When news broke recently that Google’s star developer Noam Shazeer moved to OpenAI, my initial reaction was, “High-value people always jump ship, don’t they?” But as I thought about it, something seemed off.
If someone exists whose departure can shake an entire company,
doesn’t that imply the company already has a list of people they absolutely cannot afford to lose?
I became curious about who makes that list and how the company treats them once they’re on it. I wondered if it’s a position where you can finally work in comfort, or if it actually makes things harder.
I happened to be reading an issue report by the POSCO Research Institute (POSRI) based on interviews with Apple and Google officials, and I found an answer completely different from what I had imagined.
In many domestic companies, ‘key talent’ or ‘high-potential’ employees are often associated with receiving awards or going on overseas training.
However, this report shows that Apple and Google approach it differently from the start. They have shifted the focus of talent development from ’education’ to ‘actual work experience.’ Selection, development, and even removal all take place within the context of real-world tasks.
How many people are selected as key talent?
Apple puts only 0.8% of its total workforce on this list, while Google includes 1%. In terms of numbers, these odds are lower than the probability of getting promoted to an executive role in a large corporation. That’s how high the barrier to entry is.
Apple requires, as a baseline, being in the top 10% of performance evaluations for three consecutive years.
They target all roles at the manager level and above at headquarters (excluding overseas regions), narrowing the pool to key functions like hardware, software, AI, design, operations, and marketing. Out of about 186,000 employees, they cap the pool at around 1,500 people—exactly 0.8%.
And one year of stellar performance isn’t enough. The point is that they require a three-year track record. It means only those who have consistently performed well can even get near the list.
Google’s criteria are slightly different.
They require an ‘A’ grade in at least 3 out of the last 4 semi-annual evaluations, while also having experience in at least two different positions and two different business units (BUs). It means they value people who have reproduced results across different environments more than those who only excelled in one department. Out of about 190,000 employees, they also keep this at around 1% (approx. 2,000 people).
Interestingly, this designation doesn’t start with a unilateral notification from the company. The door opens only when the individual expresses interest, followed by a process of aligning the company’s expectations with the employee’s.
Not decided by one boss — Apple’s multi-stage collective validation
Here’s the part I found most interesting.
Apple doesn’t decide this ruthless 0.8% list based on the judgment of a single direct supervisor.
They hold a validation process called the 「Talent Planning Session」 12 times a year, spending two days per function each quarter. The ‘function’ here doesn’t refer to a BU, but to professional functional organizations that cut across products, such as Hardware Engineering, Software Engineering, AI/Machine Learning, Design, Operations, and Marketing.
Instead of dividing the organization by product like an iPhone team or Mac team, Apple operates on a structure where these functional organizations collaborate across multiple products, and key talent validation is carried out meticulously and regularly within these functional units.
On the first day, manager-level and senior manager-level peers who have worked with the candidate gather to evaluate performance and collaboration skills from a practical standpoint.
On the second day, higher-level leaders (Senior Manager and above) step in to re-evaluate leadership and growth potential from an organizational and business perspective.
As the stages progress, the pool of participants actually gets smaller. It’s designed so that those who saw the execution up close and those who viewed the potential strategically use different criteria. The candidate themselves is not involved in these discussions.
On top of this, they add a 「Challenge–Defend」 structure.
General participants take on the role of ‘Challenge,’ actively raising risks, limitations, and areas that need further verification, while the direct manager and senior manager who know the candidate best take on the ‘Defend’ role, specifically justifying the performance and potential against those points.
I thought about why they would create such a cumbersome structure, and the answer was simple. The moment key talent is determined by the praise of just one person, the list stops being a list of top performers and turns into a list of people with good connections. Planting opposing opinions at every stage of validation is like institutionalizing the organization’s habit of questioning its own biases.
Why does Google require approval from leaders in other departments?
Google’s approach is a bit more blatant, starting with the premise that “our department’s bias is real.”
Even if a candidate is recommended by their BU, they undergo a ‘Cross-validation’ where 5 Director-level leaders from other BUs participate to re-verify leadership suitability. This is a mechanism to minimize the bias of specific bosses or organizations. Requiring approval from leaders in other BUs means they treat “doing well in the current organization” and “doing well in another organization” as two distinct questions from the start.
Google uses five criteria known as ‘Googleyness’ for this assessment.
Learning Agility: Learning quickly in new situations,
Adaptability: Responding flexibly to environmental changes,
Collaboration: Exerting influence by working with other organizations,
Resilience: Continuing to deliver results even amidst failure,
Humility & Openness: Accepting feedback and acting in a team-centric way.
None of these five can be satisfied just by the fact that someone is “doing well right now.”
Since you need to look at how quickly they learn when moving teams, how they handle new people, and their patterns of bouncing back from failure, these are the true criteria for potential. It implies that the core of potential is not whether you are doing your current job well, but whether you can replicate that performance in an unfamiliar organization or role.
They separate performance and potential into different drawers and judge them separately, which I felt was quite brutal. It means even if you’re doing well, you are always shadowed by the suspicion, “Aren’t you only doing well because of this specific position?”
Being chosen as key talent is where the real work begins
If you’ve read this far, you might think, “So, doesn’t life get easier once you’re chosen?” In reality, it’s the exact opposite.
Apple throws high-difficulty strategic challenges that the company hasn’t been able to solve for years at their designated key talent. These aren’t case studies for training, but actual business problems, and they assign them to cross-functional teams, not individuals. They even exempt them from some routine work to focus on this task for about a year. This means they are observing the entire 12-month execution process rather than just short-term results. The task itself serves as a real-world validation stage to test complex stakeholder management, cross-organizational collaboration, strategic judgment, and execution.
Google is no different. They assign tasks that are clearly more difficult than existing roles or cross-functional projects that go beyond one’s area of expertise—again, actual business issues, not training simulations. The experience of clashing with new stakeholders and creating results in unfamiliar areas becomes the stage to test Learning Agility and Adaptability.
Both companies have moved beyond “training then application” to a model where “work performance itself is development.” In the past, the reward for key talent might have been a leadership training program or an overseas trip, but now, the company places actual business headaches on their desks. Instead of giving them a prize, they’ve given them another test.
5 pages of evaluation — How closely are key talents observed?
I noticed a specific detail in Google’s data.
For key talent, managers and next-level managers write detailed evaluation materials spanning about 2.5 to 5 pages. Considering that a general employee evaluation usually ends in a paragraph or two, key talent is scrutinized several times more deeply, covering performance, leadership, potential, and development needs. This evaluation doesn’t end with a grade; it serves as the foundational data for designing individual development programs and future roles.
Skill development is also handled not by external instructors, but by Senior Leaders within the organization who are recognized for their expertise, acting as instructors and coaches.
They create 4-12 week programs focused on practical skills like strategy formulation, persuasion and influence, and technical roadmap building. VPs and HR analyze individual evaluation results and weaknesses to recommend customized programs. It’s a method of prescribing different remedies for different people, which, if you look at it differently, means the company knows the weaknesses of each individual in great detail. Some might call this meticulous care, but I felt ‘precise observation’ was more accurate.
It’s not just a one-off lecture by an outside expert; it’s a structure where someone who has already proven themselves within the organization passes on their know-how through coaching. Since the instructor is someone who has actually experienced the company’s real problems, the content is more likely to be methods that actually work within that organization, rather than just theory. Ultimately, key talent is observed through pages of evaluations and receives a customized curriculum based on those very evaluations.
Key talent has two mentors, plus executive oversight
At this point, it feels like the pressure rests entirely on the individual, but in reality, the company invests significant resources.
Apple assigns two ‘Guiding Mentors’ to each key talent. The ‘Main Mentor’ is not the direct boss, but a leader from a different department, providing feedback free from the biases of the home organization. Mentors, direct leaders, and next-level leaders create a development plan together and link it to organizational strategy. Furthermore, top management—like the CEO, CFO, and Senior VP—personally review the progress of strategic tasks through a quarterly ‘Leadership Committee.’ Excellent tasks are often integrated into actual business projects, and as one rises to the candidate pool for senior leadership, the intensity of observation and verification from management also increases.
Google uses Succession Planning itself as a development process.
Leaders at the VP level or higher designate multiple successor candidates for each level below them, managing several people at once rather than a single candidate. The candidate pool progresses through steps: BU CEO candidate (Senior VP) → Alphabet Leadership candidate → Alphabet C-Level candidate. They analyze leadership trajectories by synthesizing multi-year ‘Stretch Assignment’ results, supervisor evaluations, external coaching feedback, and even ‘Google Geist’ scores (an internal survey measuring organizational culture and leadership perception). I found it a bit strange that so many people look at one person’s growth from so many angles at once. It feels more like surveillance than interest.
Up to 35% of key talent are dropped every year
And this was the most impressive part of the report: once you’re on the list, it doesn’t mean you stay there.
Apple conducts a ‘Pool in/out’ process every year, and if you don’t meet expectations, you are, in principle, removed from the list. However, they do leave room for a grace period if there are reasonable circumstances, like temporary personal issues.
Google is even more explicit. They state that they maintain a pool volatility of 25–35% annually, meaning the list changes for about 1 in 4 candidates every year.
Apple also informs individuals of their key talent status privately to minimize internal friction, and while they provide differentiated rewards through additional salary increases and bonuses, they also use ‘Counter Offers’ to retain talent who express an intention to leave by renegotiating terms.
Google does the same, offering salary increase rates 8–15 percentage points higher than general employees, and they will renegotiate terms up to three times if someone expresses an intent to resign. They are as aggressive in retaining talent as they are cold-blooded in removing them from the list if their performance wavers.
The title ‘key talent’ is not a permanent badge of honor, but more like a license that requires passing a renewal exam every time. This is exactly what both companies emphasize. They don’t view dropping out of the pool as a failure or a stigma, but rather leave the structure open for re-entry depending on roles, performance, and growth stages. If you are accustomed to an organizational culture where you rarely fall off a promotion list once you’re on it, this part will likely feel the most unfamiliar. I’ve laid out the structures of both companies in the table below.
So, does being labeled a ‘key talent’ make life easier?
Going back to the question I first posed, the answer is clearly “No.”
From the moment you’re chosen, you are repeatedly validated by multiple leaders, you’re tasked with problems the company couldn’t solve, management checks your progress quarterly, and you’re put back on the test bench every year.
This is not a stable perk, but a position that comes with constant observation and re-testing.
However, the rewards are also clear.
Salary increase rates, promotion paths, access to management, and the difficulty and impact of the problems the company gives you—everything moves at an accelerated pace.
Apple and Google confirmed one thing in common:
The way to keep good talent is not to make things easier, but to keep giving them difficult problems and have multiple people watch that process together.
Whether you can transfer this directly to your own company or team is another matter.
Apple and Google have hundreds of thousands of employees, so selecting 0.8–1% yields a sizable pool of 1,500–2,000 people, leaving enough room for multiple leaders to share the burden of validation.
If you are in a smaller organization, you don’t need to copy this structure exactly; applying just two principles would be enough.
One is to not let promotions or appointments be decided by the judgment of a single boss.
The other is to not give an easy position just because someone was picked as key talent, but instead, give them a problem the company hasn’t solved yet. That would be a much more honest validation than just agonizing over who to put on a list.
After reading this report, I only had one thought:
It’s time to abandon the expectation that the company will treat you comfortably the moment you’re called ‘key talent.’
Instead, it’s more realistic to see that the density of observation increases several times over from that moment on. However, the story changes if that observation is not based on one person’s whim or favoritism, but structured so that multiple leaders share the responsibility.
Whether you view that observation as surveillance or investment is what distinguishes the two perspectives.
And it’s also worth remembering that the right to receive that investment isn’t earned once, but must be proven every single year.
Ultimately, this position was closer to a permanent test bench. There was no moment of ‘getting comfortable.’ That was the answer this report gave me.