for you. The fix we often employ is ID verification at the start of the process. We take a look, make sure the person taking the assessment is who they say they are. And then we also proctor the entire assessment so that the friend isn't walking in halfway through. The third one is really what's changed and caused primarily that 4x increase: using AI to generate answers. This is a tricky one because for some folks AI is considered cheating and for others it's not. If you are in that camp of saying we don't want candidates to use AI on an assessment, you're going to want to make sure you can flag when they do. You'll want to look at things like behavioral patterns — typing and speech analysis, browser activity, where they're clicking into — and make sure you're controlling AI access. The fourth one is really scary, and there is no single solution for this one: deepfakes. Train your interviewers and make sure your assessments are looking for some tells that deepfakes might have. AI has problems doing things like moving their head naturally. You'll also want to make sure that you're recording the entire assessment session to see if there is any blip in activity.

**Katie Fairbank**

What you're going to want to do at the end of the day is protect your integrity. Integrity means different things to different people, so you need to sit down with your company and say, in the roles that we're hiring for, is using AI a critical part of the job? If so, you may want to consider allowing AI in your assessment and in fact embedding it. One thing that often gets overlooked: if you say candidates can use AI, some candidates are paying for ChatGPT, some aren't. Some have access to Claude, some don't. So you want to make sure that everyone is given equal access and equal opportunity. And you want to communicate your policy really clearly. We've been on a tour talking to a lot of candidates over the last few months, and one of the big things that comes up is candidates saying, I really don't want to cheat — I want to be honest — and I'm afraid that I'm going to do something that is considered cheating. So make sure you clearly communicate that policy out to them. And make sure your hiring process aligns to your AI stance and represents the work that candidates will actually be doing if they get the job.

**Katie Fairbank**

Now let me show you what this works like in action. Here we're going to start in Ashby, and we have a candidate that's applied. We're going to go ahead and send her a CodeSignal assessment. On my side as the candidate, I get an email inviting me to take the assessment, and I can click right through to get started. I'm in an IDE that looks like what I'd use on the job — debuggers, a front-end preview, an integrated terminal — and I can even embed Claude Code directly in the IDE if my company allows it. This isn't just for technical roles though. For a product manager role, a relevant work sample might be building a PRD that includes diagrams, clear written explanations, scoping, and so on. For a sales example, you might want to see how your candidates handle a negotiation and procurement conversation before moving them forward. For finance, working in a spreadsheet and using all the functions you'd expect as part of that workflow.

**Katie Fairbank**

After we're complete with the assessment, you as the recruiter are going to get a report. We're going to assign an overall score, built by PhD I/O psychologists using millions of data points to be super predictive. We'll also break down each of the skills that were measured, assign a proficiency level to each, and allow you to dig into each of the questions that were asked. We give you a qualitative overview of how well they did against a rubric, and you have the full context of what was asked and what was submitted. In this case, one of our candidates actually did something that was potentially suspicious — we have flagged that on the coding report, and broken down on each question what specifically happened. You can click through to view what the suspicious activity was. In this case, they pasted something suspicious, and I can even click through to see what that paste event was. If you say I don't have time for that, we have a full service proctoring team that could do that for you.

**Katie Fairbank**

By the end of the year, there are going to be over 3,000 CodeSignal assessments for pretty much any role you can do at a computer. How do you know which one to use? We have Cosmo for that. Cosmo is our AI partner and agent. I can just ask Cosmo, hey, what assessment should I use for this role? Or I can paste in a job description. Cosmo is going to look through CodeSignal's vast assessment library, look through what you have in your organization, and match you to an assessment that's the right fit. We'll tell you all of the skills that this is assessing, and we'll even give you a sample report so you can see what the output would look like. We have talent scientists do research and validation studies on all of our assessments, so we'll link through white papers where we have additional technical data. Let's go back into Ashby and see how all of this flows back through. Here we have a candidate that's a little later in the process. They've already done their CodeSignal assessment, and we can see we have that score directly in Ashby, so we don't have to leave the system to see it. We can even click through to view that full report if we want to dive in deeper.

**Katie Fairbank**

If I were to leave you with some action items: first, think about taking a skills-based hiring approach if you haven't already. Second, use your assessment at the top of the funnel to establish skill as the very first signal. Third, make sure you have strong safeguards to make sure you can trust the results — because if you can't trust those results, you're passing bad data all the way through the funnel. Fourth, make sure your assessment is job relevant — that's going to lead to the most predictability. Use work samples like we looked at today. And finally, make sure your hiring process aligns to your AI stance and represents the work that candidates will be doing if they get the job. After doing all of this, you'll save time, you'll make the right hires, and you will focus your efforts in the right place. Thank you so much.

In this session, see how Ashby and CodeSignal put real skill measurement at the center of the entire hiring workflow. Presented in partnership with CodeSignal.

Skills-Based Hiring From First Click to Final Offer

Speaker

Recommended Sessions

AI Across the Hiring Lifecycle

What's New in Ashby

Product

Resources

Company

Developers