AI hiring: 86% of US colleges already use AI in job search coaching

The number: 86% of US colleges use AI in job search coaching

The National Association of Colleges and Employers survey sets a clear threshold: as of April 2026, 86% of US higher education institutions report using AI in one-on-one job search work with students, as Fast Company reports^[1]. The AI hiring market therefore starts long before the first interview.

Almost all of those institutions buy the tool from an outside vendor.

The number measures uptake. It measures how many career centres have switched a piece of software on, and that remains a different thing from the capacity to design that software. The gap between those two measures carries this entire analysis.

For employers the effect lands immediately. An entry-level candidate's CV, cover letter and interview preparation now pass through a system the hiring organisation has rarely seen in operation. The filter acts upstream of selection, outside the perimeter of whoever eventually decides.

Dartmouth moves from buying to building

Dartmouth College has created a chief product officer role inside its career centre, the first of its kind at an Ivy League institution. The role went to Kiley McEvoy, a former senior product manager at Google and an alumna of the college.

The product under development accompanies the student throughout their undergraduate years, with regular questions about interests, skills and professional ambitions.

Joe Catrino, executive director of the Dartmouth Center for Career Design, describes the classic offering of career services as something that has stayed identical for decades. The statement matters more than the single hire: it defines career guidance as a product, and therefore as something you design, measure and correct.

Bringing a product role inside a university shifts the question. You move from "which tool did we buy" to "who holds the skill to design the journey". The second question applies to every organisation that manages people.

The first door is designed upstream

The entry point into the labour market remains the most exposed segment. Entry-level roles absorb first the reorganisation of the tasks AI makes automatic, and careers tend to compress toward the top.

US labour market data has for months recorded mounting pressure on new graduates: relative unemployment rising and entry-level hiring falling compared with the rest of the workforce.

The Dartmouth case adds something new to that picture. When the infrastructure that prepares candidates is designed by people with product expertise, the quality of that preparation stops being accidental and becomes an engineered variable.

For employers the consequence is concrete. The signals arriving from a candidate (CV format, clarity of goals, fit with the role) also reflect the quality of the system that trained them. Reading those signals as pure individual merit therefore becomes a measurement error.

Adoption and design capability remain two different measures

The 86% figure tells a story of broad adoption. It also tells a story of dependency: most institutions hand a vendor the logic by which a student gets steered toward a sector, a role and an expected salary.

The same dynamic runs through hiring organisations.

Buying a screening platform takes a signature. Redesigning selection around that platform takes internal expertise, manager time and a clear definition of what counts as potential. The first happens in a few weeks, the second over several quarters.

The bottleneck in adoption remains the readiness of those who lead, more than the willingness of everyone else. Excellent tools in the hands of unprepared leadership deliver less than mediocre tools in the hands of literate leadership. The Dartmouth case shows the positive version of the same rule: in-house product expertise raises the level of ambition across the whole service.

The academic objection and the answer on the merits

The criticism exists and deserves room. A Chronicle of Higher Education newsletter dated 16 September 2026 called the Ivy League's rush toward AI senseless, in an opinion piece that captures the unease across part of the academic world^[2].

The objection hits a real risk: replacing the educational relationship with an automated stream of suggestions.

The useful answer lies in the difference between delegating and designing. An institution that buys a turnkey tool hands the criteria to a vendor. An institution that hires a product role keeps the criteria in house and makes them debatable, measurable, correctable.

The same holds for employers. The value of a selection technology depends on who owns its rules, and ownership of the rules remains a governance choice before it is a technical one.

Bias enters from both sides of the process

The recruiter side brings a second problem. A large share of HR teams already use some form of automated screening, stated confidence in AI-conducted interviews is rising, and alongside it rises the number of professionals reporting that they have observed bias in the results.

Confidence is running faster than measurement.

When both ends of the process are mediated by models (the student trained by one tool, the candidate filtered by another), the outcome depends on the alignment between two systems built by different vendors, with different objectives.

A university optimises placement for its own graduates. A screening platform optimises reduction in the volume of applications to read. The two objectives barely touch, and the person in the middle absorbs the entire difference.

What the organisations closing the gap actually do

High-functioning organisations move on three fronts, all visible in labour market data from the past two years.

• Converting internal talent before turning to the external market
• Redesigning entry roles around the tasks AI leaves uncovered
• Transparency with candidates about the role of models in selection

The first front is about cost. Redeploying a person into a new role costs less than firing and rehiring, and organisations with an up-to-date skills map remain a narrow minority.

The second front is about the content of the work. An entry role built around tasks that are now automatic loses its point within months, while a role built around judgement, client relationships and verification of outcomes survives the transition.

The third front is about trust. Universities are already explaining to students how the tools they use with them work, and the same question will soon reach employers.

The metrics the board should see

The Dartmouth case produces four operational questions, one for each decision-making table.

• CEO: is the organisation's front door a designed process or a historical leftover?
• CHRO: who, inside the team, holds product expertise on selection tools?
• CFO: what share of HR technology spend funds internal capability rather than external licences?
• Talent & Compensation Committee: which metric measures the quality of entry-level hires at twelve and twenty-four months?

The 86% of US colleges is a figure to take into the boardroom. It describes a change in how every future cohort of candidates is prepared, and it arrives before any internal decision about recruiting.

One college decided to treat career guidance as a product and bought the expertise to do it.

The design question for anyone leading an organisation stays symmetrical: which part of your people's entry journey is currently designed by someone who answers to your board, and which part is designed by a licence agreement signed three years ago?