How do you recruit people to work with technology that evolves faster than job descriptions? In an era where algorithms can sift through thousands of applications in seconds, and candidates mass-tailor their CVs using AI, the traditional approach to talent acquisition is becoming a thing of the past. To talk about how to build agile, globally connected teams of the future in practice, we speak with Joanna Trzaska-Lorenz – Group Talent Acquisition & EB Director at Comarch Group. Joanna reveals how to flawlessly identify "tech empathy" today, where the boundaries of HR automation lie, and which competencies of tomorrow we must look for in candidates right now.
Joanna, you are facing a major challenge at Comarch – centralizing TA processes on a global scale. In an era of global transformation and the proliferation of AI tools, how do you maintain the right balance between process standardization and a unique Candidate Experience across such culturally diverse teams, so that in the pursuit of efficiency, you don't lose sight of the crucial "human touch"?
I have a simple rule: we standardize the backbone, not just the conversation. Globally, we standardize what the candidate doesn't see — process stages, SLAs (Service Level Agreements), data, tools, decision criteria, and the interview loop. This gives us comparability, quality, and speed. However, the human touch layer — language, style, cultural context, touchpoints with the recruitment team, and the interview loop itself — must reflect the specific character of the local environment, the role, the challenges, and the business context the candidate is stepping into. Regardless of location, a candidate goes through the same process, but not necessarily the exact same conversation.
The second rule: we automate administration, never the relationship. AI can schedule a meeting, send a feedback reminder, or prepare a brief for the recruiter. It cannot build trust for us or make a decision about a person.
And the third, most important point: we want to measure efficiency not only by time-to-hire, but also by what candidates say about the process. If Candidate Experience drops in the pursuit of speed, that isn't efficiency — it's a debt that the market will eventually present to us to settle. That is why at Comarch I am also focusing on transforming candidate experience. In IT, news travels faster than job offers.
This lightning-fast verification works both ways. Today, nearly every candidate lists AI proficiency on their CV. How do you distinguish between a mere "user" — who treats technology as a mindless crutch for generating ready-made templates — and a true "AI partner" who can engage in a deep, critical dialogue with it during an interview?
First of all, I don't ask if someone uses AI — today, that’s like asking if they use the internet. I ask how they think when they use it. I'll say: "Tell me about the last time AI gave you a wrong answer. How did you spot it?". A mere user doesn't have such a story because they don't verify anything. A true partner has several and tells them in vivid detail.
I also like to reverse the exercise: instead of asking the candidate to generate something with AI, I give them AI-generated output and ask for a review. A user praises it because the text "looks professional." A partner disassembles it, asks for sources, and spots where the model started hallucinating. AI is like a brilliant intern: fast, hardworking, sometimes overly confident, eager to prove itself, and occasionally lying through its teeth. I look for people who know this and can still build a great team alongside such an intern.
You talk about team building and maturity in working with technology, and ethics is key to all of this. Your background with the European Commission and the Institute for Human Rights and Business gives you a valuable perspective. Automated tools can screen thousands of applications in seconds — in your opinion, where is the safe, ethical red line when delegating decisions to algorithms so that AI doesn't exclude non-traditional talent from hiring processes?
My red line is sharp: algorithms may organize, but humans must decide. AI can prioritize, cluster, suggest, and search — but an automated rejection of a candidate without a human eye is a line we do not cross. Not only because the EU AI Act rightly classifies recruitment as a high-risk area, but primarily because models learn from our past — which includes our historical mistakes and biases.
This sensitivity didn't come to me from pure theory. Starting in 2012, as an expert and advisor in roundtable discussions for the European Commission's "Sector Guidance Project," I helped draft guidelines on business respect for human rights across three sectors: employment and recruitment agencies, ICT, and oil & gas. The guidelines were based on the UN Guiding Principles on Business and Human Rights, and the project itself was one of the priority actions announced by the Commission in its 2011 Corporate Social Responsibility communication. The fact that the recruitment sector was included alongside oil and gas says everything: recruitment is not just a standard operational process. It is the moment a company genuinely impacts someone's opportunities, dignity, and access to work. That is why I view automation through the lens of human responsibility, not just efficiency — due diligence applies to us equally when a decision is suggested by an algorithm.
This responsibility and process-oriented approach are rooted in concrete business realities. You’ve worked in organizations with very diverse cultures, such as Allegro and IBM. What best practices in data-driven recruitment from your previous companies can be adapted most quickly to the ongoing transformation at Comarch to build an agile organization of the future?
Three things, in this order. First: treating the recruitment funnel like a product. We measure conversions between every stage, not just the final outcome — because only then can you see where you're really losing people and time. Second: capacity planning. Recruitment doesn't start with an open requisition; it starts with the business plan — if TA learns about a hiring need on the day it arises, we're already a quarter late. Third: quality of hire over time-to-hire. A bad hire brought in quickly costs more than the right hire brought in slowly.
From IBM, I brought scale and process discipline — respect for the fact that in a global organization, repeatability is a asset, not a bore. From Allegro, I brought speed and the courage to experiment — a culture where a solution is piloted with one team before being rolled out everywhere. Comarch, in its transformation, needs precisely this hybrid. And there is one caveat that has stuck with me for years: data is a compass, not an autopilot. A dashboard has never hired anyone.
Right in the context of this hands-on work and early career steps, we are facing a recruitment "junior paradox." AI handles basic coding tasks brilliantly, so traditional beginner tests no longer make sense. What should we expect today from people starting their careers in IT, and how can we evaluate their potential when technology does the foundational work for them?
We have to be honest: we've stopped hiring juniors for what they know today. We hire them for the speed at which they learn and with a future leaders trajectory in mind. Since AI does the foundational work, foundational work has ceased to be a valid test, as it was never a particularly accurate predictor anyway.
How do we test potential? We assign a task with AI explicitly allowed, and we evaluate the process rather than the end result: what questions did the candidate ask, what did they challenge, how did they verify the output, which decisions did they make on their own, and why. A junior who can say, "The model was wrong here, so I took a different approach," demonstrates exactly what I'm looking for. Currently, in our recruitment process, we have implemented these types of assessments within the Codility platform, allowing us to perform targeted evaluations of the thought process (and not just for junior talent).
Beyond that, I look for three things no technology can replicate: curiosity, a sense of ownership over the outcome, and the ability to say, "I don't know, but I'll find out." That last sentence is the best recruitment signal I know.
Taking this learning speed a step further — in an era where technology evolves week by week, unlearning (the ability to quickly unlearn old habits) becomes critical. How can you effectively assess a candidate's cognitive flexibility and readiness for continuous change during an interview?
I don't ask candidates what they've learned; I ask what they've abandoned. "What were you doing a year ago that you no longer do today — and what convinced you to stop?" This question cleanly splits candidates into two groups. Some have a concrete story: a discarded framework, a changed approach to estimation, or an internal process rewritten from scratch. Others speak in generalities about "continuous development" — and that answer tells me the most, even if it's not what the candidate intended to convey.
The second question in this series: "When was the last time you changed your mind based on data or someone else's argument?" Cognitive flexibility is also visible during the conversation itself; you just need to drop a counterexample to the candidate's premise and observe what happens. Some get defensive, while others get curious. Let's hire the latter. In a world where tools change quarterly, being attached to being right is more expensive than lacking a specific technology on your CV.
This leads us to a broader question about the future of the market. Looking ahead 5 to 10 years, how will the philosophy of career building itself change? In a world where AI takes over routine tasks, will recruitment rely solely on culture fit and human superpowers, or will technology create entirely new specializations?
Culture fit makes sense — provided we define it properly. I don't see it as asking, "Is this candidate like us?" because that's a direct path to cloning the people we already have, and cognitive diversity is the last thing organizations should sacrifice in the AI era. I understand culture fit as evaluating the right attitudes and behaviors: adaptability, openness to change, cross-functionality, and the quality of communication and collaboration. These are things that can be evaluated behaviorally — based on specific situations, not gut feeling — and they truly dictate whether someone will thrive in the organization of the future.
Otherwise, there will be a hybrid: technology will relieve humans of repetitive tasks, while humans will remain where risk, accountability, and relationships lie — because someone must be accountable for a decision, and trust cannot be automated.
Will jobs emerge that we can't even name today? Absolutely. Ten years ago, no one was planning a career as a prompt engineer — and that job has already emerged and begun to evolve. The deeper change, however, will be in career philosophy: it will cease to be a ladder and become a portfolio of competencies that we rebuild every few years.
Among these new competencies in the portfolio, the ability to give precise instructions to machines is most frequently cited today. In your opinion, is prompting a strictly technical skill, or rather a new soft skill — linked to logic and communicative empathy?
It's a soft skill in a technical wrapper. A good prompt is nothing more than good delegation: context, goal, constraints, success criteria, and an example of the expected result. I've noticed a simple correlation — people who delegate well to humans prompt machines well. Conversely, someone who throws a vague "figure this out somehow" to their team talks to a model the exact same way, and then complains about the results from both.
There's also something unforgiving about it: AI exposes messy thinking faster than any coworker because it won't guess or cover up your lack of precision with its own experience. You get exactly what you asked for — which can be a painful lesson in what you actually asked for. That's why I treat prompting as a workout in precise thinking and communicative empathy simultaneously: you must be able to view your own instruction through the eyes of a recipient who knows nothing beyond what you wrote. It's a skill of the future not because it involves AI, but because it involves thinking.
This ruthless technological precision leads many candidates to attempt to mask their shortfalls and prove their "AI-readiness" at all costs. What do you consider the biggest red flag and anti-pattern in their attitude, applications, or presentation of experience today?
AI-washing. A CV written entirely by a model — linguistically perfect and completely hollow. Ten buzzwords about "leveraging generative AI," but during the interview, not a single concrete example: which tool, for what purpose, what went wrong, and how they handled it. The paradox is that the people who boast the loudest about being ready for the AI era often demonstrate the opposite — that they've surrendered the steering wheel to it.
However, what worries me most is a phenomenon I call the outsourcing of thought: a candidate in an online interview who answers brilliantly, but with a four-second delay while their eyes scan a second screen. It doesn't bother me that they used AI — it bothers me that they don't trust their own head enough to talk without a teleprompter. Because at work, moments will come when there is no teleprompter: an customer escalation, an outage, or a tough team conversation. I want to know who I hired in that moment — the human or their chat interface.
You talk about trusting your own mind, but even with the best tools, a human ultimately makes the final call, and humans are fallible. In your career, have you ever made a mistake in evaluating a candidate — unfairly writing someone off or trusting someone who turned out to be a mis-hire — and what did that situation teach you?
Of course I've made mistakes — any recruiter who claims otherwise simply hasn't checked what happened to the people they rejected. My most painful lesson: early in my career, I passed on a candidate who was nervous, soft-spoken, and lacked charisma during the interview. Yet, he spoke with real substance. He lacked — as I called it back then — that "certain something." He didn't convince me, so I parked my doubts and compartmentalized him. A few years later, I watched him build a stellar team at a competitor. An incredible expert, leader, mentor, and practitioner.
It taught me that we routinely confuse confidence with competence, and eloquence with the quality of thought. They are two different parameters — and weakly correlated ones at that. There was also a mistake in the opposite direction: I trusted someone who was mesmerizing during the interview, but at work proved to be a master of talking about work rather than actually executing it.
Since then, I stick to a rule: an interview measures the ability to conduct an interview — everything else must be verified differently. That's why today I measure process, not claims. Situational tasks rooted in the actual realities the candidate will work in — not abstract brainteasers. Retrospectives: what went well, what didn't, and why. Drawing conclusions: "What would you do differently today — and what convinced you to change your approach?" That question tells me more about a candidate than an entire CV, because it shows whether someone learns from their experience or merely collects it. Human nature? We are predictably fallible. A good hiring process doesn't eliminate human fallibility — it acknowledges it and designs safeguards around it.
Over your 18+ years in HR, you've likely seen almost everything. Do you have a story in mind that perfectly illustrates this design of safeguards and the collision of human and technology — a situation where the human factor or an unexpected use of AI completely turned a standard process upside down?
It was a recruitment drive for a Senior HR Manager role. At the case study stage, the candidate was presenting their concept. The presentation was flawless — structure, aesthetics, storytelling; every slide looked straight out of a textbook. Except the longer I listened, the more clearly I felt something was missing: not a single personal opinion, not one uncomfortable observation, not a single trace of critical thinking. Everything was smooth, correct, and strangely impersonal. A classic profile of work done entirely by AI — from the first slide to the last.
So I did one thing: I closed the presentation. And I asked: "Now, please challenge everything you just presented. Where does this concept fall apart? What would you attack if you were sitting on my side of the table?" And the truth came out. The person who, a minute earlier, had been fluently "presenting" the strategy could neither defend it nor challenge it — because they had never actually thought it through. They didn't know their own assumptions, so they couldn't see their flaws. For me, that is a textbook story of the AI era in recruitment: technology can generate the perfect answer, but it cannot generate convictions.
That's a great takeaway on verifying convictions. If you were to reveal your one favorite "keyhole question" to test whether a candidate can critically and safely collaborate with new technologies — what would it sound like, and what are you looking for in the answer?
My keyhole question always stems from the candidate's real-world experience, not theory. It sounds something like this: "Let's go back to a specific project where you worked with AI. Tell me about it like a retrospective: what worked and why, where did you have to make a decision against what the tool suggested, what was your biggest breakthrough moment — a true gamechanger — and where did you experience a painful failure?"
It's not just one question; it's an invitation to tell a story — and that's precisely why it works. Theory can be learned in an evening; a retrospective of your own work cannot be fake-generated.
What am I looking and listening for in the response? Specifics regarding execution: which tool, for what task, with what outcome. Critical thinking: the moment the candidate said "no" to the model and can justify why. Honesty about failures: people whose repertoire consists exclusively of successes are either not experimenting or not drawing conclusions — I'm not sure which is worse. And genuine enthusiasm: because mature collaboration with technology isn't cold detachment; it's the ability to hold two thoughts in one's head at the same time — "this is a gamechanger" and "this can fail me painfully." A candidate who can talk about both in detail — that person has truly worked with AI, not just read about it.
Finally, let's step away from technology, strategy, and juggling so many things at once. Managing global teams and leading digital transformation at Comarch sounds like a recipe for digital sensory overload. In all this daily information noise, how do you clear your head and gain distance after an intense day in the tech world?
I have my "runaway place," a hidden cabin in the woods. It's my sanctuary — tucked away in the Polish Jurassic Highland (Jura Krakowsko-Częstochowska), and that's part of the ritual: a place that exists on no map of my calendars, notifications, or meetings. That's where I go for a regular tech detox — the phone stays in a drawer, and the only "feed" I follow is the forest outside the window. And that feed, unlike all the others, genuinely calms me down.
The more I work with technology, the more I value a place that doesn't update, doesn't sync, and doesn't need a charger. Maybe that's what this entire AI era is really about — knowing what to delegate to machines, and what to keep exclusively for ourselves. I'm keeping the forest for myself.