Recruiting AI, Chatbots, Text Messaging & John Sumser

Recruiting AI Chatbots Text Recruiting and John Sumser: What the HR Examiner Interview Revealed

When 90% of career site visitors leave without applying, text recruiting and recruiting AI chatbots stop being nice-to-haves and become the operational fix.

That is the thread running through a conversation between John Sumser — Principal Analyst at HR Examiner and one of the most respected voices in HR technology — and Jonathan Duarte, co-founder and CEO of GoHire. The two had a long professional history before this interview. Their 2018 HR Examiner conversation about recruiting AI chatbots, text recruiting, and the future of conversational hiring technology is as relevant now as the day it was recorded.

This page captures the full transcript of that conversation, adds context on what has changed in recruiting AI chatbots and text recruiting since 2018, and highlights the key insights Jonathan shared on the Gartner curve for AI, the difference between scripted and AI-powered chatbots, and why text recruiting outperforms career sites for hourly and frontline hiring.

What Recruiting AI Chatbots Text Recruiting and John Sumser Covered on HR Examiner

John Sumser has been covering HR technology since before most HR software companies existed. Jonathan Duarte first met him around 1998 — the early days of internet recruiting — at a cocktail party John hosted in Mill Valley, CA for a small group of HR technology founders attending the SHRM conference in San Francisco. The group was small then. The conversations ran long, ending in discussions about mergers, acquisitions, and the future of recruiting technology in a hot tub. A memorable evening by anyone’s standard.

By 2018, both Jonathan and John had watched that small HR tech world explode. Jonathan had pivoted from early internet job boards into building GoHire, a platform combining text recruiting and recruiting AI chatbots for high-volume hiring. John had become one of the most sought-after analysts in the space, running HR Examiner’s Executive Conversations podcast series as a platform for sharp, unscripted discussions about where hiring technology was actually going — not where vendors said it was going.

Jonathan joined John on the podcast to talk through recruiting AI chatbots, text recruiting, the Gartner hype curve for AI, and the ethics questions that come with automating hiring decisions.

Before the transcript, two “Sumser Quotes” are worth preserving — they capture how John Sumser challenges received wisdom in HR tech:

“Who would ever want to join a social network of other unemployed people? That makes no common sense. Who wants to hang out with a bunch of equally depressed unemployed people?”

John’s response to the question “What do you think about Jobster?” at the Jobster launch party at ERE in San Diego, 2004. Another little company would surface from that era: LinkedIn.

“Here’s the problem with solving diversity in the workforce. Say your company has been successful for years, and the one thing that’s unique about your culture is that everyone wears a clown suit. As business continues to grow, you need to bring on more people. So you interview some great people and hire who you think will be a great employee. Then, on the new recruit’s first day of work, they show up bright and early… wearing lederhosen! Now what do you do? Do you make them change to a clown suit? Do you make everyone else wear Lederhosen? Do you find a way to embrace clown suits and lederhosen?”

If you ever get a chance to sit down and have some deep conversations with John, highly recommended. Bring an open mind, good ideas, and data.

Key Themes: Recruiting AI Chatbots, Text Recruiting, and the Hiring Technology Curve

The HR Examiner conversation with Jonathan Duarte covered four topics that have only grown in significance since 2018. Here is what each one revealed about the state of recruiting AI chatbots and text recruiting at the time — and what has changed.

Why Career Sites Fail and Text Recruiting Works

Jonathan’s core argument for text recruiting starts with a conversion problem. Over 90% of career site visitors leave without applying or providing any contact information. That number — tracked across 2.5 million clicks and applies via Appcast and Jobvite — means most recruitment marketing dollars are paying to send candidates to pages that lose them. For an hourly worker standing in front of a “Now Hiring” sign, typing a 22-character URL into a mobile browser to reach a desktop-formatted career site is a broken candidate journey.

Text recruiting solves this by replacing the career site as the first step. A candidate texts a keyword to a phone number. GoHire’s text recruiting platform receives that message, engages the candidate immediately with a short structured conversation, collects contact information, and qualifies the candidate — all in the messages app they already use. No app download. No resume required. Completion rates on SMS-based conversational applications consistently outperform career site application completion rates because the candidate never leaves their native texting environment.

Scripted Chatbots vs. AI-Powered Recruiting Chatbots

One of the most useful distinctions in the HR Examiner interview is Jonathan’s taxonomy of recruiting chatbots. He separated the landscape into two categories: scripted chatbots and AI-powered chatbots. Scripted chatbots run a deterministic decision tree — question, expected answer, next question, fork on response. They use no machine learning or natural language processing (NLP). They are less expensive to build and easier to implement, and they solve real problems for the majority of the mid-market.

AI-powered recruiting chatbots go further. They use NLP to interpret freeform candidate responses, handle unexpected inputs, and improve over time. A candidate who types “yeah I can work mornings or whenever” to an availability question is understood and routed correctly by an NLP-enabled chatbot — a scripted bot would fail or escalate. In 2018, Jonathan’s assessment was that full AI chatbots were real but expensive and primarily accessible to enterprise buyers with multi-year implementation timelines. The mid-market needed the scripted model first.

GoHire’s AI recruiting chatbot has since evolved to incorporate both layers: structured conversation flows that deliver high completion rates, combined with NLP interpretation for edge cases and freeform responses. The cost and complexity gap between scripted and AI chatbots has narrowed significantly since 2018.

The Gartner Curve and Where Recruiting AI Actually Is

Jonathan referenced the Gartner hype cycle in the interview to explain why the gap between AI headlines and AI reality was so large in 2018. Every new technology generates a peak of inflated expectations — venture capital pours in, vendors claim capabilities they can’t fully deliver, and enterprise buyers get burned by implementations that overpromised. Then the trough of disillusionment. Then, gradually, the slope of enlightenment where companies with genuine product-market fit survive and deliver real value.

In 2018, recruiting AI was at the peak of inflated expectations. Jonathan’s view was grounded: the market was not ready for the fully autonomous AI recruiter that the hype suggested. Workflows in HR and recruiting vary enormously across companies. ATS stacks were fragmented across 220+ vendors. The technology needed to catch up to the operational complexity of real hiring environments before AI would deliver consistent mid-market value.

By 2026, recruiting AI has moved well past the trough. The companies that survived — including GoHire — did so by solving specific, high-value problems: pre-screening at scale, automated interview scheduling, and candidate re-engagement. These are not the autonomous AI recruiter of 2018 conference decks. They are narrow, reliable, high-ROI applications of machine learning and conversational AI in a hiring context.

AI Ethics, Candidate Bias, and Compliance in Recruiting

John Sumser pressed Jonathan on ethics in the interview — specifically the diversity and bias implications of automated hiring systems. Jonathan’s response was candid: GoHire, at the time, was not doing automated decision-making on applicants. The platform was capturing candidate information and routing it to ATS systems for human review. No one in GoHire’s system was classified as an applicant — they were candidates whose information was being collected for recruiter evaluation.

That distinction matters for compliance. Systems that make or influence automated employment decisions — particularly pre-screening that advances or eliminates candidates — fall under EEOC disparate impact doctrine. NYC Local Law 144, enacted in 2023, requires bias audits for automated employment decision tools used in New York City hiring. The Colorado AI Act, effective June 2026, expands these requirements further. Any recruiting AI chatbot operating in these jurisdictions needs a clear answer to the question John Sumser was already asking in 2018: who is responsible for the automated decision, and how is bias being audited?

GoHire’s approach today: pre-screening is human-in-the-loop, qualification criteria are role-relevant and documented, and the platform provides explainability at the candidate level. Bias is a design constraint, not an afterthought. This aligns with what the regulatory environment is now requiring — and what John Sumser was asking about before most of those regulations were written.

Full Transcript — Recruiting AI Chatbots Text Recruiting and John Sumser (HR Examiner, 2018)

Complete transcript of the HR Examiner Executive Conversations episode featuring Jonathan Duarte, co-founder and CEO of GoHire, on recruiting AI chatbots and text recruiting:

[00:00:00]

John Sumser: Good morning and welcome to HR Examiner’s Executive Conversations. I’m your host, John Sumser, and today we’re gonna be talking with Jonathan Duarte.

John Sumser: Hey,

Jonathan Duarte: how you doing, John? Great. Great. I’m ready to roll.

John Sumser: Cool. So why don’t you take a moment to introduce yourself.

Jonathan Duarte: Yeah, so my name’s Jonathan Duarte. I’ve been in the space quite a long time. I started one of the first internet job boards back in 96 and have been on the forefront of a lot of technology, but in the kind of back of the atmosphere, but have basically been running software companies that build technology from the top of the funnel all the way back down the background checks and connecting all the dots.

John Sumser: How’d you end up doing this?

Jonathan Duarte: Yeah. Long story short, so I was an ERP consultant back in the day for JD Edwards, a couple years out of college. So this is now like 93. And there really was no real internet, at that point, or not the web. And I got into it because in 96 when I started Go Jobs, a friend of mine had told me about this internet thing like the year before. And I was consulting, living in hotel rooms and said, this just isn’t gonna work. I was 23, you’re supposed to be out having fun. And I was working in hotel rooms consulting. So I quit my job and said, okay, I’m gonna find something new to do. And the whole internet thing was just blowing up that year. Netscape went public. And I was technologically savvy enough to understand where it was going, and jumped on board.

John Sumser: So what are you doing today? You’re a pioneer and now you are doing it again. What’s the story?

Jonathan Duarte: Yeah, so I think three years ago I got really excited about text messaging. I’ve got two kids, an 11 and 13 year old, and that’s how they communicate. And I got interested in how it’s working on the B2B side because it basically is not being used at all other than in marketing and then in some smaller areas. And since I started in 96 my premise has always been how do we get candidates closer to employers and employers closer to candidates at the right time and the right place. And what I saw with text messaging and just messaging in general is it is almost like the holy grail of being able to communicate at scale in rapid fire. So I started building chatbots in that space to try to build recruiting automation using that platform.

Jonathan Duarte: What does that look like? A real life example is: In the US right now, unemployment is pretty much at a 50 year low. And if you look at any retailer, you go to IKEA, you go to Walmart, you go to Sam’s Club, every retailer has a sign out front that says “Now Hiring.” But the problem is most of those signs lead a candidate on a mobile phone to type in 20 characters to some career site. That doesn’t work. The conversions are terrible. So the alternative to that is use mobile friendly solutions, which is a text messaging platform, and have somebody type in the word “jobs” and text that to a phone number and instead of redirecting someone to a website, you just get their contact information immediately. You ask them what their first name is, their last name, what kind of job they’re looking for, where they’re located. And you get 90% of the people that complete that conversation versus a career website which only converts less than 10% of the users. So you lose 90% of the people that ever show up. Recruiters then have the ability to communicate over text and get that kind of scalability, which since nobody’s opening emails anymore it just seems to be working and working in a very big way.

John Sumser: That’s interesting. Let me see if I can get what you’re saying here. If you use your smartphone to look at a career site — while they have made them fit the screen, the interface is still as if it were on a desktop. And so it’s very hard to fill out a form. When people come to career websites on their phones, the failure rate is quite high. Is that right?

Jonathan Duarte: Yeah, absolutely. With your two disposable thumbs, trying to type in 20 characters like homedepot.com/careers, and then try to communicate — web hasn’t really gone that far. In the nineties we created Web 1.0, and then we created Web 2.0, which were now forms that just dumped stuff into a database. We haven’t really got to Web 3.0 because the web interface is one dimensional. But text messaging is two-way, and that’s the difference — you can communicate at scale, which you couldn’t do on a webpage.

John Sumser: That sounds pretty counterintuitive. If you’ve got a form that can be filled out, so you ask all the questions at once, it’s pretty counterintuitive to say that a better way to do that is ask the questions one at a time.

Jonathan Duarte: Yeah. And I agree. When we were first testing this stuff, it was like, this is just weird. But we were going to test a hypothesis. My co-founder Mike and I created a chatbot on Facebook Messenger in 2016. And our hypothesis was, let’s just see if people would communicate over messenger and look for a job. We had no idea what was gonna happen. Surprisingly enough, that chatbot — GoBe, our first one — went viral to 103 countries in 30 days with no marketing. And we’re like, wow. People will use their thumbs instead of a keyboard. And that’s the majority of our workforce. They don’t sit behind computers every day.

John Sumser: The majority of the workforce doesn’t sit behind computers every day. Do you have stats about that? That’s really interesting.

Jonathan Duarte: Yeah. The hourly part-time workforce in the US is something like 60 million. And if you look at retail, hospitality, construction, logistics, truck drivers, those numbers start going over 50% of the US workforce pretty quickly.

John Sumser: So now we’ve got a system that asks you the questions one at a time. Is there anything to a chatbot that’s more than that?

Jonathan Duarte: Yeah, absolutely. You can see what we call the Gartner curve, about early adopters and how interesting everyone got in 2017 about AI. The reality is you can see that early part of that Gartner curve where everyone gets really excited about a new technology. And the reason you see that is because everyone’s throwing ideas at it, but it takes a lot of ideas to actually get to solve real problems. So that curve starts dropping off. Then you start coming back with the entrepreneurs who really got what we call product market fit. And I think we’re still on that kind of top of a lot of people throwing stuff at AI. Our perceptions are that we should be driving cars autonomously. But recruiting is still a very workflow intensive process. And I think AI is going to be coming — it’s coming — but are we gonna see it in the normal mid-market right away? No, it’s still gonna take a little time because everyone, especially in HR, their workflows are different. The tech stacks are different. We’ve got 220 ATS companies. It’s gonna take some time before the technology catches up to the reality of HR and recruiting and a human workforce.

John Sumser: So when you talk about chatbots, you’re talking about a tool that uses text to acquire the exact same information that you’d have on a job application on a website. Is that right?

Jonathan Duarte: Yeah. And it might be a little bit — I would say it’s not quite as good as the data you could get from a webpage because someone sitting behind a keyboard has a lot more keys and it’s easier to type. We know that there’s a failure rate of candidates applying via mobile, which is about less than 30% of candidates apply via mobile, but 70% of candidates are using mobile to find jobs. So there’s a drop off. The professional workforce that’s computer-driven — the one with Word document resumes — is still gonna be web based. But your hourly server at TGI Fridays, or the hotel front desk, or the Uber driver — these individuals don’t need resumes to get hired. They just need a structured sequence of questions, and those are usually short and quick.

John Sumser: So I keep hearing people talk about chatbots that are much smarter than this sounds. This is like you pay somebody to stand on the street corner and read the questions on the job application to the person applying and collect the answers one at a time. And I keep hearing about things more sophisticated than that, but it doesn’t seem like it’s actually happening.

Jonathan Duarte: Yeah. There are companies out there putting together really sophisticated AI and NLP programs, and we use some NLP or natural language processing within the chatbot as well. But we’re not trying — I think this is the difference. As a 20 year exec in this industry, it’s about trying to solve the problem today. We’ve got really big problems today. That’s where the market’s going to buy for the most part. There are companies and early adopters in the enterprise who have bigger problems and they need a longer term solution. So they have the money and effort and resources to invest in technology for five or ten years. But that’s not gonna hit mid-market anytime soon. And when you say, oh, we have a chatbot that can essentially just read your brain — that’s not really gonna happen. But you can have a chatbot that will walk you through a script, help employers get more candidates, especially in retail where they’re just dying for candidates right now, very quickly. And the conversion rates are much, much higher than a webpage.

John Sumser: Do you call those two different kinds of chatbot different things?

Jonathan Duarte: There are really two things. There are scripted chatbots — customized question, answer, question, answer, question, answer — and you can have some derivations or fork those conversations. That’s lower tech. Then there’s having a computer try to understand open-ended questions and then relate that to a specific job versus a category of jobs. So the artificial intelligence is really about trying to understand open text, responding accordingly. In a scripted technology it’s much cheaper and easier to implement than trying to use the AI part of it. So that’s where we’re focusing more — on the scripting side. That’s where the immediate potential for the majority of the market is. While the AI is coming, it’s just gonna be really expensive for a lot of people to implement.

John Sumser: Tell me about the work required to install the kind of chatbot that you guys do.

Jonathan Duarte: So our clients are looking for either a web chat that sits on a website — imagine a candidate showing up to your career site, and 90% of the people who visit your career site just disappear — or there’s the text chat, which is really just texting into a phone number. What it takes to actually implement one of these things: we have the technology in the backend. What we do is help the client write the script for what they want to do. We’ll go ahead, write the script, provide the technology, then hand it over to the client and give them a phone number. In many cases we can do that in 24 hours.

John Sumser: Are you doing thousands of these or how far along are you?

Jonathan Duarte: We’re still into the hundred mark. It’s the education process. There are a lot of other real big problems in recruiting and trying to take the time to understand this as a TA leader — what the benefits are — has been hard. So it’s new technology. We just don’t have the bandwidth of the VPs of HR to understand what this stuff can do. And of course, because it’s a new market, there aren’t that many case studies. So it’s an education process. As the market gets educated and they understand that it’s quick to implement, it doesn’t take a lot of internal resources, and the return on investment is much higher — it’ll get to market faster. That’s where I think most of us are. There’s a lot of venture-backed companies in this space and we’re self-funded. So it’s a little bit different challenge of how do you get to all these people as quickly as possible.

John Sumser: One of the things I’ve noticed is it’s very difficult for buyers to tell what’s a good company to do business with and what’s not. People go, oh those folks with the really nice offices with the big view of San Francisco and $50 million in backing — that’s a good business. And these little people in bad offices with bad lighting and lots of pizza boxes — that’s a bad business. How do you know?

Jonathan Duarte: Totally. And there’s a lot of this. Here’s the thing: as a VP of TA, if you’re gonna go walk into the CEO’s office and tell them you’re gonna work with this company who can barely get a presentation together because they’re not marketing people but their technology is awesome and the ROI is great — you’re taking a risk. Versus the guy who shows up on Forbes magazine with the greatest thing since sliced bread. That’s an easier sale and the risk seems less. So how do you get across that? I think the best way of looking at it is: when you’re looking at a platform of technology, find a vendor who knows your space and can solve your problem. Any technology in HR is about solving a problem. If the vendor can solve your problem and understands your problem and has an innate understanding of your business and your workflow — they’re the people you wanna work with. It doesn’t matter how much venture capital they’ve raised and lost because they don’t understand your problem, or how great their marketing looks. It’s really about finding a company that can understand your problem and implement what you’re trying to do.

John Sumser: That’s a tough message to carry. That’s the old smelly person in the corner is better than the guy in the shiny suit.

Jonathan Duarte: Yeah, absolutely. When you see those 40-by-40 booths at HR Tech in November, just ask the sales person how many years of experience or what kind of workflow problems they’ve been able to solve. And then head right over to the cheek booths on the side and the non-venture guys. You’ll find somebody who can rewrite your entire code base in a couple of hours.

John Sumser: A lot of what goes on in AI involves using open source stuff. Tell me about that.

Jonathan Duarte: We actually started our first chatbots on the Facebook platform using a tool called Motion.ai — which we quickly realized wasn’t scalable. Somebody else owned the code. That was good to do what we call a proof of concept or MVP. Then for the AI side, there are lots of tools for NLP that you can use — like Wit.ai, which was purchased by Facebook, and Motion.ai, which was bought by HubSpot. There’s also IBM Watson — more expensive, but stable, and many people are using it. So there’s that question: do you use something free or low-cost that might get you there, or do you pay for something more stable? After some time doing this and doing pilots with clients, we actually built our own framework. We own our own code, every line of it — versus other companies that are outsourcing parts of their code, which then causes problems when a client needs a customization. Anytime you outsource part of your code with open source, you can get to market faster, but you have the potential that you don’t own what you’re doing. There are chatbot companies in this space saying “we’re the greatest thing since sliced bread” and they don’t own any of their code, so their scalability is going to become a problem.

John Sumser: Last thing — there’s a cloud of ethical issues around the new technologies. What are the big ethical issues in your work?

Jonathan Duarte: Here’s my take on AI. In my opinion, AI is going to be a hard sell to the VP of HR when you’re talking about a computer answering a question on behalf of the company. You can’t give a computer the opportunity to answer a compliance question yet, and I don’t know if we’re ever going to completely outsource that. I don’t think any VP of HR is ever gonna sign a check to a company who says “we’re gonna outsource your compliance.” So I think we’re at this point where AI can be useful, but it’s going to always be a kind of a partner. It might be the person who sits next to you as a tool, but it’s not going to be the front person for a lot of solutions. And the other side is the diversity and bias problems you have within automated systems. We don’t actually deal with that much because we’re not trying to scrape resumes and match them. We stay away from the applicant side. At this point, SMS is not a good application for an actual apply — it’s a good communication tool, but it’s not the be-all end-all for OFCP-type compliance. No one in our system should be considered an applicant. We forward all candidates into the ATS to go through the application process.

John Sumser: Is there anything you want to make sure readers take away before we’re done?

Jonathan Duarte: Going back to what you mentioned about picking a vendor who knows the industry: find someone who actually understands your problems and your company, and leverage their knowledge base instead of just listening to the sales pitch. That’s my whole pitch — because I’m one of those guys. I don’t have the 40-by-40 showing up at HR Tech, but I’m sure we could have some great conversations about solving problems in your business pretty quickly.

John Sumser: So reintroduce yourself and tell us how to get hold of you.

Jonathan Duarte: My name’s Jonathan Duarte, co-founder and CEO of GoHire, and that’s gohire.com. You can reach me at jd@gohire.com.

John Sumser: Thanks so much, Jonathan. It’s been a great conversation. You’ve been listening to HR Examiner’s Executive Conversations.

What Has Changed in Recruiting AI Chatbots and Text Recruiting Since 2018

Jonathan and John Sumser recorded this interview at a particular moment: AI in recruiting was generating maximum hype and minimum proven mid-market outcomes. Much of what Jonathan said in 2018 turned out to be accurate, and the evolution of recruiting AI chatbots and text recruiting has followed the arc he described.

The scripted-vs-AI distinction Jonathan drew has largely resolved in favor of hybrid systems — scripted conversation flows with NLP layers that handle edge cases. The cost of NLP has dropped dramatically as transformer-based models became commercially accessible. GoHire’s AI recruiting chatbot now interprets freeform candidate responses, not just multiple-choice inputs. The candidate who types “yeah I can do mornings” is understood and routed correctly.

Text recruiting has moved from experimental to standard for high-volume hiring. SMS reaches every mobile device without a platform account requirement. GoHire’s platform operates on 10DLC-registered infrastructure, TCPA-compliant, with full opt-out mechanics built in. Candidates apply by text, get pre-screened over text, and receive their Text Invite — “Reply YES and I’ll send a few available times” — over text. The recruiter’s calendar syncs interview slots back into the SMS thread automatically. No phone calls. No email chains.

The ethics questions John pressed Jonathan on in 2018 are now regulatory requirements in several jurisdictions. NYC Local Law 144 mandates bias audits for automated employment decision tools. The Colorado AI Act, effective June 2026, requires disclosure and impact assessments for AI systems that influence consequential decisions including hiring. These were philosophical questions in 2018. They are compliance requirements now. Jonathan’s instinct in the interview — that human-in-the-loop review needed to precede any AI-influenced hiring decision — aligned with where the regulations have landed.

GoHire’s Recruiting AI Chatbot and Text Recruiting Platform Today

The platform Jonathan described building in the interview — a text recruiting and recruiting AI chatbot system for the high-volume, hourly hiring market — has compounded considerably since 2018. GoHire now processes thousands of candidate conversations per month across retail, hospitality, healthcare, and logistics verticals.

The core product combines four capabilities: Apply by Text (candidates text a keyword to apply in under 60 seconds), two-way text recruiting (TCPA-compliant platform for recruiter-to-candidate messaging), an AI recruiting chatbot (pre-screening, FAQ handling, 24/7 availability), and automated interview scheduling (candidates self-schedule via the Text Invite SMS flow). Per SHRM, 94% of qualified candidates accept the first offer they receive. GoHire’s platform is designed around one principle: engage the candidate first, faster, before your competitor does.

GoHire customers consistently reduce time-to-offer from 7-14 days to 24-48 hours. 92% of qualified candidates self-schedule their interview within 24 hours of receiving the Text Invite. No phone calls. No email follow-up chains. The recruiting AI chatbot handles pre-screening around the clock so recruiters spend time on the candidates who qualify — not chasing the ones who don’t.

To see the same principles Jonathan described in 2018 working in production, explore the top 11 ways to use a recruiting chatbot or review what GoHire’s full recruiting AI platform delivers today. For the text recruiting fundamentals, the text recruiting guide covers everything from TCPA compliance to campaign setup.

AI Ethics and Compliance for Recruiting AI Chatbots

John Sumser’s questions about ethics were ahead of the regulatory curve in 2018. The issues he raised — bias in automated pre-screening, diversity outcomes, accountability for AI-influenced hiring decisions — are now codified in law in several jurisdictions. Any employer deploying recruiting AI chatbots for pre-screening needs to understand the current compliance landscape.

NYC Local Law 144 requires bias audits for automated employment decision tools (AEDTs) used in New York City hiring, including chatbot pre-screening systems. Bias audits must be conducted by independent third parties and results published. The Colorado AI Act, effective June 2026, requires disclosure to candidates when AI systems are used in consequential employment decisions and mandates impact assessments. California has introduced additional AI transparency requirements that affect recruiting AI systems operating in the state. At the federal level, the EEOC’s disparate impact doctrine applies to any pre-screening system — automated or human — that produces differential outcomes across protected classes.

GoHire’s recruiting AI chatbot is designed with these constraints as first principles. Pre-screening questions are role-relevant and documented. Human-in-the-loop review precedes any candidate advancement decision. Algorithmic bias is monitored through regular audit processes. Explainability is built into the qualification trail — if a candidate asks why they were screened out, the system can provide a reason that is both accurate and legally defensible. The approach Jonathan described in 2018 — AI as a partner to the recruiter, not a replacement for human judgment in consequential decisions — reflects how the regulatory environment has evolved as well.