AI PM: Ship Your First Customer-Facing Agent
Move beyond internal demos and ship agents that survive contact with real customers — learn best practices for system design, agent ergonomics, evals, and monitoring.
Led by Bharat Batra
Why we built this
Anyone serious about AI in 2026 can build prototypes for internal users. Putting one in front of paying customers at scale is a different beast altogether. The failure modes multiply, the polish bar is much higher, and nobody is watching over the model's shoulder when things go wrong.
This course helps you take that leap. Together we will build a real customer-ready agent — scoped, polished, evaluated, and instrumented — using a working codebase crafted by Luminary instructors.
By demo day, you'll know how to elevate your product beyond "just a wrapper" and sharpen your value proposition so customers actually want to use it. And you'll know how to launch and iterate without burning trust — with guardrails, evals, and observability built in from day one.
This cohort is built for people who already have a working understanding of core AI concepts — LLMs, prompting, and agents. We'll do our best to stay fully inclusive regardless of background — but by enrolling, you're confirming you meet that bar.
Who we built it for
AI PMs & Builders
Learn how you can shape the product, not just the roadmap — by getting hands-on with agentic design principles. Develop the taste to know what's worth building, long before it ships.
Engineers & Tech leads
Learn software design patterns that help you build agents that stand the test of time. Master the core architecture and evaluation frameworks that separate real AI teams from vibe coders.
Founders & Operators
Learn to build the moat that keeps your agent from being cloned in a weekend. Go beyond the wrapper — scope, ship, and own the workflow that makes your product hard to replace.
What we'll cover
Skip the theory — week one starts inside a working customer-facing agent, taking it apart piece by piece. Learn what an agent is actually made of, then get started building your own.
- The anatomy : the core loop, tool calls, context, memory, and the system prompt — what each part does, where it breaks, and how to fix it
- The build-or-plug call: when your product needs its own agent — and when an MCP or skill inside a general assistant is enough
- Your takeaway: your agent v0 is running + you understand the frameworks to scope an agentic product
Capstone — Demo Day
Showcase your agent to a panel of experts; get actionable feedback that elevates your product experience to world class
Bonus Module — Background Agents
The future of AI products. Learn what differentiates a vanilla chatbot from a multi-purpose agent that can work while you sleep
Guest Lecture - Solving Agent Quality
Tactics from leading industry practitioners who think about this problem 24/7. Learn what's already used and in-flight at the best companies in the world.
How we'll interact
Live sessions
Fridays at 9:00 AM IST · 90 min each · all recorded for async catch-up.
Sep
Module 1
Sep
Module 2
Sep
Module 3
Sep
Module 4
Weekly office hours — drop-in Q&A and live feedback on your agent between sessions.
Private cohort community — ongoing questions, peer review, and shared resources, always on.
Peer network — build relationships with founders and builders shipping agents to customers.
Who you'll learn from
Bharat Batra
Product Leader at Superhuman AI
Bharat Batra is a product leader at Superhuman AI. He has spent years building products from zero to launch — learning firsthand what separates a product people rely on from one that quietly churns.
His conviction: with AI, the moat isn't features — it's craft. Getting code out is easy. Making the core loop work extremely well, and keeping it reliable in front of real users, is the grueling part most teams skip. In this cohort he teaches exactly that craft for agents you put in front of customers.
What people are saying
Nobody has taken this yet.
This is the founding cohort — there are no reviews to show, and that's the point. You'd be among the first through it, and the first to say what it was worth.
Everything else you need to know
No — you need to be willing to build. The cohort provides a working codebase and scaffolding; if you can follow a guided build and use AI tools honestly, you're equipped.
Cohort-Based Courses
Other programs you might like
Work closely over multiple weeks alongside real-world experts and dedicated peers.
AI Strategy: Optimizing Customer Operations
A strategy course for CX and operations leaders bringing AI into complex, high-volume customer operations.
Starts Sep 4 · Founding rate
Regular price $1,400Founding price $1,000