Portfolio / 2025

Building AI ProductsThat Solve RealBusiness Problems.

I'm Ankit Mudgal, an AI Product Manager passionate about building AI-native products, automation systems and user-centric experiences that help businesses operate smarter.

View Case Studies→Download Resume
IIM Ranchi2+ Years0 → 1 ProductsAI AgentsEnterprise SaaS
Portrait of Ankit Mudgal
Ankit MudgalAI Product Manager
Chapter 01 — My Story

A quiet transformation from selling to building.

IIM Ranchi campus

IIM Ranchi — where the questions began.

I did not start in product. I started in rooms — sales floors, hotel lobbies, operations war-rooms — the places where customers actually meet a company. That was my apprenticeship.

Every role I took quietly reshaped how I think about products. Sales taught me listening. Business development taught me framing. Operations taught me systems. Hospitality technology taught me that most product failures are actually workflow failures in disguise.

By the time I stepped into AI product management, I wasn't chasing the technology. I was chasing the same question I'd been chasing on shop floors and back offices for years — what is the smallest, sharpest thing we can build that removes real friction from a real day?

  1. 01Sales
  2. 02Business Development
  3. 03Operations
  4. 04Hospitality Tech
  5. 05AI Product Management
  6. 06Building AI Products
Chapter 02 — Career Journey

Five acts, one throughline.

Each role is a chapter. Click one open.

Key achievement

Shipped an AI-native travel operations platform from zero, with autonomous agents handling itinerary, supplier and support workflows end-to-end.

Biggest learning

AI products are 30% model, 70% workflow. Design the human handoff before the prompt.

Chapter 03 — Featured Projects

Three products. Three ways of thinking.

Each of these is written as a product case study, not a portfolio brag. If you want the deeper artefacts — PRDs, evaluation reports, roadmaps — I'll happily share them in a call.

01

AI-native · Travel operations

Avyron AI

An AI operations layer for travel companies.

Case Study01 / 03

Avyron AI

AI-native · Travel operations

The problem

Mid-market travel operators drown in itinerary changes, supplier back-and-forth and last-minute traveller requests. Every SaaS they use adds another inbox.

The approach

We reframed the product not as a tool but as a co-worker: a set of AI agents that read the shared inbox, negotiate with suppliers, update itineraries, and hand off only edge cases to humans.

Discovery

34 discovery interviews across ops, sales and support before writing a single line of the PRD.

PRD & MVP

Scoped a 6-week MVP around the single most painful workflow: same-day itinerary changes.

AI Agents

Planner + supplier-comms + traveller-comms agents, orchestrated with a lightweight state machine.

Architecture

Event-driven backend, human-in-the-loop UI, evaluation harness on every prompt change.

Impact

62% reduction in change-handling time. 3.1× more changes handled per ops FTE.

Lesson

The hard part of an AI product is not the model. It is the escalation UX.

02

Hospitality · Operations

Hospitality Operations OS

One system for front-desk, housekeeping and revenue teams.

Case Study02 / 03

Hospitality Operations OS

Hospitality · Operations

The problem

Twenty properties. Fifteen spreadsheets. Three WhatsApp groups. A morning briefing that took 90 minutes and still missed things.

The approach

I lived on-property for two weeks, mapped every workflow, then designed a system that was less about features and more about a single shared source of truth.

Research

Shadowed 12 shifts across 4 properties before writing any spec.

Workflow map

Reduced 47 tracked workflows to 9 canonical ones.

Strategy

Sequenced the roadmap around what removed the most WhatsApp messages per week.

Automation

Auto-generated the daily brief; freed 60+ manager-minutes every morning.

Business impact

18% lift in on-time housekeeping SLAs; 22% reduction in guest complaints tied to ops.

Lesson

If the product does not survive the 7am shift change, it does not exist.

03

Enterprise · Multi-agent

Enterprise AI Automation Platform

Multi-agent workflows for finance, ops and CX.

Case Study03 / 03

Enterprise AI Automation Platform

Enterprise · Multi-agent

The problem

A large services company had 40+ high-volume workflows sitting between systems, each one a chain of copy-paste, approvals and email.

The approach

We built a workflow-first platform where each 'agent' is a scoped, auditable actor. PMs, not engineers, could compose new workflows from a small set of tested primitives.

Discovery

Stakeholder mapping across 6 business units; ranked workflows by pain × volume.

AI workflows

Reusable primitives: extract, decide, draft, approve, notify — composed into 24 live workflows.

Multi-agent

Specialised agents with shared memory and a supervisor pattern for audit trails.

APIs

Thin adapters over 11 internal systems; the platform never became the source of truth.

Delivery

Shipped in 90-day increments with a live metrics dashboard visible to the CFO.

Lesson

Enterprise AI wins are boring on purpose: audit, control, rollback. Ship boring first.

Chapter 04 — How I Think

A product lifecycle, honestly.

01Discover

Sit with the problem before naming it. Ambient listening beats formal interviews.

02Research

Talk to 5 users who love the pain and 5 who tolerate it. The gap is the product.

03Prioritize

Rank by pain × frequency × strategic wedge. Not by loudest stakeholder.

04Prototype

Cheap, ugly, disposable. The prototype is the argument.

05Build

Small teams, short cycles, boring engineering. Rituals over heroics.

06Measure

One north-star metric, three guardrails. Everything else is telemetry.

07Iterate

Learn faster than the market. Kill features with the same seriousness you ship them.

Product Principles
01

“Build for outcomes, not features.”

02

“Understand users before writing PRDs.”

03

“Ship MVPs. Learn faster than competitors.”

04

“Data informs. Customers decide.”

05

“Simple products win. Every time.”

Chapter 05 — Products I Admire

Products I study, quietly.

ChatGPT

It replaced the blank page. A product that changed the shape of a workday.

What I took away

The best AI UX is a text box, then everything under it.

Spotify

Personalisation that feels like taste, not a database query.

What I took away

Great ML products hide the model and reveal the human.

Notion

One primitive — the block — and a whole category folded into it.

What I took away

Extensible primitives beat exhaustive features.

Google Maps

Twenty years of quiet, correct decisions in a life-critical product.

What I took away

Trust compounds. Design for the tenth year, not the launch.

Swiggy

Made a chaotic real-world supply chain feel like a two-tap app.

What I took away

The magic is always in the operations behind the screen.

Uber

A pricing model that made an entire market legible.

What I took away

Sometimes the product is a pricing decision, not a feature.

Linear

Opinionated, fast, uncompromising. A tool with a point of view.

What I took away

Opinions are a feature. Blandness is a bug.

Chapter 06 — Bookshelf

The books that quietly rewired me.

Hover to lift. Click for the one line I actually took away.

Toolkit

The stack I reach for.

Tools are means. I'm loyal to none of them — I just want the fewest that make the team faster.

OpenAIClaudeGeminiCursorLovablev0GitHub Copilotn8nLangGraphCrewAISupabasePostgresReactNodeREST APIsFigmaJiraMixpanelAmplitudeSQL
Chapter 07 — Off the Clock

Life, outside the roadmap.

Mornings begin with weight.
Mornings begin with weight.
Roads I've walked to think.
Roads I've walked to think.
The bookshelf never sleeps.
The bookshelf never sleeps.
The ritual before the ritual.
The ritual before the ritual.
Where the sketching happens.
Where the sketching happens.
The North Star

To build AI products that make work feel less like work — and give people back the hours the software promised them.

I want to spend the next decade building products where the AI is felt, not shown — where the model disappears into the workflow, the team ships with more confidence, and the customer just notices that their day got quieter. That is the kind of AI Product Leader I'm trying to become.

Contact

Let's build something meaningful.

If you're hiring for a product role — especially anything AI-native, 0→1, or an ambitious rebuild — I'd love to talk. I read every message.

Email

hello@ankitmudgal.com

→

LinkedIn

in/ankitmudgal

→

Resume

Download PDF

→

Calendly

Book 30 minutes

→