You Don't Need a $312-a-Month SaaS to Hire Two People

Two open roles. Four SaaS seats. $312 a month, signed for a year, before anyone had opened a single resume.

You Don't Need Four SaaS Seats to Hire Two People

Picture Jonas. He runs a twelve-person plumbing and electrical outfit outside Manchester, and he needs two people this quarter: an apprentice electrician and a dispatcher who can juggle the phone and the job board. He gets on a call with an AI hiring platform. The demo is smooth. A sourcing agent, a screening agent, a scheduling agent. Three problems, three agents, done. The quote lands at four seats, $78 each, $312 a month, on an annual contract. He is about to sign.

Stop. The vendor is right about the pain. Sourcing eats a full day. Screening eats another. Scheduling eats a third in fifteen-minute increments spread across two weeks. That part is true and it is expensive in the currency that actually matters to a small business, which is the owner's attention. What the vendor does not say out loud is that none of those three jobs is a hard problem. They are four automations and one API key.

What Does an AI Hiring Agent Actually Do?

Not a chatbot. A chatbot answers a question and stops. An agent finishes the job.

A chatbot bolted onto a careers page can say "thanks for applying, we'll be in touch." An agent watches the inbox, pulls the resume text out of the PDF, scores it against a rubric you wrote, writes that score back to a spreadsheet, and pings you the moment someone clears the bar, all without you opening the folder. The difference is not intelligence. It's follow through. The chatbot greets the applicant. The agent does the work that used to be your Tuesday.

This is the same shift that has already hit customer support and invoicing: the software stops being a form you fill in and starts being a worker that finishes the loop on its own.

Why This Is Suddenly Cheap Enough to Build Yourself

That number is not a typo. Reading and scoring a resume against a rubric is a small, well-defined task for a large language model, so it costs fractions of a cent per candidate. The expensive part was never the intelligence. It was the plumbing that got the resume in front of the model and the decision back out to a human, and that plumbing is exactly what a workflow tool like n8n or Zapier is built to do for free or close to it.

The SaaS platform in Jonas's quote is selling the plumbing back to him at $312 a month, which works out to roughly $3,744 a year, for two hires. Once you have built the four workflows once, they run for every hire after that at close to zero marginal cost.

Isn't This Just Automated Bias?

Fair question, and the honest answer is: it can be, if you build it lazily. Ask a model to just "rate this resume" and it will happily invent enthusiasm it has no evidence for. The fix that matters is a constraint, not a bigger model: force the AI to return three sentences of reasoning tied to specific evidence from the resume, alongside the score. That single rule turns a black-box guess into an auditable decision a hiring manager can actually check before trusting it.

It also is not a compliance shortcut. In New York City, Local Law 144 already requires a bias audit before an automated tool can screen candidates for hire, and other jurisdictions are moving the same direction. Build the rubric, keep a human approving every final decision, and check your local rules before you point this at real applicants. The agent should make the shortlist faster. It should never make the hire.

The Four-Workflow Hiring Stack, Step by Step

Here is the actual shape of the system, based on a real sprint: a six-person agency hiring one developer and one ops role, 184 applicants, $1.47 in AI spend, 9 hours and 12 minutes of recruiter time reclaimed over three weeks.

1. Sourcing, no AI required. Three triggers feed one table: an inbox watcher for applications forwarded from LinkedIn and job boards, a webhook fired by your own careers page form, and a scheduled check of the niche boards you actually post to. A normalizer step converts all three into the same seven fields, name, email, role, resume link, cover letter, source, timestamp, and drops them into an Airtable base with status set to "new." This step is plumbing, not intelligence, and it is the one most tutorials skip because it is unglamorous.
2. Screening, where the AI earns its cost. A new row triggers the workflow. It pulls the resume, extracts the text, and sends it to Claude with the job description, a rubric of five weighted criteria, and the resume itself. The model returns a score from 0 to 100, a tier of A, B, or C, and three sentences of reasoning tied to specific resume evidence. The score and reasoning write back to the row, and every A-tier candidate posts to a private Slack channel within minutes of applying. On the real sprint, 184 applicants went in and 22 came back A-tier. The hiring manager only ever opened those 22.
3. Scheduling, the part that feels like it should be hard. When the manager approves an A-tier candidate with one click in Airtable, a webhook fires: it generates a booking link scoped to the manager's real calendar availability, drafts a warm, specific email naming the role and one detail pulled from the resume, and sends it. A second workflow watches the reply thread. If the candidate asks a question instead of booking, the AI drafts an answer for the manager to approve in one click rather than typing from scratch. A confirmed booking auto-creates the calendar event with the resume and screening notes already attached.
4. The nurture nobody builds and everybody should. Every B-tier or rejected candidate enters a fourteen-day sequence instead of a form-letter "no" and silence. Day one: a genuine rejection with specific feedback drawn from the screening reasoning, not a template. Day seven: a check-in with two other open roles that might fit. Day fourteen: an invitation to a small talent pool for future openings. On the real sprint, the client filled their junior role from this exact list six weeks later, at zero additional sourcing cost.

What It Actually Costs: SaaS vs Build-Your-Own

The SaaS platform is not lying about what it does. It is charging rent on plumbing you can lay yourself in a weekend, and once it's laid, it is yours to change without asking anyone's permission.

What You Have After One Hiring Round

A week in, you have stopped opening 184 resumes and started opening 22.

A month in, you have a rubric that gets sharper every time you tweak the prompt after a bad hire or a good one, and a talent pool quietly filling with the strong candidates you had no role for today.

A quarter in, hiring stops being the thing that eats your Fridays. It becomes four workflows that run themselves while you do the part that actually needs you: the interview, the offer, the decision. Once your new hire's paperwork is done, the next bottleneck is usually how you onboard them, and that is worth automating next.

Same team. Same budget. A completely different Tuesday.