AI agents for HR that answer employees and run onboarding inside your access rules
AI agents for HR are software in which a language model handles routine people operations work: answering policy and benefits questions from your own documents, running onboarding checklists across systems, and preparing cases for HR staff to decide. Goldenroz builds them on your HRIS and document stores with per-user permissions, masking of personal data and a full audit log. Projects start at $25,000, fixed price per phase after a free one-hour scoping call.
From $25,000, fixed price per phase
Updated 2026-11-14. By Shai Goldenberg
What usually goes wrong
- The HR inbox repeats itselfHow much leave do I have, where is the parental policy, who approves a relocation. A small team answers the same questions by hand, in several time zones.
- Open enrollment floods the teamFor a few weeks each year benefits questions multiply. The answers are in plan documents nobody reads, and every employee's situation differs slightly.
- Onboarding spans five systemsA new hire needs an HRIS record, payroll setup, accounts, equipment and training assignments. One missed step means a first day without a laptop.
- Policies vary by locationLeave, overtime and notice rules differ by state and country. A generic assistant gives the headquarters answer to an employee it does not apply to.
What is included
- Policy assistantAnswers from your handbook, benefits documents and local policy addenda, with the source passage cited and the employee's location and employment type applied.
- HRIS connectionRead access to systems such as Workday, BambooHR or ADP through their APIs, so an employee can see their own balances and records and nobody else's.
- Onboarding and offboarding runsA checklist the agent drives across HR, IT and payroll systems, with each step logged and exceptions sent to a named person.
- Privacy controlsPer-user permissions, masking of sensitive fields before they reach the model, and rules for which categories of data are never sent at all.
- Escalation and audit logSensitive topics such as complaints, health and accommodation requests route straight to HR staff. Every question, lookup and answer is recorded for review.
AI for HR: service tasks first, decisions never
The safe starting point with AI for HR is employee service: questions with an answer in a document or a record. These are high in volume, low in risk when a person can be reached, and easy to measure. Case preparation is the second tier, where the agent gathers the relevant policy and history and a human HR partner decides.
Decisions about people are a different category. Several jurisdictions regulate automated tools in employment. New York City's Local Law 144 requires a bias audit and candidate notice for automated employment decision tools, the EU AI Act treats many employment uses as high-risk, and US anti-discrimination law applies to an employer whatever tool it uses. We do not build agents that screen, rank or reject candidates or employees. Goldenroz is an engineering firm, not a law firm or a bias auditor, and your counsel should review any use that touches employment decisions.
Employee data and who can see what
HR data is the most sensitive data most companies hold, so access design comes before prompts. Each request runs as the person asking. An employee sees their own record, a manager sees their direct reports, and HR staff see what their role allows in the source system. The agent uses approved queries and never writes its own.
Shai, our CTO, scopes and builds the agent using the practices on the AI agent development page. His on-prem MCP gateway product was built around the same controls: read-only access, per-user permissions, PII masking, an audit log and prompt-injection defenses. Your HR, legal and security leads review the permission model and the excluded data categories before production data is connected.
How the work runs
Scoping call
Before the projectWe map the process, the systems involved and the cost of a wrong action. You get a fixed price for the first phase.
Prototype
About 1 to 2 weeksA working agent on one narrow task, run against real examples in a test environment. You see where it fails before you pay for production.
Production build
Roughly 4 to 8 weeksIntegrations, permissions, fallbacks, approval steps, logging and the evaluation set. The range depends on how many systems the agent connects to.
Launch with limits
At launchThe agent starts on a share of the volume, or with every action approved by a person. Limits widen as the logs show it holds.
What it costs
Every phase has a fixed price agreed before work starts. Model usage is paid to the model provider from your own account.
- Focused agent or agentic workflowFixed price per phaseFrom $25,000
- Larger AI platformsSeveral agents, many integrations or a full product$100,000 and up
- Scoping callEnds with a written scope and a price$0
- Model usageBilled to your own accountPaid to the provider
Who it is not for
- Automated screening, ranking or rejection of candidates or employees. We do not build that.
- Legal or compliance advice on employment law. We build the technical controls and your counsel decides what is permitted.
Who does the work
Shai leads the technology side: business software, SaaS and mobile apps, payments and escrow systems, Postgres security, AI agents built with Claude Code and MCP, and IT and infrastructure projects with senior infrastructure engineers.
Questions buyers ask
What can AI agents for HR do safely?
They answer policy and benefits questions from your own documents, show employees their own balances and records, run onboarding and offboarding checklists across systems, and prepare cases for HR staff. Decisions about hiring, pay, promotion and termination stay with people.
How much does it cost to build an AI agent for HR?
Projects start at $25,000 and are fixed price per phase after a free one-hour scoping call. Larger programs across several countries and systems run $100,000 and up. Model usage is paid to the provider from your own account.
Do you build AI for HR recruiting, such as resume screening?
No. Automated screening and ranking of candidates carries legal obligations in several jurisdictions, including bias audits and notices, and the cost of a wrong output is high. We build scheduling, candidate communication and document preparation, where a person makes the decision.
Does employee data get sent to the model provider?
Only the minimum needed to answer a specific question, after masking, under your own agreement with the provider. The integration layer runs on your infrastructure and every call is logged.
Which HR systems does HR automation connect to?
Systems with an API, such as Workday, BambooHR and ADP, can be connected, and the scoping call confirms what your tenant and plan expose. Document stores for the handbook and policies are connected read-only. Systems with no API get a custom connector scoped as its own phase.
Guides on this topic
- Build a working MCP server in TypeScript with the official SDK, then add what production needs: permissions, PII masking, audit logs. Follow the steps.
- MCP does not replace REST APIs. It wraps them so AI applications can discover and call them. Compare MCP vs API and see when an MCP server is worth building.
- MCP is an open standard that connects AI applications to your tools and data. Learn how hosts, clients and MCP servers work, and what to secure first.
Talk to the person who will do the work
One hour, no sales team. You talk to Shai Goldenberg, the CTO, about your systems or your product, and you leave with a clear next step and a price range.