One of the privileges of being your SourceCon Editor is that I get to connect with luminaries in the space, those we affectionately call ‘The OGs.” Josef Kadlec is certainly worthy of the designation. His book, “People as Merchandise,” was autographed and given to me on September 24, 2014 and still resides on my bookshelf as one of my favorite sourcing reads. His latest book will soon sit beside it, and I suspect, will be heavily highlighted and read through multiple times as well.
Josef was kind enough to sit down with me for an interview to answer questions about his book, his career and the future of sourcing. Enjoy.
Josef, it takes a lot of passion to write a 550 page book. What fueled this initiative?
I undertook this project for the second time, so I was aware of the challenges ahead. My goal was to surpass the quality of my previous work, ‘PEOPLE AS MERCHANDISE’ from 2013, not only in content quality but also in the visual presentation of the book. I aimed to create an even more appealing book. This wasn’t just because it is 200 pages longer and in a larger, hardcover format that would look impressive on a bookshelf.
The main impetus, however, came from the rise of generative AI. I found myself frequently invited to companies and public training sessions to discuss AI in recruitment and HR marketing, which motivated me to create a clear guide for those without a technical background. The initial spark probably came from our own employees. As a former software engineer, many people expected this of me, so it felt like a duty of sorts.
What key challenges do you believe HR professionals face today that make AI integration a necessity?
I think that from the AI excitement in 2023 we are going to the stage of AI deployment in 2024. So regardless of if an individual HR professional is going to use some AI tools or not, HR professionals are going to face AI features more often when using their regular tech stack. It can be for instance a feature in your ATS which creates a description of a candidate for a hiring manager with a click of a button (we’ve recently developed this feature for Datacruit ATS). It can be a video call solution which makes candidate interview notes automatically. Or it can be LinkedIn where we can already use AI features on our profiles or in LinkedIn Recruiter premium accounts. And maybe you don’t even know that these features are fueled by the elements of AI.
So we are going to adopt new habits of doing particular parts of the recruitment (and HR) process such as intake meeting, sourcing, job ads creation, interviewing, etc. regardless of if you use some premium ATS suites or you run it yourself using ChatGPT, Claude, Bard, Midjourney, DALL-E, etc.
Could you share insights on how AI tools can be effectively utilized to improve the overall candidate journey in recruitment?
I believe that a practical strategy, which we’ve also implemented in our organization, is to segment and simplify complex tasks. Some solutions attempt to replace an entire role, such as a recruiter, talent sourcer, or HR professional, with AI. However, these roles are too complex to be substituted without breaking them down into smaller, more manageable parts. Moreover, the complexity increases with various scenarios – for instance, in our organization, the process and tech stack differ between permanent agency recruiters and our RPO recruiters.
Nevertheless, there are numerous applications of AI in HR processes that can enhance the candidate journey for candidates, recruiters, or both at times. Take interview notes, for example. Instead of jotting down notes during the interview on paper or into the ATS, an HR professional can concentrate fully on the candidate. This enables better reading of the candidate, asking more insightful questions, and perhaps even showing more empathy. This is undoubtedly beneficial for both parties.
Another scenario is creating a digital replica of your HR specialists (let’s say in an ATS) for various tasks, ranging from initial candidate outreach, sending informative video messages throughout the recruitment process to boost engagement, all the way to onboarding and Learning & Development activities.
Not to mention HR marketing activities, such as tweaking job descriptions, job ads, and posts, including text, images, and videos. For instance, the demand for copywriters is already declining, as are their wages.
How does your guide address the ethical considerations surrounding AI implementation in HR, particularly in relation to candidate experience?
Firstly, as previously described, the implementation of AI in HR processes is not about allowing AI to make hiring decisions. AI can indeed assist in the process. For instance, you could take a LinkedIn profile and a job description, apply a Large Language Model (LLM), and prompt it with a request like, “Give me 3 reasons why this person is or is not a good fit for this role.” This approach seeks insights rather than a definitive yes or no answer. It’s like seeking a colleague’s opinion, but it’s risky to leave the entire decision to current narrow AI, especially considering how language models like GPT, PaLM, Gemini, Claude, Jasper, etc., function.
Secondly, there is a growing preference, possibly supported by studies, for people to communicate with a computer rather than a person. Typically, when seeking information, we first turn to Google, then perhaps a chatbot, and only as a last resort do we make a phone call. For instance, the staffing service Shepherd faced an issue where their sales staff was overwhelmed and couldn’t meet demand in a timely manner. Market-available voice chatbots, like SameDay (gosameday.com), can automate a salesperson’s role in such scenarios. People are given the choice to get an immediate response from a robot or wait (and potentially pay more) for a real assistant. Surprisingly, many prefer the robotic option.
In your opinion, what are the most exciting advancements in AI for HR that HR professionals should be aware of?
That’s a great question! In my talent sourcing and digital recruitment trainings, I’ve demonstrated emerging technologies that were, at the time, practically unusable prototypes in real-life scenarios. Many of these technologies have since advanced to the point where they are now readily usable. For instance, any HR professional, even those with little technical background, can be excited about tools like the HeyGen suite, where you can create a digital clone, including your own voice, capable of speaking any language. If I were to produce e-learning or onboarding videos, for example, I wouldn’t use anything else. No need for a studio, cameras, and it’s possible to distribute in any global language. Just three months ago, this wasn’t so easily achievable or affordable.
Another AI technology worth noting is the one that can eliminate accents, such as a French or Indian accent in English. This advancement in AI has the potential to bring nations closer together by enhancing mutual understanding, for example, through automated subtitles or dubbing in platforms like Teams and Zoom.
In general, any application of machine learning in recruitment is thrilling, be it cloning the style of recruitment emails or the style of images on career websites.
However, there are areas where the excitement isn’t as high as one might expect. One example is candidate searching and matching. There are significant barriers here, such as legal access to candidate profile data (e.g., LinkedIn profiles) or the quality of data in your ATS.
What impact do you foresee AI having on HR analytics, and how can HR professionals stay ahead of the curve in this evolving landscape?
HR analytics, as well as data analytics in general, is an area where the capabilities of current generative AI are greatly assisting those who know how to use it. The challenge in data analytics often lies in dealing with issues like incoherent data, data cleaning, and handling data sources in various formats and structures. For instance, consider an Excel table where names are listed in one column, but the entries vary—some have just a first name, others just a last name, or no name at all. Tools like the Advanced Data Analysis mode in ChatGPT can clean this data with a single prompt. This mode of ChatGPT often suggests solutions autonomously, functioning somewhat like an auto GPT.
Therefore, if you’re looking to explore relationships between variables, such as overtime and employee retention, and you have sufficient data, it’s possible to resolve this using Large Language Models (LLMs) like ChatGPT. However, it’s crucial to be mindful of data protection laws and avoid uploading private data into public language models.
How does your book address concerns about potential job displacement and the role of HR professionals in managing the human side of AI implementation?
Consider movies like ‘2001: A Space Odyssey’ by Stanley Kubrick, based on the novel by Arthur C. Clarke, which serves as an example of how people in the 1960s envisioned the world in 2001. It’s quite different from reality, I would say. Therefore, while it’s important to look forward to the future, I believe focusing on the mid-term future is more practical. Instead of assuming that AI will replace us (which is clearly not the case), we should aim to enhance our human skills with AI.
I often say that if a robot can do something, let the robot do it and focus on tasks that are still challenging for it. Reading my book ‘HR ROBO SAPIENS’ or any other source on this topic might be a good starting point to stay ahead in the HR field.
How does your guide cater to HR professionals looking to enhance their skills and knowledge in AI, especially those who may not have a technical background?
We are living in a remarkable era, one where, to a certain extent, software engineers might worry about their job security. Programming is likely to evolve significantly in the next decade, possibly becoming more akin to natural language. The key takeaway is that many projects are now being implemented by non-technical people because it has become feasible. Think about the evolution of website creation from a decade ago to today. Now, add tools like Make or Zapier, which allow for app integration without any coding knowledge. And then, consider the capabilities of today’s General AI, which can efficiently code everything from simple Excel formulas to Python scripts.
My book is structured to be accessible for anyone, enabling them to make AI their digital assistant in tasks related to talent acquisition and HR marketing. You don’t need an expensive platform or complex implementations. That would be a topic for a different book.
What advice would you give to HR professionals seeking to convince organizational leaders about the value and necessity of AI integration in HR?
Evolve or die. This concept is similar to the adoption of LinkedIn over a decade ago – some companies embraced it early, some later, and some are still adopting it now. The key difference between LinkedIn and AI is their impact areas. While LinkedIn significantly influenced talent sourcing and recruitment, AI spans across all HR activities – talent sourcing, HR marketing, interviewing, onboarding, Learning & Development (L&D), payroll, HR analytics, and more. Being an early adopter of AI will significantly strengthen your competitive advantage.
Can you discuss the potential cost savings for organizations that implement AI in HR processes, and how your book guides HR professionals in maximizing these benefits?
I believe this boils down to the cost considerations inherent in any digital technology implementation. For example, we regularly contemplated integrating CV parsing technology. However, given the costs and the level of glitches it still presented for an organization of our size, we ultimately decided against implementing it.
A similar situation applies to more sophisticated AI-based matching systems. In these cases, you have to account not only for the implementation cost but also for every request made to ChatGPT. Fortunately, this pricing has the potential to decrease. There are methods to condense the size of the prompts, such as compressing a CV, LinkedIn profile, or job description into a minimal vector. This approach reduces the number of tokens used and, consequently, lowers the cost per prompt.
My book, ‘HR ROBO SAPIENS,’ primarily addresses these issues from the perspective of individual HR professionals (though it’s also applicable to companies). In these cases, the costs of an AI tech stack are much easier to manage. Moreover, many of these tools are available free of charge.
How does your book approach the delicate balance between automation and maintaining a personalized touch in HR, particularly during the interviewing and candidate experience stages?
This challenge has always existed, even before the advent of AI. Tools like LinkedHelper and Robot.works are used to automate LinkedIn messages, while MixMax can automate emails and LinkedIn InMails. Generative AI technologies, such as GPT, Bard, or specific tools like Compose AI, Jasper, and Monica, can create highly personalized candidate messages. However, the challenge lies in maintaining the integrity of the text. The problem is that when using a candidate’s LinkedIn profile as input, the information can vary greatly – some profiles are very informative, while others are quite sparse. Consequently, some messages turn out well, but others not as much, or they might even include inaccurate, ‘hallucinated’ information.
At our organization, GoodCall, we don’t use AI for sending bulk messages. Instead, we use templated messages with placeholders filled with information directly from our ATS. However, when recruiters are sending messages in a semi-automatic mode (even if these messages are prepared in advance in a spreadsheet with GPT – using the Google apps extension GPT for Sheets and Docs), they utilize AI to generate one personalized paragraph based on the candidate’s social media profile. This paragraph is then inserted into the message template, ensuring the output is free of glitches and includes a level of personalization that would normally require manual effort by a talent sourcer.
In your experience, what are some common misconceptions HR professionals might have about implementing AI, and how does your book address these misconceptions?
I have trained several hundred HR professionals in my AI training course for HR pros, and I’ve noticed that the main misconceptions stem from understanding what today’s generative AI (GenAI) is and what ‘intelligence’ means in this context. Once you comprehend that it operates on the basis of a language model (a probabilistic algorithm that more closely mimics human intelligence than actually possessing it), you’ll be better equipped to predict whether the prompt you’re typing makes sense or not.
For example, if you ask ChatGPT or Bard to create a list of IT security companies in France, you might get a relevant list (using a Large Language Model with or without web browsing capability). However, if you request a list of IT security companies sorted by the number of employees on LinkedIn, you shouldn’t expect a reliable output, as LinkedIn data cannot be accessed in that way, even with web browsing capabilities. It’s possible that someone might have compiled such a list in a blog post, so you could still try and get results, but you see my point.
How can HR professionals ensure a smooth transition to becoming AI-powered HR pros, and what ongoing support and resources does your book recommend for staying updated on AI advancements in HR?
Many people argue that it’s premature to engage with AI currently, given its rapid evolution. While I agree that AI is evolving swiftly, the fundamental principles were established some time ago. Now, the various verticals, such as text-to-text models, text-to-image, and text-to-video diffusion models, are merely undergoing refinement. Therefore, understanding the basics is certainly beneficial. To stay informed, I recommend following AI-focused newsletters, LinkedIn influencers in AI, or joining AI groups.
The source doesn’t matter much as key information tends to proliferate across platforms. For deeper insights into a specific category, subscribing to newsletters from specific AI tools and platforms is an effective way to keep abreast of the latest developments – this is a strategy I often employ.
Thank you for your time Josef. I wish you continued success!
…
JOSÉ KADLEC, a former software engineer, has been at the forefront of revolutionizing stagnant industries like recruitment through analytical strategies and technological advancements since 2006.
He made waves as one of the pioneering LinkedIn recruiters, detailing his innovative talent sourcing techniques in his best-selling book, People as Merchandise. Even today, José trains and advises talent acquisition teams at both global and local giants like Oracle, Cisco, Siemens, Accenture, Zalando, DHL, Barclays, and Microsoft.
José co-founded the 140-employee recruitment titan, GoodCall, along with ATS provider Datacruit and the recruitment certification authority Recruitment Academy. With a decade-long presence in the market, these enterprises have received accolades, including recognition in Deloitte’s TOP50.
“I began programming neural networks during my studies at the Faculty of Nuclear Sciences and Physical Engineering. I’m fortunate to be living in an era where we can all integrate AI into our daily lives! This is sooo powerful that I wanna die: )”