Public work example
An onboarding automation proof of concept for a practical AI workflow
Onboarding often involves gathering information, coordinating people, and preparing repeatable next steps. This personal proof of concept explores whether a small AI-enabled workflow can assist with that work while keeping people responsible for review and decisions.
It is a public experiment, not a client case study, production system, or claim of a particular business result.
What it explores
Test the workflow before treating it as a solution.
A bounded workflow question
The proof of concept explores how an AI-enabled workflow might organize onboarding information and prepare useful next steps without treating the output as a final decision.
LangGraph orchestration
LangGraph coordinates the workflow steps so the experiment can make the sequence, handoffs, and review points explicit rather than hiding them in one prompt.
Local language-model experiment
The workflow runs a local language model. That makes it a useful way to examine cost and control tradeoffs, not a promise of lower cost or the right deployment choice for every team.
Build context and limits
Rapid exploration still needs careful follow-through.
I built this proof of concept in an evening using Codex and ChatGPT 5.6 Sol. The rapid build was useful for testing the workflow idea; it does not replace the design, evaluation, and operational work a production system would require.
A real implementation would need to define the data it may handle, integrations, security controls, quality measures, exception paths, and the points where a person reviews or approves consequential output. Some workflows will be better served by conventional automation or no new technology.
Have a workflow worth exploring?
Bring one real process, and we can discuss whether AI is a reasonable fit and what a sensible first step could be.