Who builds AI agents for recruiting agencies?
Four kinds of providers build AI agents for recruiting agencies. How to choose between them, what drives cost, and what to ask before you sign.
AI agents for recruiting agencies are built by four kinds of providers: ATS and CRM vendors that add AI features to their own product, packaged AI recruiting products that plug into the tools an agency already uses, custom development firms such as Rubatt that build an agent around one agency's workflow, and in-house or freelance builders working with no-code agent tools. Which one fits depends on whether your bottleneck is a standard task, such as parsing resumes, or a process that only your desks run.
What are the four ways a recruiting agency can get an AI agent?
- Built into your ATS or CRM. What you get: AI features switched on inside the system your recruiters already use. Fits when: your need is a standard task such as parsing, matching or search. Watch for: features can depend on your plan and on the vendor's roadmap, and you cannot change how they reason.
- Packaged AI recruiting product. What you get: a ready-made agent for one job, such as sourcing, screening or scheduling, that connects to your ATS. Fits when: your process looks like most agencies' and you want a fast start. Watch for: pricing that grows with seats or volume, and limited control of prompts and logic.
- Custom development firm. What you get: an agent built around your brief format, your client rules and your systems, running in your accounts. Fits when: the work crosses several systems or follows rules no product covers. Watch for: you need a clear scope, an owner on your side and a way to measure the result.
- In-house or freelance build with no-code tools. What you get: a workflow assembled by your own staff or a contractor. Fits when: the task is small, internal and low risk. Watch for: testing, logging and maintenance often get skipped, and the knowledge can leave with one person.
What does "AI agent" mean on a recruiting desk?
An AI agent is software that reads incoming work, decides what the next step should be, and drafts or performs that step using your tools. An ATS (applicant tracking system) is the system an agency uses to hold candidates, roles and submissions.
On a recruiting desk, that usually means reading applications against a client brief, preparing a shortlist with reasons, drafting outreach, and proposing updates to the ATS. We cover the screening workflow itself in our guide to AI resume screening automation for recruiting agencies.
When is a custom build the wrong choice?
Often. Check these first:
- If your ATS plan already includes the feature you need, turn it on and measure it before buying anything else.
- If a packaged product covers your workflow at a price that works for your volume, buy it.
- If two recruiters would screen the same role differently, agree the process before anyone automates it.
Custom makes sense when the work crosses systems no single product connects, when your rules are specific to your clients, or when you want to own the logic instead of renting it.
What drives the cost of a custom recruiting agent?
We do not publish a flat price, because the same five things move the number on every project:
- Integrations. Each system the agent reads from or writes to adds work: the ATS, job boards, email, calendars.
- Write actions. Reading and drafting is cheaper to make safe than changing candidate records or sending messages.
- Evaluation. Testing on closed roles where you already know the outcome takes time, and it is the part that shows whether the agent works.
- Data handling. Candidate data is personal data. Where it is stored, who can see it and what a model provider receives all need design.
- Running costs. Model usage and hosting continue after the build and scale with application volume.
What should you ask any builder before you sign?
- Who owns the code, the prompts and the test set when the contract ends?
- Can the agent reject or advance a candidate without a recruiter approving it?
- Does each assessment show the evidence it relied on, or only a score?
- How was it tested on our past roles, and were results compared across groups of candidates?
- Is every action logged, and can we produce that log for a client or a candidate?
- Where does candidate data go, and which model provider sees it?
- Can we switch an integration off without a rebuild?
Who is responsible when the agent screens candidates?
In all three markets the agency using the tool carries responsibility, whoever built it. This is general information and not legal advice, so confirm your position with a qualified adviser.
US. Rules vary by state and city. New York City's Local Law 144 covers employment agencies as well as employers. The city's FAQ says "the vendor that created the AEDT is not responsible for a bias audit of the tool" and that employers and employment agencies "are ultimately responsible for ensuring a bias audit was done before using an AEDT" (an automated employment decision tool) (NYC DCWP, AEDT FAQ).
UK. The ICO says "both recruiters and AI providers have a responsibility for data protection compliance" and lists six questions to ask before procuring a tool, starting with "Have you completed a DPIA?" (a data protection impact assessment). It also says the controller and processor roles must be recorded clearly in a contract (ICO, key data protection considerations for AI in recruitment).
Canada. In Ontario, as of January 1, 2026, an employer that uses AI to screen, assess or select applicants for a publicly advertised job posting must say so in the posting. The province's guide says this "applies equally where the employer engages a third party (for example, a recruiting firm)". Employers with fewer than 25 employees are exempt (Ontario, requirements for publicly advertised job postings).
Whichever route you pick, ask the builder how the design supports these duties. A system that keeps a recruiter on every decision and records the evidence for each assessment is easier to explain to a client, a candidate or a regulator.
How does Rubatt build for recruiting and HR firms?
Rubatt Inc builds custom AI agents and workflow automation for small and mid-size businesses in the US, UK and Canada, including recruiting and HR firms. The agent reads and drafts, a person approves anything that can't be undone, and everything is logged. Clients own the code, prompts and evals. Our team is in Pakistan and works hours that overlap US, UK and Canadian business days.
We built an AI recruiting platform as our own product, where every assessment links back to the part of the application that produced it. The AI recruitment platform case study explains how, and our approval gates page shows how the human approval step is enforced.
If you have one recruiting workflow in mind, see custom agent development or book a call. We start with one workflow, ship it, measure it, then expand, and we will tell you plainly if your ATS or a packaged product is the better choice.
Questions
Who builds AI agents for recruiting agencies?
Four kinds of providers: ATS and CRM vendors that add AI features to their own product, packaged AI recruiting products that connect to an agency's existing tools, custom development firms such as Rubatt that build an agent around one agency's workflow, and in-house or freelance builders using no-code agent tools.
Should a recruiting agency build or buy an AI agent?
What drives the cost of a custom AI agent for a recruiting agency?
Is the vendor or the agency responsible for how an AI screening tool is used?
Who owns a custom recruiting agent after it is built?
Rubatt engineering
We build agent systems for operations, sales, research and support teams. Everything here comes from something that broke in production first. Reach us at hello@rubatt.com.
