The Creativity Companies Want Most Often Comes From the People They Most Want to Cut

If a company genuinely wants breakthrough innovation, the first contradiction it faces is not in the market but in its own management system. It hires with one competency model, evaluates with one performance standard, and filters with one promotion process. That system is good at identifying "qualified" people. It is not necessarily good at identifying "irreplaceable" ones.

Alex Karp put the contradiction bluntly in an interview: the creativity a company most wants may come from exactly the people it most wants to eliminate. It sounds like a management platitude, but his explanation is not tame. The problem is not that managers are insufficiently enlightened. It is that there is a structural conflict between how organizations select talent and what they expect that talent to produce.

1. A "No Weaknesses" Filter Cannot Produce "Non-Standard" Results

Traditional companies favor the "hexagonal warrior": someone who communicates, collaborates, executes, and manages well enough, with no obvious weak spot. The preference makes sense. Such people are predictable, easy to work with, and easy to slot into existing processes and hierarchies.

But Karp argues that the people who create genuinely breakthrough value tend to have extremely uneven abilities. They may be poor communicators, dislike being managed, and struggle to adapt to standard processes. In one specific domain, however, they possess a kind of creativity others cannot understand. Put it in numbers: most abilities sit at 60, while one reaches 99.

The problem is that hiring forms, performance reviews, and promotion criteria usually reward the person with "nothing below 60," not the person with "one 99." That produces an absurd situation. Companies select talent in a standardized way and then expect that talent to produce non-standard results. They hire, evaluate, and promote with the same ruler. What remains is the most manageable people; what gets cut may be the least standard person who is most likely to break through. Management efficiency rises, while the organization's ceiling for innovation quietly comes down.

This is not an argument that specialists always beat generalists, or that every difficult person deserves to stay. Karp's claim is narrower and sharper: when a company measures everyone with a single standard, it has already decided in advance how much innovation it is capable of.

2. Creativity's Real Adversary Is Authority

Why do organizations struggle to accept genuinely creative people? Karp's answer is that real creation often means challenge.

Suppose a young employee is right, and his judgment contradicts a decision his manager made in the past. The issue stops being a question of ability and becomes a question of authority. To admit that this person is stronger is to admit that management was wrong. For many organizations, that admission brings not a learning opportunity but a humiliation.

So a common reason innovation fails is not that managers dislike innovation. It is that managers cannot accept innovation coming from someone they cannot understand—someone they may even need to learn from. Karp puts it more starkly. A company with no geniuses may simply lack geniuses. A company that has them and cannot keep them lacks the capacity to hold them. The latter is more common and much harder to cure.

Yet Karp does not romanticize exceptional people. He acknowledges that they are not "always right." The hard part is making two things true at once: judging whether their ideas have value, and managing them without destroying their distinctive creativity. That demands a higher order of management—not getting everyone to submit, but getting people who are stronger than you in some dimension to stay and do their best work.

There is an easily missed boundary here: an organization's cognitive ceiling should not be identical to its leader's. If every thought that challenges authority is suppressed the moment it appears, the company loses more than one person's idea. It loses the organization's capacity to correct its mistakes and find new directions.

3. Technology Can Build a Product Without Changing a Company

Having discussed talent, Karp pushes the question to the other side: even if a company assembles the best geniuses, will that create business value?

Not necessarily. Being technologically ahead is not the same as creating business value. If a team builds a world-leading AI product, will customers automatically want to use it? If a company buys the most advanced system, will its existing way of working change on its own? Of course not. A vast "innovation chasm" sits in between.

Karp offers a concrete metaphor. Someone asked him: how do you get the first olive out of the jar? In other words, how do you get the first customer to try your product? But Karp argues that the truly difficult part is not getting the customer to take the first bite. It is changing the customer's palate—getting them to genuinely understand why this olive is worth eating.

Behind the metaphor is an important business logic: getting a customer to try is not the same as getting them to approve. Getting them to approve is not the same as getting them to adopt. A technology demo can surprise a customer without changing the company. "Adoption" means redesigning processes, shifting roles, redrawing lines of accountability, and sometimes challenging the existing power structure. The biggest bottleneck for a technology may not be what it can do, but whether the organization can understand and adopt it.

In other words, technological innovation and value realization are not the same thing at all.

4. FDEs: Put Technology Into the Workflow, and Let the Field Define the Product

This is why Karp places such weight on the FDE, or Forward Deployed Engineer. FDEs do not just deliver software. They go into the real business environment, understand the problem, adjust the system, and make technology actually enter the company's workflow.

More interestingly, Karp stresses that this mechanism of technology transfer also shapes, in reverse, what products the company should build. The logic changes fundamentally here. You do not build technology first and then search the market for customers. You solve real problems and, in the process, keep discovering what value the technology should create.

That sounds like a tweak to product methodology. In practice, it changes who is qualified to define the product. If product definition happens only inside the office, engineers optimize features. If it happens at the customer's workplace, engineers optimize outcomes. Building a product is a technical problem; changing how a company behaves is a business problem. A product's real value lies not in how many features it has, but in what results it changes.

5. The New Efficiency Trap of the AI Era: Execution Gets Cheap, Judgment Gets Expensive

AI is amplifying this contradiction further.

Today, models are getting stronger, tokens are getting cheaper, and the barrier to using AI is falling. Yet many companies are discovering that AI usage keeps rising while real business returns do not keep pace. Karp mentions CEOs who are furious because they are paying for enormous numbers of tokens that are not producing meaningful business results.

A new efficiency trap emerges. Employees generate dozens of reports a day; AI automatically executes hundreds of tasks. Productivity appears to have exploded. But if no one uses the reports and the tasks do not improve business outcomes, what exactly did these AI systems create?

Consider a simple thought experiment. A process that created no value used to require ten people. Today one or two might keep it running. Efficiency has clearly improved, but the "no value" part has not disappeared. It is simply being produced faster and more cheaply. AI can amplify efficiency, and it can amplify waste just as well.

That is why Karp keeps insisting that AI cannot be measured only by how many tasks it executes or how many labor hours it saves. It must be measured by whether it changed a real business outcome. When the cost of execution approaches zero, what becomes scarce is no longer "doing it faster" but "judging what is worth doing."

6. Anyone Can Buy a Model; No One Can Buy Tacit Knowledge

The AI era has another, subtler cost: the leakage of proprietary knowledge.

A company's years of supplier selection experience, production scheduling methods, and procurement judgment rarely appear in full on its financial statements, yet they genuinely determine its competitive advantage. Why can an experienced procurement manager spot a supplier's latent strengths at a glance? Why can a seasoned sales leader tell which customers have real intent to buy? These abilities rarely come from a textbook. They come from long practice, trial and error, and business understanding—and they cannot always be fully written down.

In the past, this experience depended mainly on being passed down by a few people. AI opens a new possibility: turning individual tacit knowledge into a judgment system the organization can reuse continuously. But this is not as simple as putting documents into a knowledge base. What is actually required is letting AI understand the company's business relationships, decision rules, and workflows so that experience can be called upon, tested, and corrected.

If a company fails to control the scope and flow of this proprietary knowledge, the money it saves may not offset the cost of that knowledge leaking. Models are available to anyone on the open market. The tacit knowledge, judgment rules, and feedback records accumulated over years are the part that cannot be copied.

7. Talent, Organization, Knowledge System: A Complete Innovation Value Chain

Put Karp's arguments together, and the same logic runs beneath them.

Exceptional talent discovers problems others cannot see and creates possibilities others cannot imagine. Organizational capability turns those possibilities into real business results. A knowledge system turns one-off innovation into a competitive advantage that compounds.

These three form a complete innovation value chain. Remove one link and the chain fails.

Talent without organization: creativity may stay trapped in an individual, never becoming a product, a process, or revenue. Organization without distinctive judgment: the company simply executes rules more efficiently without asking whether the rules still hold. AI tools without accumulated knowledge and real feedback: the company will struggle to build a lasting advantage, because the intelligence it uses is the same intelligence everyone else is using.

8. The Real Risk Is Grading Your Most Valuable People as "Unqualified"

The stronger AI becomes, the cheaper execution gets—and the more expensive judgment becomes.

Companies once relied on standardization to let more people execute reliably. As standardized work becomes easier for AI to perform, what becomes scarce is not just execution capacity but the ability to find problems, create differences, and judge value.

So could the distinctive abilities once treated as flaws be turning into advantages? Could the people once considered hard to manage be the ones holding the hardest creativity to copy? And could organizations that demand everyone conform to the standard be eliminating their own future by hand?

Karp's warning can be compressed into one sentence: the greatest danger for a company may not be a shortage of geniuses, but a seemingly reasonable management system that turns its most valuable people into "unqualified" ones.

When an era no longer lacks executors, the most expensive capability is no longer obedience to the standard, but the ability to create a new angle.