AI and Education

The conversation about AI and education has been dominated by two concerns: that AI will make it easier for students to cheat, and that AI tutors will eventually replace teachers. Both concerns are real in some form. Neither is the most important thing happening.

The most important thing happening is that AI is creating new paths around the credentialing system — and those paths are beginning to matter in ways that will change who gets access to economic opportunity.

What the existing system does

The formal education system does three distinct things that are often conflated.

It transmits knowledge and builds capability. A student who learns mathematics, writing, programming, or any other discipline is developing a capability that has genuine value.

It signals capability to employers. The credential — the degree, the institution, the GPA — is a signal that employers use to make hiring decisions because they cannot directly observe the underlying capability.

It provides access to networks. The alumni network of a well-regarded institution, the relationships formed with peers and professors, the exposure to professional contexts that educational environments provide — these are genuine assets that graduates carry.

AI is relevant to the first of these in meaningful ways. It is almost irrelevant to the second and third, which is where most of the economic value of formal education is concentrated.

What AI actually changes about learning

Access to expert-level explanation is no longer scarce. A student in a well-resourced school has always had access to teachers who can explain concepts at the level the student needs, identify where understanding has broken down, and adapt the explanation accordingly. A student in an under-resourced school has had access to the same curriculum but lower-quality explanation, feedback, and adaptation. AI tutoring tools are beginning to close this gap — providing the kind of responsive, adaptive explanation that was previously available only to students with access to high-quality teachers or expensive tutors.

This is genuinely important for students in under-resourced educational environments, including the rural India context that I know directly. The student who couldn't afford a tutor and whose school had 60 students per class now has access to a patient, infinitely available, highly capable explainer. This matters.

The cost of reskilling is dropping. Learning a new technical skill — programming, data analysis, digital marketing, a new language — has historically required either formal education (expensive, time-consuming) or self-study (possible but high-effort, high-failure-rate). AI tutoring systems make self-directed skill acquisition more accessible and more effective. The person who wants to learn Python can now do so faster, with better feedback, and at lower cost than was possible five years ago.

The feedback loop on writing and thinking has shortened. One of the most valuable things a good teacher does is give specific, actionable feedback on student work — not just "this is good" or "this needs work" but "the argument breaks down here because the causal claim isn't supported by the evidence you've cited." AI writing tools are beginning to provide this level of feedback at scale. For students who haven't had access to teachers who give this kind of feedback, this is a new capability.

What AI does not change

The credential still signals. Employers hiring at scale use credentials to filter candidates because they have no better tool. An employer processing 500 applications for a software engineering role will not individually assess every applicant's coding capability from a portfolio. They will filter for candidates from institutions they recognise, which is a proxy for the capability they actually want. AI learning tools produce capability. They do not produce credentials that employers recognise.

The network advantage is unchanged. A graduate of a well-regarded institution has access to alumni networks, institutional brand associations, and the professional contacts formed during their education. These advantages do not exist for someone who learned the same skills via AI tools. The network advantage is structural — it depends on membership in a social institution — and AI does not create social institutions.

The paths being created

What AI is creating is not a replacement for formal education. It is a set of paths around the credentialing bottleneck for the specific cases where the bottleneck can be bypassed.

Software development is the most established example. The ability to demonstrate capability through code — through open source contributions, through portfolio projects, through technical assessments — has created a path to well-paying employment that doesn't require a formal credential from a prestigious institution. AI tools that make learning to code faster and more accessible are expanding this path.

Similar dynamics are developing in data analysis, digital marketing, content creation, and other fields where the work product is directly evaluable. They are not yet developing in fields where the credential is regulatory (medicine, law) or where the network is the product (finance, consulting).

The people who benefit most from the paths AI is creating are those who have the capability and motivation to develop real skills but have lacked access to the environments where those skills are taught well. That is a significant population in many parts of the world, including the parts I come from.

The ceiling on how far those paths can take someone is still set by the credentialing system's grip on the most lucrative professional pathways. AI loosens it in some domains. It does not break it.