Start With the Pain, Not the Tool: John Keenan on AI in Recruiting at Gong

Show notes

In this episode of the Leaders in Talent podcast, host Adriaan Kolff interviews John Keenan, Director of International Talent Acquisition at Gong, the revenue AI company, where he leads recruiting across EMEA and APAC from Dublin. John has spent nearly 20 years in talent acquisition, from agency recruiting to in-house roles at LinkedIn, Pluralsight and now Gong, and he is also an executive coach.

John's core rule for AI in recruiting is to start with the pain point, not the tool. He walks through how he runs his week through Claude as an operating system, connectors and MCPs feeding a weekly-planning agent off a now.md file, why ATS data integrity now decides the quality of every insight, and the RAPID decision framework he uses to give hiring managers clear ownership. He and Adriaan get into the moment AI-generated content started irritating everyone, why Gong assesses for AI literacy at the top of the funnel, and where AI interviewers actually fit: useful for junior and volume screening, off the table for senior hiring. Not now, and he would not be comfortable being interviewed by one either.

Timecodes

01:09 Welcome and John's background

02:17 What Gong does: revenue AI, 1,400 people

03:14 Don't start with the technology, start with the pain points

05:17 Claude as an operating system: connectors, MCPs, a weekly agent

07:14 Building the planning agent and the human stack

09:05 Why ATS data integrity now decides everything

10:03 Rolling Claude out to the team: licenses, training, AI for coaches

12:34 The AI-generated proposal that irritated everyone

14:07 An agent for mid-year reviews from 1:1 recordings

15:39 AI literacy as a hiring signal

16:18 Fake and AI-polished candidates: is inbound a bottleneck?

18:03 The RAPID framework: clarity on who decides

20:07 AI for bias mitigation and interview consistency

22:03 Will AI interviewers replace recruiters?

24:54 The mobile-phone news clip, and what normalises

25:45 Making time to think and build with AI

27:04 Where to find John

___________________________

Connect with us on LinkedIn: https://www.linkedin.com/company/matchr/

Get in touch with us: https://www.matchr.io/who-we-are/contact/

___________________________

Connect with John Keenan: https://www.linkedin.com/in/johnkeenanlinkedin/

Connect with Adriaan Kolff: https://www.linkedin.com/in/adriaankolff/

___________________________

RSS feed: https://media.rss.com/leaders-in-talent/feed.xml

Transcription

[00:01:09] Adriaan: All right, ladies and gentlemen, welcome back to another episode of the Leaders in Talent podcast. Today I'm joined by John Keenan, who is the International Director of Talent at Gong. John, welcome.

[00:01:22] John: Thanks very much for having me. Pleasure to be here.

[00:01:24] Adriaan: John, can you give our listeners and viewers a little more context on some of the companies you've worked with, what your current role as Director of Talent at Gong looks like, and what Gong does?

[00:01:39] John: Sure. So I've been in TA in various roles for nearly 20 years. I spent my first five years in agency recruiting, and then pivoted into tech over the past 14, 15 years across LinkedIn, Pluralsight, and now Gong, where I lead recruiting internationally. That entails everything in EMEA and APAC, and I've been with Gong for just over four years.

[00:02:16] Adriaan: And what does Gong do?

[00:02:17] John: We're a revenue AI company, very much within the AI space. We have about 1,400 people globally, and our European headquarters is Dublin, which is where I'm based.

[00:02:34] Adriaan: What does revenue AI mean?

[00:02:36] John: In its simplest terms, it's a platform which helps salespeople get insights at scale, to drive performance, coaching, next steps, deal efficiency, and forecasting, all in very much an orchestration system. That's what we do.

[00:02:56] Adriaan: So the company is at the forefront of everything happening with AI. I could argue that is a huge opportunity, but also potentially a risk, given how easily some data is now available and what people can build themselves. At the same time there's the unlock of all the different touchpoints. Before our conversation, you mentioned you're doing things with AI within the recruitment team, and that you don't start with the technology, you start with the pain points. Can you elaborate on that?

[00:03:33] John: More than ever, it's very easy with AI to look everywhere, to ask where can I put it in. All these bright, shiny new tools, and they can end up not addressing the pain points or the business challenge or the issues in your specific funnel that need attention. So diagnosing the pain point is key. From a recruiting perspective, it's about breaking down the funnel and understanding, number one, where can AI be implemented really efficiently, and then, what's the major business challenge? One example would be driving alignment with interviewers. Some of the tools we use can help us do that. There's a lot of that kind of work, designing job specs, Boolean strings for specific searches, reporting and analytics, interview intelligence, recording interviews and deriving insights. Those things are here already. And then you've got the likes of Claude and the frontier AI companies, which are becoming the operating system that everything drives through. That's almost a different conversation.

[00:04:56] Adriaan: You're using Gong, BrightHire, Claude, and an ATS to drive that. Tell me more about that setup, how you use it day to day, and what its strengths are.

[00:05:17] John: Like everybody else, I'm on the journey with AI, but I'm very much an avid Claude user. I use Claude as an operating system: everything flows through it. Anything I can get Claude connected to, I connect, which is another valid point, getting the MCPs and the right connectors in place. I have all my project folders in there, reporting and analytics, team-based performance. I feed that in on a weekly basis. As an efficiency gain, I have an agent that looks at my priorities for the week, helps me define those by pulling info from these systems. I come in at the start of the week and the agent says, you need to prioritize these, you need to look at that. So everything flows through it. It's an operating system, and then it's about plugging it into the core systems you use. That's really where the power is.

[00:06:22] John: Some companies, like Greenhouse, just launched an MCP recently, and we're starting to use that. BrightHire is another system we use for interview intelligence, and they're defining an MCP too. So it's about getting all those connectors set up. In the next six months to a year, everything flowing through these frontier companies and connecting with them is going to be the key.

[00:06:56] Adriaan: I love that. Help me understand practically and operationally: how did you create that weekly update for yourself? What data pieces go into it so you have a clear view of what to focus on?

[00:07:21] John: This was my first agent, I'll own that. The interesting thing about building an agent is where you start, which is what's called building a human stack. These are files you build about your working style, who you are, what your voice is, so you almost program the AI with that. For weekly planning, there's a now.md file which, on a weekly basis, you update with your priorities, and then it pulls everything else from the other systems and marries it in and tells you where to focus. For example, it looks into Slack and asks, have you missed anything there? If my priority is performance reviews, it looks into previous data on my team and comes back with action items and points to work on.

[00:08:32] Adriaan: Is there also something, or is it work in progress, where you look at the funnels across different roles and it flags, be aware, this funnel is behind on interviews?

[00:08:47] John: That's a work in progress, and the reason is those MCPs are still in beta, like Greenhouse. That's absolutely the intent of what these agents will do. But access to the data is key, and making it connect, because right now some of it is manual, you're loading things into Claude to get some of these insights. The core thing a lot of recruiting leaders are seeing is that data integrity in the ATS has always been important, and it's even more important now, because the process flows have to be really adhered to. What you put into AI, what you get out is only as good as what's there. If the data is incorrect, it goes to bad data and gives you poor insights. If the data is on point, you get really crisp insights at scale. So a lot of TA leaders, including myself, are very much on that journey of data integrity.

[00:09:46] Adriaan: How do you ensure that gets done, given some recruiters are better at it than others? How do you do that with your team? Does everyone have free access to Claude? Are there specific guardrails?

[00:10:03] John: On the Claude side, we started the journey about four or five months ago with enterprise licenses and built from there. In terms of training, I wouldn't over-engineer it. Claude do a very good certification, Claude 101, which is good for getting up to speed. I'm also an executive coach, and I'm doing an external course in training AI for coaches, which has been excellent. It starts with building your understanding: what's the model for you, what's the LLM for you to use, how to program it, the human stack, agentic building, and how to coach with it. AI for coaches is a really interesting area. It can derive patterns you might not see when you're coaching someone, and that feeds into the team as well.

[00:11:02] Adriaan: Are you using the AI to coach you as a coach?

[00:11:09] John: No, it's to coach other people. When I'm coaching somebody, you record the session and feed it into Claude, and you prompt it accordingly, so it's all about the quality of the prompt. It gives you insights and takeaways, and as you build that within a specific project, the intelligence builds and it gets to know these individuals more. There are confidentiality things you do: you don't mention names. Then it builds at scale. For example, after a coaching assessment I need to build a one-pager with action points and next steps for that individual. That can be automated and done after you leave. At its best it's pattern recognition, things you might miss, plus automated follow-ups. But you absolutely need the human nuance. Merging an AI output with human nuance and making it sound credible is an emerging skill, because I'm noticing now people send me things made with AI, and you can tell.

[00:12:34] Adriaan: I was about to bring that up. We're proposing a big RPO project tomorrow, and I spent tons of time back and forth with Claude creating the proposal. All the input was mine, only the output and the way it looked was AI-generated. It spoke in my tone, no dashes anymore. I was genuinely excited by what it looked like, the visuals, and I sent it to my co-founder and to the lead of the project, and they were both so irritated. Being on the receiving end, when someone sends me an AI-generated email, it's so obvious.

[00:13:13] John: It's almost insulting, isn't it? And that's happened fast.

[00:13:26] Adriaan: So in our management team we made an agreement: we do not share any AI-generated content in email unless we say specifically at the top, AI touched it up, but these are my own words. We're very careful with it, because at one point I had a disagreement with my co-founder and we were both using AI to convince each other over email. It was horrendous. At some point you lose each other.

[00:14:00] Adriaan: It's one of the things I'm looking into now, especially given your executive coaching work. We just went through mid-year reviews, and the problem I always have is that the examples I remember are from the last four weeks. So what I'm doing now is building an agent that collects all the recordings of our one-on-ones and, within the mid-year review framework, surfaces the stuff from four or five months ago based on real conversations. Any thoughts on anonymization and doing that the proper way?

[00:14:49] John: I totally agree with that approach. It's about how you use it. With Claude, it's ensuring you've got it programmed and that you do it in specific projects it remembers, so it derives insights as the intelligence builds. I'm doing exactly the same with performance reviews. I'm living in Claude, that's the operating-system piece. You have to be quite clear in how you classify things: your team, your reporting and analytics, and then you're feeding the memory all the time so it can curate and derive insights. The challenge is that companies are training this, but people are learning at different speeds. It's almost a whole new concept, AI literacy.

[00:15:39] Adriaan: Definitely. Several of our clients, and it's early days, are now hiring with AI fluency frameworks. Is that something being discussed within Gong?

[00:15:52] John: It's been assessed. AI literacy is dependent on the role, obviously. When we're hiring, we're very much assessing at the top of the funnel for AI literacy. We're not pushing too hard on it yet. Curiosity, and the ability to conceptually understand it, is the starting point, the minimum standard I'd say. I expect that to become a lot more refined very quickly.

[00:16:18] Adriaan: One thing a lot of our clients are facing, I'm curious if it's the case for you, is tons of fake candidates, AI-generated motivation letters and resumes, and inbound becoming a bottleneck rather than something that helps bring in the right people. Are you experiencing that?

[00:16:38] John: Fraudulent candidates are definitely becoming more prevalent. From talking to my colleagues, and you'd know more about this, it seems to be more of a thing in the US.

[00:17:10] Adriaan: Yes, the US is predominantly where we see this happen. US and Asia, actually.

[00:17:16] John: So I've seen it less personally, but in terms of the higher inbound flow, yes, you're seeing CVs that are a lot more refined. You can update your CV with Claude in 15 seconds, it's very easy. But with the right screening in place and a good recruiter supporting it, you're going to be able to cross-check against a LinkedIn profile. So personally I haven't seen it as a massive issue, but I'm aware it's becoming one.

[00:17:50] Adriaan: In terms of recruitment and business AI, is there anything you're doing to help hiring managers be better at the recruitment and interview process?

[00:18:03] John: I mentioned earlier that driving alignment is one of the biggest challenges in recruiting, and it always has been. Good recruiting teams tend to have that alignment. So decision-making frameworks are built up. One we use is the RAPID framework, which is a Bain decision-making framework. It gives clarity to who in the process is making a decision. R is recommend, A is agree, so that's like the veto, P is process, who processes, I is input, that's the interview panel, and D is decision-maker, so that's typically the hiring manager. Recruiting sits as the R, they recommend. You literally use that language. At the outset of a process you make it very clear who sits where. That matters before any AI implementation, because what I've found over the years is that more junior hiring managers can be heavily influenced by more senior people. If a senior person interviews and they're not sure, that can dissuade a hiring manager from making the decision, which is theirs. It's good to empower them to make it, particularly for their own team.

[00:19:31] Adriaan: How does the RAPID framework help with that particular case, especially when a junior hiring manager might be persuaded by more experienced people? Is it because the D is so clear, it's you that makes the final decision?

[00:19:48] John: Yes, you empower them to make that decision, and it reduces friction in decision-making. When you don't have that clarity at the outset, a hiring manager either doesn't make a decision or defers it, and then you have paralysis in the funnel and things aren't moving. What AI is doing here, and this is something we're working on as a team, is bias mitigation. There's probably never been a better opportunity to look at that as a recruiting leader. A lot of the time I'd be dependent on gut feel, thinking, I'm not sure why there are blockages in this funnel. Now, with tools like BrightHire, that divergence is clear. You can build up those insights at scale and see, for example, that this person, if they're developing rapport with somebody, isn't probing as much as they would with someone else. So it's about driving consistency, making sure they're asking the same questions, being clear on the competencies. A lot of this is simple, but AI gives you the ability to scale it more efficiently. You set the competencies and the decision-making framework at the outset, drive that into the system, and then review at scale how people are addressing them, and debrief at the end. That's a use case we're working on.

[00:22:03] Adriaan: Do you think part of the role of the recruiter is going to be taken over by AI interviewers, in six, 12, 18 months?

[00:22:19] John: It's an interesting topic. I've spoken to BrightHire about this, and the terminology seems to be pre-AI screening. Do I see it replacing recruiters right now? No, I don't. For certain roles, very junior, volume, sales development, there's definitely potential upside with AI screening, for two reasons. There's efficiency: instead of coordinating an interview, you send a link and they have 72 hours to do it. You can program the questions and be clear on what you want to assess. As a recruiter you could go to bed and wake up with 20 people screened in APAC. That all sounds great. But there's downside too: people will be chasing the humans on the team for feedback, so extra volume creates other tasks. The more senior the role, I absolutely would not use it, and I think people would opt out. So there's a place for it, and it will evolve, but it won't replace recruiters right now. It can give efficiency gains, but it's not there for complicated or more senior hiring.

[00:24:27] Adriaan: If you were a candidate, would you be okay speaking to an AI interviewer?

[00:24:32] John: I wouldn't be comfortable at all. But like we've seen over the years, Zoom interviewing 10 years ago, people weren't really comfortable recording calls, and there was a time when people weren't comfortable applying through an ATS. Things can change, but the technology's not there yet.

[00:24:54] Adriaan: There's this fascinating old news clip from when the mobile phone just came out. They're interviewing people on the street asking, would you buy a mobile phone? The clip is edited, so I don't know the sample size, but of the six people they interview, all of them say no. Why would I want a mobile phone, people calling me while I'm in transit, I have a phone at home. And that's a question I ask myself: is it just that I'm thinking I want to speak to another human because I'm not used to it, or is this technology that all of a sudden will just be there?

[00:25:38] John: If it catches up and there's more nuance to how it's done, it could become more normalized, but right now, no.

[00:25:45] Adriaan: You mentioned there's no bigger time to innovate, yet time is our most scarce resource in terms of time to think about it. How do you carve out time to think about the things you want to build, and how do you motivate your team to do so?

[00:26:04] John: One leads to the other. When you're using it the right way, it's giving you time back to actually think and strategize. So look at the tasks that were previously time-consuming: reporting and analytics is one, and with Claude you can do that much more efficiently, so you're literally saving time. Team management, how you organize yourself, planning your week, all of those can take a lot of time, and there's time saved there. So it's about looking at where AI gives you those efficiency gains, so you can spend the thinking time on what you can build. It's the same with my team. The more you put into it, the more time you'll have to do the things you actually want to be doing as a recruiter.

[00:27:04] Adriaan: John, this was such a pleasant conversation. I really enjoyed it, some real nuggets of wisdom. I'm going to look into the RAPID framework. It's a great example of the one thing that's so important as leaders: to communicate in a clear, concise and consistent way, with something that's easy to remember. Language is critical to getting anything done, for people to embrace whatever change or process you want them to follow.

[00:27:38] Adriaan: John, what's the best way for people to connect with you or follow you? LinkedIn?

[00:27:43] John: Yes, John Keenan on LinkedIn. Feel free to connect and follow. Thanks for your time.

[00:27:49] Adriaan: John, thanks so much for your time.