AI Has Made Applying Easier. It Hasn’t Made Hiring Better.

Getting another application is easy. Working out whether the person behind it can do the job is still hard.

AI can rewrite a CV, tailor a cover letter and prepare interview answers in minutes. That can help good candidates explain themselves more clearly. It also makes a polished application a weaker signal than it used to be.

For Seed–Series B AI founders hiring engineers and GTM talent in the US, the response should be a better assessment process. Adding more filters to an unclear brief simply automates the confusion.

The application problem is now an evidence problem

Greenhouse’s research published on 20 October 2026 found that 77% of surveyed US job seekers had used AI in their search in the previous year. It also found that 81% had been rejected during that period without speaking to anyone at the employer. Those figures describe the US sample of 1,200 respondents within a wider international survey; they are not measurements of every US hiring process.

Our interpretation is straightforward: both sides have more tools, but tools alone do not create understanding. A founder still needs to establish what somebody personally owned, how they made decisions and whether their experience transfers to this stage of business.

A CV saying “built an AI-powered growth engine” tells you very little. What was built? Who used it? What changed commercially? Which part was the candidate’s work? Those are recruitment questions, and no amount of polished wording answers them automatically.

Stop screening for the ability to sound right

Early-stage hiring is particularly exposed because the role itself is often moving. A founder wants a marketer who can own positioning, create pipeline, run events and build the website. An engineering brief combines product delivery, infrastructure, model evaluation and customer support.

A candidate can match every keyword without being the person you need first. Before publishing the role, agree three or four outcomes for the first six months and separate essential experience from skills someone can learn.

For a founding engineer, an outcome might be shipping a reliable customer workflow with sensible evaluation and monitoring. For a founding seller, it might be finding a repeatable route into a defined buyer group. Neither is adequately tested by asking whether someone has “worked with AI”.

Our guide to AI engineers, ML engineers and research scientists explains why defining the actual work matters more than choosing a fashionable title.

Ask for the decisions behind the result

The useful part of a candidate’s story usually starts after the headline. “We increased conversion” is an opening, not evidence of ownership.

Ask what the starting point was, which options they considered, what they personally changed and what went wrong. Then follow one example through in detail. A strong candidate can usually describe the constraints and trade-offs without turning the conversation into a perfect success story.

This also avoids unfair assumptions about presentation. Someone may use AI to improve their written English, organise their experience or reduce the administrative burden of applying. That does not tell you whether they lack judgement. Equally, confident delivery does not prove competence.

We should be testing the work, rather than guessing who used a writing assistant.

Make practical assessment resemble the real job

A short, relevant working session can reveal more than another broad interview. Give an engineer a small system design problem with incomplete information. Ask a product marketer to critique positioning for a defined audience. Ask an AE how they would qualify a buyer whose enthusiasm is ahead of their budget.

Explain the exercise, the time required and the criteria in advance. Provide reasonable adjustments where needed, avoid extracting usable unpaid client work, and pay for substantial assignments.

Be clear about AI use too. If the job involves AI tools, banning them throughout assessment may test the wrong thing. You can allow their use, ask the candidate to explain how they checked the output, and include an agreed discussion without assistance to explore understanding.

The US Office of Personnel Management’s structured interview guidance offers a useful foundation: consistent job-related questions and common assessment standards. A startup can apply that discipline without building a corporate interview marathon.

More stages will not repair weak decision-making

If nobody owns the hiring decision, adding a final founder conversation or another take-home creates delay rather than confidence. Decide who evaluates technical depth, commercial judgement and working style before interviews begin.

Have interviewers record evidence independently before the debrief. Separate a missing requirement from a preference. When a candidate meets the agreed bar, make a decision instead of reopening the search because there might be somebody better somewhere.

Review the process as well as the funnel. Are qualified people dropping out? Are interviewers repeating questions? Does rejection feedback match the original brief? An application count is an activity measure; it is not evidence that your hiring is improving.

Where a recruiter should add value

A recruiter sending more profiles into this system is adding volume. The useful work happens earlier: challenging the brief, explaining the market, approaching relevant people and establishing the context behind their achievements.

At Dataworks, our work recruiting for Seed–Series B AI businesses across the US and Europe puts engineering and GTM requirements alongside the conditions of the role: stage, founder involvement, compensation and onsite expectations. A plausible CV is the start of that conversation.

For candidates, the same principle helps. Use AI to organise your application if useful, but make your contribution specific. Explain what you built, sold or changed, what support you had and what you learned when the first approach failed.

Questions founders should settle this week

• What counts as success? Write the outcomes before choosing screening keywords.
• What will we verify? Select a small number of relevant examples and a practical exercise.
• Where is AI allowed? Tell candidates before assessment, with clear boundaries.
• Who decides? Assign decision ownership and a realistic timetable.

AI has made presenting an application easier. Better hiring still depends on doing the harder work behind the decision.

If your pipeline looks busy but the shortlist is weak, talk to Dataworks about the brief and assessment process. You can also read what we look for in an early-stage AI seller for a practical example of evaluating beyond the CV.

Photo by Christina @ wocintechchat.com on Unsplash.