The gap we kept hitting
We kept seeing the same pattern across our HR clients. Engineers wanted to ship AI features fast. HR leaders wanted full control over the rules. The two groups spoke different languages, and every change cost a week. Kiework needed a layer that both groups could live with. OpenClaw is what we built.
What OpenClaw is
OpenClaw is a small, opinionated workflow engine for AI flows. Every flow is a directed graph of steps. Each step is a model call, a tool call, a condition, or a human approval. HR leaders configure flows in a visual builder. Engineers extend the library of available steps. The same flow definition runs in dev, staging, and production.
Why we did not just use LangGraph
LangGraph is for engineers. OpenClaw is for product teams and HR analysts. We kept LangGraph for the deeply technical workflows and built OpenClaw for everything that needed configurability without code. The two coexist inside Kiework.
| Decision | OpenClaw choice |
|---|---|
| Who edits flows | Non-engineers via a visual builder |
| Where flows live | Versioned JSON in the database |
| How flows fail | Always with a human-readable reason |
| How flows recover | Resume from the last successful step |
| How flows are tested | Eval set per flow, run on every save |
Real use cases inside Kiework
- Leave approvals. Read the policy, check balances, draft a response, route to a manager if anything is ambiguous.
- Onboarding. Generate offer letters, schedule the first week, surface the right documents to sign.
- Compliance Q and A. Retrieve the relevant policy passages and answer with citations.
Lesson learned: The single biggest unlock was the human approval step. The HR team stopped fearing AI the day they realised they could put a human checkpoint anywhere in a flow with two clicks.
Where to learn more
Try Kiework at kiework.ai. For the more autonomous side of our agent work, see HermisAgent and Open Agent.
