What APAC Measures (and Why CVs Miss It)

A CV is a record of the past. It tells you what someone has done, where they did it, and which institution put a stamp on it at the end. That is useful information, and we read every CV that lands in our inbox. But a CV does not tell you how someone thinks under pressure, how they hold a hard conversation across technical and non-technical audiences, or whether they will push back on a shipping decision when the room wants to ship anyway. In AI work, those are the qualities that decide whether a hire becomes a great one or a costly one. They are also the qualities that no résumé format was ever designed to capture.

This is why we built APAC. APAC is Radical's proprietary four-dimension assessment, and it is the spine of how we select every AI professional we accept into the community. The four letters stand for Adaptability, Personality, Awareness, and Connection. The framework is named, defined, and applied with discipline, because vague phrases like "cultural fit" are how bias hides in plain sight. APAC does the opposite. It names the qualities, measures them in a structured way, and lands the result in a real human conversation. Candidates take it for free at radicalnetwork.nl in ten to twenty minutes, and the report stays with them through their entire career at Radical.

I designed APAC after two decades in executive search. The recurring lesson across those years is simple. The hires that worked, almost without exception, were the ones where I had been able to read the human qualities and not only the credentials. The hires that did not work were the ones where I had let the CV do too much of the talking. APAC is the instrument I wish I had been able to put in every intake conversation from the start.

This piece is the long version of why. It walks through each of the four dimensions, what each one looks like when it is present, how it shows up when it is missing, and how we read it in the report. It also covers how the assessment actually works, why having no answer key matters, how the report follows the candidate, and why a structured framework is more ethical than gut feel or pure CV screening, especially under the EU AI Act. If you are a candidate, this is what we are reading when we read you. If you are a company, this is what you are buying when you ask us for a Radical.

A thoughtful candidate looking up, the human qualities a CV cannot show

Why CVs are a poor instrument for AI hiring

Software hiring has always been imperfect. AI hiring is worse, and the reason is timing. Technical surface moves faster in AI than in any subfield of engineering I have worked with. According to the Stack Overflow 2024 Developer Survey, 76 percent of professional developers now use or plan to use AI tools in their work, and 62 percent are using them today, up from 44 percent a year earlier. New model families ship every quarter. Frameworks rise and fall in months. A CV that lists fine-tuning techniques from 2024 is already a partial map by 2026. The list of "tools used" is, at best, a lagging indicator of who the person is.

The deeper problem is what AI work actually demands. The best AI professionals I have placed do not differ from the rest because they have shipped a particular framework. They differ because they make better calls. They know when a model is good enough to ship and when it is not. They know when to push back on a stakeholder who wants a shortcut. They know how to brief a board that does not read papers, and they know how to listen to a junior who notices something they missed. None of that is on a CV. All of it is in the room when the work actually happens.

The World Economic Forum's Future of Jobs Report 2025 puts analytical thinking, resilience and flexibility, leadership and social influence, creative thinking, and motivation and self-awareness at the top of the global core-skills ranking. Note what is not on that list. There is no specific stack, no specific model family, no specific certification. The skills that survive the rate of change are the human ones. They are the qualities APAC is built to read.

There is also a quieter problem with CV-led hiring. A CV is a record of access. It records which university opened its door, which employer extended an offer, which network passed the name along. Those things correlate with talent. They are not talent. A growing line of research, including Wilson and Caliskan's 2025 work on auditing AI-powered resume screening, shows that algorithmic CV screening can replicate and even amplify the demographic and stylistic patterns of past hiring, while wearing the appearance of objectivity. If a CV-screening model is unfair when it is automated, it is unfair when a human does the same screen by hand. APAC is the alternative.

The four dimensions, defined

APAC measures four things. Each one has a precise definition, an observable signature when it is present, an observable absence when it is missing, and a set of signals we read in the assessment. The table below is the compact version. The sections after it are the long version.

The four dimensions are not a personality test in the pop-psychology sense. They are a structured reading of qualities that the academic literature on personality and work has spent decades pinning down. The classic Barrick and Mount meta-analysis from 1991 and a long tail of follow-up research show that structured assessments of conscientiousness, openness, and related traits predict job performance across occupational groups in ways that interview impressions alone do not. APAC is not the Big Five. It is calibrated for AI work and read together with a human. But the principle behind it, that structured measurement of human qualities beats gut feel, is the same.

A is for Adaptability

Adaptability is the willingness and capacity to keep being effective when the ground moves under you. In AI work the ground moves constantly. A new model family ships, an evaluation method you trusted goes out of date, a regulator publishes guidance that changes what you can deploy in a quarter. The adaptable Radical does not freeze, does not double down on the old way, and does not perform certainty they no longer have. They re-read the situation, adjust, and keep moving.

What it looks like when it is present. A candidate describes a project where the constraints changed mid-flight, narrates the recalibration without blame, and treats the experience as data rather than as a wound. They show curiosity about new tools without chasing every novelty. They have a vocabulary for change that includes "I was wrong about X, and here is what I learned." They can sketch a contingency plan when asked, without sounding rehearsed.

How it fails when it is absent. The candidate tells the same story with a different label. Every project is the project where they shipped a transformer model. Every problem is the problem they have already solved. When asked what would change about their approach in light of a recent shift in the field, they retreat to authority. They invoke the framework they were trained on as if no other framework has appeared since. In real teams, this profile is the one that becomes a single point of failure during a stack migration.

Signals we read. We look for how a candidate narrates change, not how they declare comfort with it. Anyone can say "I love change." We listen for whether the change-story has a structure: what shifted, what they tried first, what failed, what they kept. We look for evidence that they have updated a strongly held position based on new information. We notice whether their CV shows movement between frameworks and contexts, or whether it is a single rail.

A candidate gazing thoughtfully, adaptability shows as the calm to keep moving when the ground moves

P is for Personality

Personality at Radical does not mean charisma. It means the analytical, growth-oriented temperament that takes ownership of outcomes and brings curiosity, creativity, and ambition to the work. It is the part of the profile that decides whether the candidate will, on a Wednesday afternoon when something is broken, lean in or lean out. AI work has a high enough density of ambiguous, painful Wednesday afternoons that this matters.

What it looks like when it is present. The candidate owns the failure stories as readily as the success ones. They describe a piece of work they were not asked to do, but did anyway because it was the right thing. They are visibly curious about adjacent problems, including ones outside their seat. They use first-person plural in success stories and first-person singular in failure stories, which is the inverse of the usual tell. They have a growth narrative, not just a credential narrative.

How it fails when it is absent. The candidate is technically strong but locates the cause of every difficulty outside themselves. Their CV is impressive, and the conversation around it is flat. They have not learned anything in the last twelve months that they can name without prompting. They describe their last team in terms of what the team failed to give them, not in terms of what the team produced together. In production, this profile is the one that needs to be carried by managers, and the one that quietly poisons review cycles.

Signals we read. We pay attention to how a candidate describes a piece of work they are proud of and a piece of work that did not go well. The Personality dimension is largely visible in the proportions. We look for the small acts of initiative, including the unglamorous ones: the rewriting of the runbook, the mentoring of the new joiner, the unblocking of someone else's project. We look for genuine analytical depth, including the willingness to sit with a hard question instead of resolving it too quickly. We listen for ambition that points at the work, not at the title.

A is for Awareness

Awareness is the dimension that distinguishes a good AI professional from a responsible one. It is the capacity to see context beyond the task: the team, the user, the regulator, the second-order consequences. In a field where a model release can become a regulatory and reputational event in days, this dimension is no longer optional. It is the qualification.

What it looks like when it is present. The candidate raises ethical and safety considerations before being asked, and raises them as engineering problems rather than as PR problems. They can describe a time they refused a shortcut, or pushed for an evaluation step that the rest of the team thought was overkill. They know which classes of decisions in their domain count as high-risk under the EU AI Act, and they have an opinion about why. They are aware of who their work affects beyond the immediate user.

How it fails when it is absent. The candidate is fluent about model performance and silent about model impact. Asked about a harm scenario, they default to "that is a policy problem" or "that is for legal." They have not thought about the difference between a system that works and a system that should work. They are competent but contextually thin. In production, this profile builds systems that pass internal review and fail public review, which is the worst possible failure mode for a company exposed to regulation.

Signals we read. We look for unprompted attention to context. We notice when a candidate raises a stakeholder, a user, or a regulator who is not in the room. We test for the language of consequences, not the language of compliance. The EU AI Act's Annex III classifies AI used for recruitment and worker management as high-risk, with binding obligations on transparency, human oversight, and bias evaluation. A candidate with Awareness can talk about that landscape in plain language, not in legalese.

A candidate looking out a window, awareness is the discipline of seeing past the immediate task

C is for Connection

Connection is the capacity to build meaningful relationships across teams, stakeholders, and the world beyond. It is rooted in values, not in extroversion. Some of the most connected Radicals I have placed are quiet people who write better than they speak, and that is fine. What matters is whether the relationship work happens. AI products are cross-functional by default. The engineer who cannot brief the board is half an engineer. The board that cannot brief the engineer is half a board. Connection closes that gap.

What it looks like when it is present. The candidate names the people who shaped their work, including the difficult ones. They describe a hard conversation they handled, not a hard conversation they avoided. They show care for the work of others, including peers they did not pick. They have a vocabulary for disagreement that is not combative. They can describe a relationship they invested in over time, professional or otherwise, and explain what changed in them because of it.

How it fails when it is absent. The candidate talks about themselves in a vacuum. Their stories are technically detailed and humanly thin. They cannot name a peer who challenged them productively in the last year. They have a low tolerance for the relational work of shipping software with other people, which they describe as "politics." In a team, this profile is the one that produces brilliant work no one else can use, and the one that burns out the people who try to integrate it.

Signals we read. We listen for whether the candidate's stories include other people as full participants, not as background. We notice the proportion of "I" to "we." We look for evidence that the candidate has invested in a relationship over time. We test for whether they can describe a stakeholder's point of view without flattening it. We watch how they speak about a previous manager they did not like.

How the assessment actually works

The mechanics are deliberately undramatic. A candidate goes to radicalnetwork.nl, creates an account, and takes the APAC test. It takes ten to twenty minutes. There is no proctoring software, no facial analysis, no eye-tracking. The questions are designed to surface how the candidate thinks, decides, and relates, not whether they can guess the "right" answer.

There is no answer key. That is not marketing copy, it is a design choice. The dimensions are read across the entire profile, not from individual right or wrong responses. There is no leaderboard. There is no comparative score that ranks candidates against each other in public. The output is a structured report on the four dimensions, written for the candidate to read first.

The report lands in a real human conversation with the Radical team. That conversation is not an interrogation. It is a reading together. We walk through the profile, ask where the candidate recognises themselves and where they do not, and we listen. The candidate is allowed to disagree with the report. Often the most useful conversations are the ones where they do.

This is what a structured assessment looks like when it is built for people rather than for sorting. It is closer to a coaching instrument than to a screening filter, and that is on purpose.

Two people in a real conversation across a desk, connection is where APAC lands

Why no answer key matters

People sometimes ask whether the test can be gamed. The honest answer is: not in any useful way. There is no answer key to game against. The dimensions are read across the entire profile and re-read against the human conversation that follows. Attempts to perform a persona surface quickly, because a persona is, almost by definition, less internally consistent than a real self.

A no-answer-key design also protects the candidate. A test with a hidden "right" answer rewards people who have learned the conventions of testing, which correlates with educational access. It punishes people who think honestly when they are nervous, which is most of us. APAC is built to read the person, not the test-taker. Honesty is rewarded. So is contradiction, when the candidate can sit with it.

Internally, the no-answer-key design also protects us. It means we cannot fall back on a number when a decision is hard. We have to look at the profile, listen to the candidate, and decide. That is the work. We chose a framework that keeps us doing it.

How the report follows the candidate

A common failure of assessment tools is that the result is used once, at the point of decision, and then forgotten. APAC is the opposite. The report is given to the candidate, discussed in the intake, and stored against the candidate record inside the Radical community. From that moment on, it follows them.

It informs the coaching they receive. A Radical strong on Adaptability and Awareness, lighter on Connection, will get different coaching suggestions than the reverse profile. It informs the mentors they are matched with, because the right mentor for someone who needs to stretch their Connection muscle is not the same as the right mentor for someone who needs to develop their Awareness. It informs every shortlist conversation about that Radical with a client, because the recruiter who pitches them knows precisely which dimensions are strongest. It informs the quarterly check-in, because we revisit the profile and update it when life or work has moved the dial.

For the candidate, that means the APAC report is identity-building. It is a structured mirror, written by people they have met, that they own. We see Radicals refer back to their own profile when they are making career decisions. That is a use case we did not engineer for, and it is one of the most rewarding outcomes of the framework.

Bias, gut feel, and the EU AI Act

Structured assessment is not perfect. No framework is. But it is a meaningful improvement on the two dominant alternatives in recruitment today: pure CV screening and pure gut feel.

CV screening, automated or manual, replicates the patterns of past hiring. If past hiring favoured graduates of a specific set of universities, the model favours them too. If past hiring rewarded a particular writing style, the model rewards it too. The Wilson and Caliskan study on AI-powered resume screening shows that even systems built to look neutral can carry intersectional bias that surfaces only under careful audit. Manual CV screening has the same problem, with the additional difficulty that the human reviewer cannot be audited at the same scale.

Gut feel, the other dominant alternative, is worse in a different way. The classic finding in the personality and performance literature, including Barrick and Mount's 1991 meta-analysis, is that structured assessment of human qualities consistently outperforms unstructured interview impressions for predicting job performance. Gut feel is also where bias hides most comfortably. When the criterion is "I just had a feeling," the feeling is correlated with the assessor's history, in-group affinity, and unspoken assumptions. Naming the dimensions, defining them, and applying them consistently is the discipline that pulls those assumptions into the open.

This is also where the EU AI Act matters. The Act's Annex III classifies AI used for recruitment, candidate evaluation, and worker management as a high-risk use case. High-risk systems carry binding obligations: transparency, human oversight, bias evaluation, documentation of the design choices. Compliance is the floor. The principle is higher. Responsible AI is the principle that there is always a person accountable for the decision, that the person can explain it, and that the person can be challenged. APAC is built to make all three possible.

What this means for the candidate and the company

For the candidate, this means a free, identity-building assessment that takes ten to twenty minutes, lands in a real conversation, and produces a report that they own. It means being read on who they are, not only on what they have shipped. It means coaching and mentoring that is matched to the profile, not stamped on top of it. It means stopping being a commodity the moment they are accepted.

For the company, this means judgment-based hires. Every Radical on a shortlist has cleared four gates including APAC. The company is hiring for the qualities that survive the rate of change in AI: adaptability, personality, awareness, connection. They are not hiring on the qualities that the CV happens to surface and the LinkedIn algorithm happens to rank. They are hiring on the qualities the Harvard Business School research on skills-based hiring has been showing for years are the qualities that compound across roles.

The system works because both sides of the table are held to the same standard. The candidate is read seriously. The company is briefed seriously. The recruiter who runs the intake stays accountable for the rationale that goes into the introduction. AI does the matching scaffolding. A human makes the call.

The instrument I wish I had had earlier

After more than twenty years in executive search, I am clear about one thing. The hires that worked were the ones where I could read the human qualities. The hires that did not work were the ones where I let the CV carry too much weight. APAC is the instrument I wish I had been able to put in every intake from the beginning. It is the framework that names what good selection has always quietly relied on, and it is the discipline that keeps us honest when the temptation is to move fast and screen by keyword.

If you are a candidate reading this, take the test. It is at radicalnetwork.nl, it is free, it takes ten to twenty minutes, and at the end of it you will be in a real conversation with us about who you are. If you are a company reading this, that is the standard you are buying. Every Radical we put on your shortlist has been read through APAC, talked through with a person on our team, and accepted into a community of people we would hire ourselves.

The human factor is the scarce resource in the AI economy. Models are abundant. Adaptability, personality, awareness, and connection are not. APAC is how we measure them, name them, and stand behind them.

Human on the loop. European by choice. Trustworthy by design.