01
It reads
Kumo HR Agent 2.0 reads CVs against the job, contracts against labour law as a first-pass HR checklist, and policies against gaps. It is looking at the document, not summarising a filename.
Teams searching for an AI HR employee usually get a chatbot, a copilot, or another system of record. Kumo HR Agent 2.0 is none of those. It is an agent colleague: it reads CVs, contracts, and policies; sees each tool's real result and keeps going; pauses for your confirm on sensitive steps; and cannot claim work it has not finished.
“Agentic” means the software takes a goal, plans the work, runs the tools, looks at what actually came back, and continues the mission. That is the difference between asking a box for suggestions and hiring an AI HR employee who finishes the job.
Kumo HR Agent 2.0 reads CVs against the job, contracts against labour law as a first-pass HR checklist, and policies against gaps. It is looking at the document, not summarising a filename.
After every tool call it looks at the real result. If the last step returned something unexpected, the plan changes. This is a loop on live results, not a one-shot answer.
Give it a goal in plain language (a mission) and it plans the steps, runs them, and files the result as an artifact. You can still click around the HR platform. You just will not want to.
Sensitive actions wait for your confirm. An honesty kernel means it cannot claim unfinished work. Every step, tool call, and result is in the log.
Most “AI for HR” is a search bar, a draft box, or a copilot bolted onto an HRIS. A traditional HCM stores the record and waits. An agentic HR employee does the work, and shows you what it did.
Copilots wait for you to paste text. An agentic HR employee opens the file. Kumo HR Agent 2.0 is built to read the document, see what the last tool actually returned, and decide the next move.
Reads a CV against the job, not a keyword cloud. Screening stays grounded in the role you actually opened.
First-pass HR checklist against labour law. This is not legal certification and does not replace counsel. It flags what a people team should review next.
Reads policies against gaps, drafts updates, and keeps the handbook from going stale. You publish after sign-off.
Autonomy is earned on trusted, repeated kinds of work. Until then, and always on anything sensitive, the agent pauses. That is what makes an AI HR employee usable in a real people team, instead of a demo that writes confident fiction.
Sensitive steps wait. Nothing that should not ship, ships. You stay the decision-maker.
It cannot claim unfinished work. If a step did not complete, the log says so. No invented finish line.
Every prompt, retrieval, tool call, and approval is logged: searchable, explainable, and built on your private data, never trained on it.
Kumo is an agentic HR operating system: hiring, onboarding, payroll, leave, performance, policies, and compliance, with the same agent available over API and MCP if you want your own stack to drive it.
Missions, artifacts, skills, the 2.0 loop, and the honesty kernel. The full product explanation.
Open /intelligence →People, hiring, pay, performance, leave, learning, culture: one source of truth the agent colleague uses.
Open /platform →Current plans, credits, and what is included. No invented numbers here. See the live pricing page.
Open /pricing →The same capabilities, scoped and audited, for Claude, ChatGPT, Cursor, and your own agents.
Open /developers →30-minute demo · live product · your real questions.