Most companies use AI to answer questions. We built one that ships code.
Imagine saying in your team chat: "Add a dark mode toggle to the settings page." Minutes later, a Pull Request appears — with the implementation done, lint checks passed, and a clear summary of what changed. No IDE opened. No terminal touched.
That's Themis — an AI engineer we built in-house at PIPI.
Not a Tool. A Teammate.
The market is full of AI coding tools. Copilot autocompletes. Cursor chats in the editor. They're useful — but they all share the same assumption: the developer has to come to them.
We flipped that around.
Themis lives where your team already works — Teams, Telegram, Linear, GitHub. You @mention it in a group chat, tag it on an issue, or call it out in a PR review. It picks up the context, does the work, and delivers the result right back to you.
The shift is subtle but powerful: from "people finding tools" to "tools finding people."
A Brain That Knows When to Think Hard
Not every message needs a powerful AI model behind it. When someone says "good morning" or asks "what's our sprint goal?", you don't need a full-blown agent to respond.
So we designed Themis with two layers:
The fast layer handles simple conversations instantly — under one second. It's lightweight, low-cost, and covers about 80% of everyday team interactions.
The deep layer activates only when the request involves real engineering work: reading code, searching issues, generating implementations, reviewing PRs. When it kicks in, Themis connects to GitHub, Linear, and Sentry through a standardized protocol, assembling exactly the tools it needs for the task.
The user never sees the handoff. Themis just "thinks a moment longer" when the question is harder. No error messages. No "I can't do that." Just a smarter response.
Code Generation That Actually Ships
Themis doesn't just suggest code snippets. It runs a complete engineering pipeline:
Understands your project — It reads your project's conventions, architecture guidelines, and tech stack constraints before writing a single line.
Works like an engineer — It finds relevant files, reads surrounding context, and makes precise, targeted changes. We enforce a strict principle: partial changes are worse than no changes. If it touches something, it follows through completely.
Checks its own work — Automated lint and quality checks run before anything is pushed. Warnings are logged transparently so human reviewers can make the final call.
Delivers a clean PR — A well-structured Pull Request with a summary of changes, file list, and linked issues — ready for review.
Every task runs in an isolated environment, so parallel work never collides. If something fails, it cleans up after itself. Think CI/CD principles, applied to AI.
Five Workflows, One Platform
Themis isn't a one-trick tool. It covers the full development lifecycle:
| Workflow | What It Does |
|---|---|
| Code Generation | Turns issue descriptions into complete Pull Requests |
| PR Review | Reads diffs, existing discussions, and leaves context-aware review comments |
| Mentions | Responds to @mentions across Teams, Telegram, Linear, and GitHub |
| Bug Reports | Analyzes Sentry exceptions and generates structured bug reports |
| User Stories | Converts feature requests into actionable stories with acceptance criteria |
Every run is tracked — duration, cost, tools used. We know exactly what Themis does and what it costs, every time.
Talk to It — Literally
Themis supports voice. A product manager commuting to work can send a voice message: "Create an issue about optimizing the checkout flow." Themis transcribes, understands, executes, and replies — by voice.
This changes where and when teams can interact with AI. It's no longer tied to a keyboard and an IDE.
AI Building AI
Here's the part that still amazes us: Themis helped build itself.
Browse our git history and you'll see commits co-authored by AI throughout the project — from voice mode implementation to admin panel features to error handling improvements. The development cycle — generate code, review, iterate, merge — is exactly the workflow Themis was designed to enable.
On its very first day in production, a senior engineer submitted a PR to ensure human contributions are always tracked alongside AI's. That moment captured our philosophy perfectly: humans set direction and maintain quality; AI handles implementation at scale.
Why We Built This
PIPI is a hospitality tech company operating across Japan, Thailand, and Malaysia, managing over 1,000 properties. Our challenge was never "can we write the code?" — it was "how do we cover multiple products, markets, and tech stacks with a lean team?"
Themis is our answer: make AI a team-wide infrastructure layer, not just a programmer's personal tool.
When everyone on the team — product managers, designers, engineers — can trigger code generation, check project status, create issues, and review code through natural language on platforms they already use, AI stops being "a coder's assistant" and becomes a capability multiplier for the entire organization.
We're not building foundation models or general-purpose AI infrastructure. We're doing something more specific and, we believe, more impactful: embedding frontier AI deeply into how our team actually works — bridging tool silos with open protocols, balancing cost with intelligence through smart routing, and ensuring engineering-grade reliability through version control and human review.
Software is eating the world. AI is redefining how software gets built. Themis is our answer to that.
Published in The Pulse by PIPI Inc · February 2026