Agentic Demos · HR Processes

What happens when an AI agent shows up to do real HR work?

Four end-to-end demonstrations of agentic AI applied to everyday HR processes — from the year-end performance management cycle to HRBP advisory work, web-form data entry, and shared-drive cleanup. Each card records what the agent did, the capability being showcased, how long the task took, and the deliverables it produced.

See the Demos Browse Artifacts on GitHub

Theme 01

Document Analysis

2 demos

Agents that read across many document types — Word, PDF, Excel, PowerPoint — and either extract specific evidence on a defined target, or aggregate and synthesize what they find against a question the user posed.

Document Analysis · 01

MRE Artifact Review

Agent runtime

~10 minutes

Process · Annual performance management review

Part of the year-end performance management cycle for Material Risk Employees. Each MRE submits a folder of evidence and a reviewer has to walk through it and confirm that the artifacts actually substantiate material risk activity. The agent showcases the ability to search through and reliably extract specific content from a variety of document types — in this run it walked 75 employee folders containing 175 artifacts across PDF, DOCX, XLSX, and PPTX, applied a defined risk taxonomy, captured verbatim excerpts with location references for every classification, and routed borderline cases to a supervisor queue rather than auto-deciding.

multi-format extraction targeted content search taxonomy classification evidence excerpting

Deliverables produced by the agent

MRE Review Log XLSX Supervisor Report DOCX Run Summary MD

Document Analysis · 02

Strategic Alignment Check

Agent runtime

< 5 minutes

Process · HRBP advisory consult

An HRBP-style consulting exercise. Given a direction from the user — "tell me where stated strategy and observed reality are diverging" — the agent demonstrates the ability to aggregate, synthesize, and analyze across a variety of document sources. In this run it ingested 14 source documents (4 strategy plans, 3 executive reports, 4 performance decks, plus supporting material — a mix of Word, PDF, and PowerPoint) and reconciled them against 8 quantitative workforce CSVs. For each line of business named in the documents the agent inferred the stated direction, compared it against observed deltas in headcount, hiring, attrition, and engagement, and assembled severity-scored findings into a one-page executive brief and a multi-tab detail workbook.

cross-format aggregation cross-source synthesis directed analysis brief authoring

Deliverables produced by the agent

Executive Brief PDF Detail Workbook XLSX Run Summary MD

Theme 02

Workflow Automation

2 demos

Agents that don't just read — they act. They drive web forms from a list, restructure messy drives, and leave behind indexes so the next agent doesn't have to start over.

Workflow Automation · 01

Web Page Submission

Agent runtime

< 10 minutes

Process · HR ops · provisioning intake

A demonstration of how agentic AI can input information directly into webforms based on lists and other instructions. The agent was given a 20-record roster CSV and a target HR platform, mapped each column to the right form field — including dropdowns, equipment checkboxes, service checkboxes, priority, approver, and delivery address — submitted a controlled test order first to validate the field mapping, then iterated through the remaining 19 records. All 20 of 20 records were filed end-to-end without manual intervention; each issued reference number landed in an audit log alongside any anomalies worth flagging (priority escalations, international shipping addresses, request-type mismatches).

list-driven input field mapping controlled test-then-batch audit log capture

Deliverables produced by the agent

Submission Audit Log CSV Run Summary MD

Workflow Automation · 02

Folder Content Analysis, Organization, and Indexing

Agent runtime

~20 minutes

Process · Knowledge management · drive cleanup

Agentic AI applied to a messy shared drive. The agent analyzed diverse drive contents, recommended a folder structure, implemented the structure, and produced an index ready for a future agent to use — so the next AI tasked with answering questions against this drive doesn't have to re-analyze every document from scratch. In this run it inventoried 62 files spanning strategy documents, performance decks, executive reports, workforce CSVs, installable analytics skills, and assorted skill outputs. The agent characterized every file, proposed three reorganization options with trade-offs, executed the approved option, and emitted both a machine-readable file index (lifecycle stage, domain, schema, key topics, example questions per file) and a human-readable data dictionary.

drive inventory reorganization planning structure execution agent-ready indexing

Deliverables produced by the agent

Folder Review MD Agent File Index JSON Top-level README MD Run Summary MD

Disclosure

What you're looking at

These demos run on synthetic and public data. None of the deliverables represent actual organizations, employees, or compensation. The point of each demo is the agentic capability — the work an AI did to get from input to output — not the contents of the fictional dataset it produced findings about. Methods are documented; everything can be inspected.

Governance Charter All Artifacts on GitHub