Replace Your ATS: Build a Custom AI Recruiting Stack for $25/Month

What You’re Actually Paying For With a Commercial ATS

If you are running a business with fewer than 50 employees and actively paying for an applicant tracking system, there is a reasonable chance you are overspending by $1,200 or more per year. This guide covers exactly how to build a working ATS alternative using Airtable and Claude, what it costs, where it outperforms commercial tools, and where it does not.

Workable starts at $129/month. That is the starter tier. By the time you have added a second admin user and turned on any meaningful integrations, most SMBs land between $200 and $400/month. Annualized, that is $2,400 to $4,800 for software that, if you are a 15-person agency or a regional manufacturer, you use maybe 4 hours a month.

The feature list looks impressive: sourcing pipelines, structured interviews, compliance packs, analytics, LinkedIn scraper integrations. But walk through your actual workflow. You post a job, applications come in, someone reads the resumes, shortlists the strong candidates, schedules a call, and takes notes. That is it. You do not need a SOC 2-compliant talent CRM to run that process.

The Real Cost of Commercial ATS Tools

Tool	Monthly Cost	Annual Cost	Best For
Workable Starter	$129	$1,548	Teams with structured HR and 20+ hires/year
Workable Standard	$299+	$3,588+	Mid-market with multiple hiring managers
Greenhouse	$400-600	$4,800-7,200	Enterprise with compliance requirements
Custom Airtable + Claude	~$25	~$300	SMBs hiring 1-10 roles/year

Greenhouse pricing reflects estimates based on publicly available sales call ranges and third-party buyer reports; Greenhouse does not publish a standard pricing page, and exact contract costs vary by company size and negotiation (see references for available public sources).

The math is direct. A custom ATS replacement stack costs about $300/year. Workable costs $1,548 at minimum. That is $1,200 or more back in your operating budget, and the custom build screens candidates faster than a tired human reading through a pile of PDFs on a Friday afternoon.

Why SMBs Overpay for ATS Software

Enterprise ATS platforms are built to solve enterprise problems: multi-department hiring pipelines, EEOC audit trails, global compliance requirements, and structured interview scorecards with six approvers. Those features cost money to build and maintain, and that cost is distributed across every customer on the platform, including you.

When you are hiring a customer service manager for your 22-person e-commerce brand, you are subsidizing features built for companies 10 times your size. That is not a criticism of Workable or Greenhouse; they are well-engineered products. It is a structural mismatch between the product and your actual needs.

What the Custom ATS Replacement Stack Actually Looks Like

Three components handle the entire workflow. Airtable handles the database and interface. Claude handles the intelligence layer. A workflow automation tool like n8n or Make handles the connections between them.

Airtable as Your Candidate Database

Airtable ($20/month for a team plan) gives you a candidate table with custom fields: name, role applied for, resume text, LinkedIn URL, screening score, stage, notes, and hiring manager assignment. You build a form view for applications and either link to it from a careers page or embed it directly. Every application lands in one base, visible to anyone on your team who needs it.

This is not a workaround or a hack. Airtable is a legitimate relational database with a clean interface. The candidate pipeline view gives you Kanban-style stage tracking. The grid view gives you sortable, filterable data. You can build a hiring manager dashboard in about 30 minutes using filtered views and locked fields.

Claude as Your Screening Layer

Claude sits on top of the Airtable base via the Anthropic API. When a new application hits Airtable, an automation triggers a Claude API call that reads the resume text and scores it against a rubric you define: relevant experience, years in role, specific skills required, and any hard disqualifiers. Claude returns a numeric score, a two-to-three sentence candidate summary, and a flag for anything that eliminates the candidate immediately. That result writes back into the Airtable record automatically.

The whole screening step, which used to take 5 to 7 minutes per resume for a human reader, now runs in about 15 seconds. A 50-application batch that used to consume 4 to 5 hours of a Friday afternoon runs overnight and is waiting in your Airtable view by the following morning.

API cost per screening call: approximately $0.01 to $0.03 depending on resume length. Screening 200 candidates in a month costs roughly $4 in API usage. That is the entire intelligence budget for what would otherwise be a multi-hundred-dollar line item.

n8n or Make as the Automation Glue

n8n (self-hosted, $0) or Make (starts free at low volume) handles the trigger logic: new Airtable record created, pull the resume text and the role-specific screening prompt, send the payload to the Claude API, parse the JSON response, write the score and summary back to the Airtable record, and send an acknowledgment email to the candidate via whatever email tool you already use.

This automation takes about 2 hours to configure if you have used n8n before, and closer to 4 hours if you have not. The n8n community templates include Claude and Airtable nodes with documented setup guides.

How to Build This ATS Alternative in a Weekend

Step One: Build the Airtable Base

Create one table called Candidates with the following fields: Name (text), Email (email), Role (single select), Resume Text (long text), LinkedIn (URL), Screen Score (number, 1 to 10), Screen Summary (long text), Stage (single select with values: Applied, Screened, Phone Screen, Interview, Offer, Rejected), and Hiring Manager (linked record or single select depending on your team size).

Create a second table called Roles with fields for Role Title, Screening Prompt, Required Skills, Hard Disqualifiers, and Status (open or closed). Link the Candidates table to the Roles table so each candidate record connects to the job they applied for.

Create a Form view on the Candidates table that captures name, email, role applying for, and resume text. Ask applicants to paste resume text directly, or accept a file upload and use a parsing step to extract the text before it hits Airtable. Publish the form URL and add it to your careers page or job postings.

Step Two: Write the Screening Prompt

The quality of your screening output depends almost entirely on prompt quality. A generic prompt produces generic output. A specific, well-structured prompt produces actionable, consistent scoring.

A strong screening prompt follows this structure: “You are a hiring screener for [Company Name]. We are hiring for [Role Title]. The role requires [Skill A], [Skill B], and [Skill C]. Candidates must have [minimum requirement]. Hard disqualifiers include [Disqualifier A] and [Disqualifier B]. Score this resume from 1 to 10 on overall fit, write a 2-to-3 sentence summary of the candidate’s relevant background, and flag any hard disqualifiers. Return your response as JSON with keys: score, summary, disqualifier_flag, disqualifier_note. Resume text: [resume text].”

Store this prompt in the Roles table. When the automation fires, it pulls the prompt for the specific role the candidate applied to. This means your screening logic updates automatically when you edit the role record, with no changes needed to the automation workflow.

Step Three: Configure the Automation in n8n

In n8n, create a workflow with the following node sequence. First, a Trigger node set to fire on new record creation in your Airtable Candidates table. Second, an Airtable Read node that fetches the full candidate record and the linked Role record, including the screening prompt. Third, an HTTP Request node (or the native Claude node if available in your n8n version) that sends the prompt plus resume text to the Anthropic API endpoint. Fourth, a JSON Parse node that extracts the score, summary, and disqualifier fields from Claude’s response. Fifth, an Airtable Update node that writes the parsed values back to the candidate record. Sixth, an Email node that sends a plain-text acknowledgment to the candidate’s email address.

Test the workflow with five to ten sample resumes before going live. Calibrate the scoring rubric by reviewing Claude’s scores against your own assessment and adjusting the prompt until the outputs align with your judgment.

Step Four: Set Up the Hiring Manager View

Create a filtered Gallery or Kanban view in Airtable that shows only candidates with a Screen Score of 7 or above and a Stage of Screened. Lock the view so hiring managers see only the fields relevant to their decision: name, score, summary, LinkedIn, and a notes field. This gives them a clean shortlist without exposing raw resume text or automation metadata.

Add a simple interface layer using Airtable Interfaces (included in the team plan) if you want a more polished experience for non-technical hiring managers. You can build a one-page interface showing the shortlist, candidate summaries, and a button to move candidates to Phone Screen stage in under an hour.

Where This Stack Beats Workable and Where It Does Not

Where the Custom Stack Wins

The custom ATS replacement wins on cost, screening speed, flexibility, and data ownership. You can revise the scoring rubric in 5 minutes. You can add a field for salary expectation or portfolio URL without waiting on a product roadmap. You can export every record at any time without vendor lock-in. You control where the data lives and who can see it.

Screening consistency is a real, underappreciated advantage. A human reading resume 47 of 60 at the end of a hiring cycle is skimming and pattern-matching against fatigue. Claude reads resume 47 with the same attention as resume 1. Consistency does not guarantee accuracy, but it eliminates a specific category of human error that quietly damages hiring quality in small teams.

The total cost at steady state is approximately $20/month for Airtable, $5/month or less for Claude API usage at typical SMB hiring volumes, and $0 to $20/month for n8n or Make depending on whether you self-host. Call it $25 to $45/month depending on your setup choices.

Where Commercial ATS Tools Still Win

Workable and Greenhouse earn their price at specific thresholds. Native job board syndication (Indeed, LinkedIn, Glassdoor) from a single dashboard is genuinely useful if you are posting to multiple boards and want unified applicant tracking without manual de-duplication. The custom stack requires you to route all applications through your Airtable form, which means coordinating multiple job board source links.

Compliance tooling matters at scale. GDPR-compliant data handling, EEOC reporting, and structured audit trails are built into commercial ATS platforms and would require custom development to replicate properly. If you are in a regulated industry or approaching 50 employees, these features start to justify the price difference.

Multi-hiring-manager permission structures get complicated in Airtable above a certain team size. Airtable’s permission model is functional but not as granular as Workable’s role-based access controls. If you have five hiring managers who each need to see only their own pipeline with no cross-visibility, the custom build requires careful view and interface configuration to achieve that reliably.

The threshold is roughly 15 open roles simultaneously or 50 employees in the organization. Below that threshold, the custom ATS alternative handles the job cleanly.

Screening Quality and Prompt Engineering

The most common concern about AI-assisted screening is accuracy. The underlying question is whether Claude’s assessment of a resume is as reliable as a trained recruiter’s assessment. The honest answer is: it depends on the prompt, and a well-engineered prompt outperforms an average human reader on consistency, though not necessarily on intuitive pattern recognition for unusual candidates.

A well-designed screening prompt for this ATS replacement workflow should include five specific required skills in order of importance, two to three hard disqualifiers stated as explicit exclusion criteria, a description of what the top 20 percent of candidates at your company look like in this role, and the format specification for the response so the automation can parse it cleanly.

Invest 30 minutes writing a strong prompt for each role before you open applications. Review the first 10 outputs manually and compare them against your own read of the same resumes. Adjust the prompt until the scores and summaries match your judgment. After that calibration pass, you can trust the automated shortlist for most of your screening volume and reserve manual review for edge cases where the score is borderline.

One practical addition: add a field to your Candidates table called Prompt Version and log which version of the screening prompt was used for each candidate. This matters when you update the prompt mid-cycle, since scores from different prompt versions are not directly comparable.

Scaling the Stack as Your Hiring Volume Grows

The Airtable plus Claude stack handles up to roughly 500 applications per month without hitting meaningful performance or cost limits. At that volume, Claude API costs remain under $15/month for standard resume lengths, and Airtable’s record limits on the team plan accommodate several thousand candidate records per base.

If you grow to the point where you are running three or more simultaneous hiring campaigns across different departments, consider creating separate Airtable bases per department rather than a single shared base. This keeps view configurations clean and reduces the risk of hiring managers accidentally accessing records outside their pipeline.

For businesses that reach the point where compliance documentation becomes a real requirement, the custom stack can be extended. GDPR-compliant data deletion workflows can be built in n8n. Application audit logs can be maintained in a separate Airtable table with timestamp fields. These extensions take additional build time but remain far cheaper than upgrading to an enterprise ATS contract.

The Total Cost Picture

Running the numbers clearly: a Workable Standard plan at $299/month costs $3,588 per year. The custom ATS replacement stack at $25 to $45/month costs $300 to $540 per year. The difference is $3,000 to $3,200 annually, recurring, for a business that is hiring fewer than 15 people per year.

Over three years, that gap compounds to $9,000 or more in recovered operating budget, not counting the time saved on screening through automation. For a 20-person business where every dollar of overhead matters, that is a meaningful number.

The build investment is one weekend of focused work, plus one to two weeks of part-time refinement to get the screening prompt calibrated and the hiring manager interface polished. After that, the system runs with minimal maintenance: update the screening prompt when you open a new role, archive closed-role records quarterly, and review API usage monthly to confirm costs remain in range.

If you are currently paying more than $100/month for an ATS and hiring fewer than 15 people per year, the case for switching to a custom Airtable and Claude stack is strong. The tools are mature, the setup is documented, the costs are predictable, and the screening quality is competitive with what you are getting from your current platform.