In the AI Era, Why Does Trust Grow Scarcer as Ability Becomes Easier to Demonstrate?

A recent engineering graduate can now use AI to produce a complete technical proposal in a day, build a working demonstration system, and even answer most interview questions.

Yet none of those outputs can answer several questions for a hiring manager. Does this person genuinely understand the trade-offs in the proposal? Can they identify and communicate risks when the system fails? If they are given a task that will last several months, will they see it through?

Exams, degrees, certificates, and portfolios have long translated individual ability into signals that others can recognize. For young people without family advantages or professional connections, relatively standardized forms of assessment have provided an important route to upward mobility. AI is not overturning that entire system, but it is changing part of it.

AI has lowered the cost of producing professional-looking work, but it has not lowered the cost of judging whether a person is reliable. As a result, “Can this person do the work?” and “Can this person be entrusted with the work?” are becoming increasingly distinct questions.

For someone entering professional life, the hardest problem is often not learning a skill. It is securing the first opportunity to be tested on a real task when no track record yet exists.

I. AI Has Lowered the Cost of Demonstrating Some Abilities #

Generative AI has improved performance on some knowledge tasks, but these experimental results do not mean that professional expertise no longer matters.

A study of 5,172 customer-support agents found that access to a generative AI assistant increased the number of issues resolved per hour by an average of 15%. Less experienced and lower-skilled workers improved more substantially. More experienced and higher-skilled workers achieved only small gains in speed, while the quality of their service declined slightly[1]. The study examined a specific customer-support setting within one company, so its findings cannot be assumed to apply to other occupations. At a minimum, it shows that AI affects different tasks and people with different levels of experience in different ways.

The more visible change today is that AI has lowered the cost of completing and demonstrating certain abilities. People can produce code, reports, charts, and proposals more quickly, and they can more easily present something that looks professional. That is a real productivity gain, but the output itself now reveals less about the person who produced it.

In the past, a complete technical proposal might have conveyed two things at once: that its author could produce the work and that the author understood the judgment behind it. Those signals can now come apart. AI can help someone produce an output, but the output alone cannot show that the user understood the constraints, checked for errors, or is prepared to take responsibility for the consequences.

Capability has not lost its value. What has changed is that an organization can no longer infer as much about capability from a single piece of work—and even less about reliability.

II. Why Capability Does Not Automatically Become Trust #

Capability answers, “Can you do it?” A track record answers, “Have you followed through before?” Reputation is the judgment others form from those records, while trust determines whether they will place the next task in your hands. An endorsement becomes important when records are scarce: a third party temporarily places some of their own reputation behind a person whose reliability is not yet known.

These concepts are related, but they are not interchangeable. Someone may be capable of completing a task without having produced a verifiable record of follow-through. They may also have completed one task without demonstrating that they can assume responsibility over a wider scope or a longer period.

Real work is rarely judged only by its final output. A project usually involves collaboration, deadlines, budgets, and risk. An organization must assess not only whether someone can produce an answer, but whether they will disclose uncertainty, conceal problems when something goes wrong, or make sound trade-offs when interests conflict.

This is also why AI cannot acquire trust on a person’s behalf. AI can provide analysis, but people must still choose which proposal to adopt, which risks to accept, and who will explain and remedy a failure. Many difficult organizational decisions have no single correct answer. They involve priorities, accountability, and the distribution of benefits and costs.

Trust, then, is not a permanent verdict on someone’s character. It is a willingness to entrust that person with future work within a particular scope of responsibility and risk. Newcomers often do not lack potential capability; they lack records that show why they should be trusted in this way.

III. Why Initial Trust Is Often Borrowed #

Societies have always faced a difficult credibility cold-start problem: why should an organization trust a complete stranger?

A common response has been to reduce uncertainty by relying on people and institutions that already possess credibility. A degree from a prestigious university, training under a respected mentor, and experience at a well-known company are primarily external signals. A teacher or former colleague uses their own reputation to provide an endorsement only when they explicitly recommend someone.

External signals cannot prove that a person is outstanding. They show only that a selection, training, or working environment with some established reputation has encountered that person. Labor-market signaling theory uses this mechanism to explain how educational background can influence an employer’s judgment[2]. Without a sufficient track record, a newcomer can temporarily rely on these signals or endorsements to obtain a limited opportunity. This is why initial trust is often borrowed rather than earned.

For people with prestigious educational backgrounds, distinguished mentors, or experience at prominent companies, those credentials may open the door to interviews, a first job, and later opportunities. But they provide opportunities to be tested, not lasting trust. As real collaboration accumulates, others will rely less on background and more on the individual’s record of follow-through. If repeated opportunities do not produce an independent track record, the original credentials lose explanatory power. Their most valuable role is to create access to real collaboration, where opportunity can be converted into credibility of one’s own.

An explicit recommendation carries more direct responsibility. Someone who has worked with you over time knows whether you procrastinate, conceal problems, make excuses after mistakes, or take steps to repair the damage. When that person tells someone else, “I have worked with them; you can entrust the work to them,” they are putting their own judgment at stake. If the person they recommend repeatedly conceals risks or breaks commitments, others will also place less trust in the recommender. Such endorsements therefore tend to cover only a small number of opportunities and a limited scope of responsibility. The value of a professional relationship lies not merely in “They know me,” but in their willingness to put their name behind yours.

Poor performance by a single student or former employee is usually not enough to damage the credibility of a school or company. Only when an institution repeatedly sends signals that do not match actual quality will outsiders begin to trust its selection and training processes less. Whether external credibility comes from a person or an institution, it cannot replace a newcomer’s own record of follow-through.

These mechanisms can also reproduce existing inequality. Prestigious universities, prominent companies, and powerful professional networks are more visible, so people without those backgrounds may struggle to secure a first chance even when they are equally capable. Explaining how external credibility reduces uncertainty does not justify allocating opportunities through personal connections. The former concerns how organizations make judgments; the latter concerns whether access is fair.

IV. How Borrowed Credibility Becomes Your Own Track Record #

External signals and explicit endorsements can provide entry, but they cannot build a person’s professional credibility for them. The opportunity a newcomer receives is usually not responsibility for a critical project, but a collaboration in which the risk can be contained.

Tasks suited to a credibility cold start usually have three characteristics: a clear scope, a manageable cost of failure, and an outcome that a third party can verify. The task might involve organizing a dataset, writing a module, completing an analysis, resolving a customer issue, or fixing a small problem in an open-source project.

Whether the task looks impressive matters less than whether it creates a clear commitment: if this work is entrusted to you, can you complete it within the agreed boundaries?

Complete one task, and someone may entrust you with a little more the next time. As the scope of responsibility expands, a person moves from carrying out tasks to managing projects, and from managing projects to exercising judgment and assuming responsibility for outcomes. Their own record of follow-through gradually replaces the external credibility on which they initially depended.

Being reliable does not mean never making a mistake. Real work is inevitably affected by estimation errors, technical failures, and changing external conditions. What matters for credibility is often how someone responds: did they disclose the problem early, conceal the risk, attempt to repair the damage, and explain the cause and consequences clearly?

The same failure can produce very different records of responsibility. Someone who raises a problem early and protects the team from greater harm may earn more trust as a result. Someone else may conceal the problem for a time, only to lose long-term credibility when the risk finally surfaces.

Early in a career, what matters is usually not a single opportunity to become an overnight success, but a series of small commitments with clear boundaries and verifiable outcomes. Capability can become credibility only through real collaboration. After credibility has been tested repeatedly, it can become reputation and support a larger measure of trust the next time.

V. How Society Can Create Entry Points for a Credibility Cold Start #

What can a young person do if they did not attend a prestigious university, have few family resources, and know no one willing to provide an initial endorsement?

Standardized examinations have historically provided one important route. A person can study and prepare independently, then use postgraduate entrance examinations, civil-service examinations, professional qualifications, or other public forms of assessment to turn private effort into a signal that strangers can recognize. These systems are imperfect, but they reduce the extent to which initial recognition depends on private relationships.

It is still too early to conclude whether AI will weaken the ability of some examinations to distinguish among candidates. Even if assessment methods change, society will still need public mechanisms that convert unseen individual capability into a first opportunity for collaboration.

Credibility, however, cannot be built entirely at home. You can learn to program alone, but you cannot independently prove that you are a reliable collaborator. Real collaboration reveals things that a portfolio cannot: how someone interprets requirements, handles disagreement, responds to feedback, and faces failure.

Open-source collaboration offers one possible model. Participants can submit code, improve documentation, respond to questions, and take part in discussions, leaving a public record of the process. Open-source contribution is not “permissionless”: maintainers still decide whether to accept a submission, and contributors must understand the project’s rules. Its value is that some real tasks have a relatively low barrier to entry, while both the process and the result can be inspected by third parties.

Internships, student projects, community work, and graduate programs can serve similar purposes. An effective entry point need not be completely open, but it should usually have several features: a sufficiently low cost of entry, a task grounded in a real need, a verifiable outcome, a durable record of contribution, and consequences of failure that remain bearable.

This responsibility cannot be placed entirely on young people. If society merely tells them to build a personal brand, meet senior people in an industry, and accumulate credibility on their own, it turns a structural problem into an individual burden. Every industry needs newcomers, and must therefore allow some people who have not yet been proven to enter with limited risk. Otherwise, people without experience will be denied opportunities because they lack experience, and will never acquire experience because they are denied opportunities.

Schools, companies, professional bodies, and public platforms can jointly build the infrastructure for a credibility cold start. They can divide real work into clearly bounded, low-risk tasks, specify evaluation criteria, give newcomers feedback, and ensure that reliable records of follow-through can be recognized by the next collaborator.

Conclusion #

Young people cannot accumulate professional credibility entirely on their own before entering professional life. Schools, former employers, and public assessment systems provide external signals; recommendations from teachers and former colleagues provide limited endorsements. Together, they help newcomers secure a first opportunity to be tested.

What ultimately determines professional credibility is how a person handles each commitment thereafter: whether they complete the task, disclose problems in time, and attempt to repair a failure. AI can help people produce results more quickly, but it cannot leave these records of responsibility on their behalf.

Capability becomes credibility only when it enters real collaboration and produces verifiable results. Once credibility has accumulated, others become willing to entrust that person with more consequential work.

References #

[1] Brynjolfsson, E., Li, D., & Raymond, L. (2025). Generative AI at Work. The Quarterly Journal of Economics, 140(2), 889–942. https://doi.org/10.1093/qje/qjae044

[2] Spence, M. (1973). Job Market Signaling. The Quarterly Journal of Economics, 87(3), 355–374. https://doi.org/10.2307/1882010