🧠 AI Employee Console

The company brain that makes every AI employee smarter

When a human resolves an ambiguity once, the system never asks again. That's the thesis. This demo performs it live.

---

The moat in AI employees isn't any single agent. It's the compounding knowledge layer underneath — the company brain — that makes every agent smarter with every deployment.

Built as a working product demo for Zamp's Product Manager role.

⚡ The Demo in 60 Seconds

1	Priya Sharma joins as a Sales Engineer in Dubai. The AI employee extracts her details, builds a cross-functional plan across IT, HR, and Finance — then stops. "Sales Engineer" maps to both Sales and Engineering. The agent refuses to guess.
2	The admin resolves it: Sales Engineer = Sales. One click. The decision cascades: IT access groups, finance tier, reporting line — all update. The resolution animates into the Company Brain as a persistent, queryable rule.
3	Omar Reyes joins as a Sales Engineer in Dubai. This time, no ambiguity. The Company Brain fires. Omar's plan resolves automatically — same IT groups, same finance tier, same reporting line. No human needed.

Corrected once, never asks again.

🚀 Run

pip install flask
python3 app.py

Open http://localhost:8080. No database, no Docker, no build step.

🏗 Architecture

├── app.py                  Flask server · 6 endpoints
├── agent_engine.py         Deterministic onboarding logic + ambiguity map
├── brain_store.py          JSON-persisted company brain · search & query
├── gpt_client.py           Templated LLM responses · preset chips are instant
│
├── static/
│   ├── index.html          Two-tab product shell
│   ├── styles.css          Linear/Vercel-inspired design system
│   └── app.js              Staggered beats · fork viz · constellation brain
│
├── data/
│   └── company_brain.json  Persistent decisions · auto-created · gitignored
│
└── research/
    ├── README.md                             Methodology & findings
    ├── research_summary.html                 Visual summary · 629 complaints
    ├── clustered_painpoints.json             11 pain clusters
    └── filtered_records_and_seed_clusters.json   Raw extracted records

🎯 Design Decisions

Decision	Why
Ambiguity escalation is deterministic	The backend checks a role-department map. The LLM provides language, not control flow. The hero moment never depends on model mood.
Brain is a real persistence layer	Decisions persist to JSON, survive restarts, and return grounded answers. Reset exists for clean recording takes.
Preset chips are templated	Three demo scenarios return instant, deterministic responses. Free-text falls through to the LLM. On camera, the product never waits.
Only two animations are dramatic	Knowledge created (fly-into-brain) and knowledge applied (rule pulse). Everything else stays calm so these moments land.

🔬 Research

Before building, I scraped 629 public complaints from Reddit and Trustpilot to map where Zamp's product has gaps.

The finding: every major ops pain point — AP, reconciliation, onboarding, vendor management — is already claimed. There is no process-level white space.

That's what led to the company brain thesis. The full research is in /research.

🧩 Full-Stack Workbench

This repo is the focused thesis demo. The expanded full-stack product version lives here:

Company Brain Workbench — approvals, audit trail, learned rules, and the Priya → Omar guided workflow in a production-style app shell.

💡 The Company Brain Thesis

Every time a human resolves an ambiguity an AI employee couldn't handle, that judgment becomes structured memory: what happened, what context mattered, what was decided, and when it should apply again.

Over time, this layer compounds. Each deployment makes the next one faster. Each correction eliminates a future escalation.

The moat isn't one agent — it's the accumulated decision intelligence of the organization.

---

Built by Karuna Sagar

Founder, Plato Tech — clinical & document AI agents shipped to hospitals