I used AI to apply for hundreds of jobs. The next step is a better argument.

AI-assisted founder authorship

AI-assisted author identity used at John N. Farmer’s request. John is OMIKINA’s founder and accountable publisher, not a fictional reporter. Text is prepared with AI assistance; authorization to publish does not itself establish completed human editorial review.

Key points

• The ledgers record 498 distinct openings from September 27 through October 9, 2026; 491 have captured receipt artifacts and seven have agent-observation records. Applications and resume sends are not interviews or offers.
• The workflow connects profile and portfolio updates with a verified resume, role-specific responses and deduplicated submission receipts.

  Sources: S1

• Hiring managers using AI to manage volume is reasonable in my view, provided they remain accountable for the criteria, evidence and consequential decisions.

  Sources: S2 · S3 · S4

• My next experiment is a concise, job-specific presentation. Better response or hiring outcomes have not yet been established.

One person, hundreds of applications

I am one person. With AI assistance, I have built a job-search workflow that can find relevant openings, update my professional materials, prepare tailored responses and carry an application through to a recorded submission. That changes the amount of work one applicant can attempt. It also makes the hiring manager’s problem easier to understand.

My rough impression was that we had reached about 600 applications. The working ledgers record 498 distinct openings across September 27 through October 9, 2026: 462 employer-form submissions or acknowledgements and 36 roles addressed through recruiting emails. The email total counts roles: 35 distinct messages covered those 36 roles. General outreach, uncertain attempts, researched openings and known duplicate attempts are excluded.

The archive retains captured submission or sent-email receipt artifacts for 491 of those entries. Seven have preserved agent-observation receipt records rather than captured confirmations. This audit reconciled the ledgers and available artifacts; it did not individually reopen every historical confirmation screenshot. That distinction belongs beside the total, not hidden behind it.

Those are application actions, not 498 interviews, employer reviews or offers. The total spans nearly two weeks, with concentrated bursts of work; it would be misleading to describe every submission as happening in just a few days. The records prove throughput. They do not yet prove that more applications produce better opportunities.

Start with a profile that agrees with the resume

The first step was not a clever cover-letter prompt. It was making my professional profile, resume and portfolio tell a coherent story. Creative direction, applied AI, interactive products, visual storytelling and healthcare innovation needed to connect to work someone could actually inspect. A hiring manager should not have to reconcile three incompatible versions of the same person.

We updated the positioning and resumed reviewing the materials as the search evolved. The resume became a verified base, not a license to invent credentials for each listing. Relevant projects and evidence could move forward; employment dates, actual experience and established facts had to stay consistent. My portfolio provides concrete work behind the summary.

Sources: S1

Sources: S1

Curated responses, with a record of what happened

The application work begins with the posting: what the employer needs, what the role actually involves, whether it fits the search, and where the official application belongs. Bounded agents can research different lanes in parallel. A shared record helps us catch duplicates and distinguish old listings from current opportunities.

Then we tailor the emphasis. A creative-director opening should receive a response about creative leadership and the relevant work. An AI-focused role should see the actual workflow integrations and products. A producer should see production experience. The response should answer that employer’s question rather than paste a polished paragraph that could belong to anyone.

The important constraint is truth. AI can overstate a scope, compress a timeline badly or confidently answer something we have not verified. Our process has held applications when required experience was unknown, and reviews have caught wording that needed correction. That is a reason to strengthen verification, not to describe the system as error-free.

Finally, the ledger separates an intention from an action. A portal confirmation, a recruiting-email send, an unfinished form and an uncertain submission are different states. We retain the role, destination, resume version and receipt. Otherwise, a busy agent can produce an impressive number that says very little about what actually reached an employer.

Hiring managers probably should use AI, too

Seeing the applicant side at this scale has changed my view of screening. I think hiring managers are probably right to use AI to organize and sort applications as it becomes a more normal part of work. If one person can pursue hundreds of roles, expecting every recruiting team to begin with a completely manual read is difficult to defend.

This is already an explicit product workflow. LinkedIn describes its Hiring Assistant producing candidate recommendations with evidence and reasoning while recruiters decide who advances. Ashby documents review against employer-defined criteria, including an undecided result and a way to flag an incorrect evaluation. Those are descriptions of particular tools, not proof that every employer uses AI or that any specific rejection came from it.

Sources: S2 · S3

The condition is accountability. AI should help a hiring team find relevant evidence and manage volume; a score should not erase the person behind it. Managers need to own the criteria, inspect mistakes, allow correction and keep meaningful human judgment in the consequential decision. NIST’s AI risk framework supports documented oversight and evaluation of risks such as bias. Tool use alone does not establish fairness.

Sources: S4

I cannot reasonably celebrate AI expanding my reach and insist that employers never use it to cope with the resulting volume. I can ask both sides to use it responsibly, explain their claims and distinguish an absence of evidence from evidence of an absence.

Sources: S2 · S3 · S4

The next experiment: a short presentation for the actual job

The next improvement I want is more substance per application. AI now makes it easier to prepare a custom presentation or slide deck around a specific posting. For a creative leadership role, that could make my thinking easier to evaluate than another general-purpose resume. It may be more effective; I have not demonstrated that it improves response, interview or offer rates.

I would keep it short: five slides covering the problem I believe the role needs to solve, relevant work I have already completed, a clearly labeled proposed approach, the AI workflow and verification I would use, and what I would measure in the first 30 days. Completed work and proposed ideas must be visibly different. A deck should show judgment and evidence, not pretend I have already worked for the company.

There is a useful principle behind this. The U.S. Office of Personnel Management describes work samples built around activities representative of the job. An unsolicited AI-assisted presentation is not automatically a validated assessment, and attractive slides do not prove competence. But the idea points toward something worthwhile: make it easier to inspect how a candidate approaches the work.

Sources: S5

For selected roles, I want to compare the ordinary tailored application with an application that includes a relevant brief, while keeping fit, resume quality and the response window in view. I would track employer responses and interviews separately from how many decks we generate. Small samples and different jobs will limit what that comparison can tell us.

AI has made applying faster. The next question is whether it can help me make a clearer, more useful case for a specific role. That is the shift I want: from increasing the number of attempts to making each well-chosen attempt easier for a human to understand.

Sources: S5

Why it matters

AI changes the volume one applicant can produce and the workload employers face. The useful next step is making truthful, job-relevant evidence easier to evaluate while measuring hiring outcomes separately from output.

Sources: S2 · S3 · S5

Sources

1. John N. Farmer — professional portfolio and resume — John N. Farmer
2. How we engineered LinkedIn’s Hiring Assistant — LinkedIn Engineering · 2025-10-21
3. AI-Assisted Application Review — Ashby Knowledge Base
4. Artificial Intelligence Risk Management Framework 1.0 (January 2023) — National Institute of Standards and Technology
5. Work Samples and Simulations — U.S. Office of Personnel Management