CASE 03 — AI-NATIVE PRODUCT, COMPLETE PRODUCT LIFECYCLE

CASE 03

Trove

Building an AI-native case study builder which is a reasoning partner, not just a case study generator for product designers targeting the best companies.

ROLE

Founding Product Designer • AI Builder

SERVICES

Complete Product Lifecycle across Research, Brand, Design,

Development and AI Systems

Complete Product Lifecycle across Research, Brand,

Design, Development and AI Systems

Complete Product Lifecycle across Research, Brand, Design, Development and AI Systems

YEAR

2026

OVERVIEW

Designers ship exceptional work but rarely have the bandwidth to document it at the level hiring panels expect. The tools that exist either don't understand what great case studies look like, or generate speed without substance. Trove is built to reason like a senior design director - one that knows what's strong, and helps you refine if it isn't.

Designers ship exceptional work but rarely have the bandwidth to document it at the level hiring panels expect. The tools that exist either don't understand what great case studies look like, or generate speed without substance. Trove is built to reason like a senior design director - one that knows what's strong, and helps you refine if it isn't.

01 — THE CHALLENGE

The best work never gets written up.

Designers ship exceptional work but rarely have the bandwidth to document it at the level hiring panels demand. As a result, portfolios go stale exactly when capability needs to be legible fast.


General AI assistants don't know what the best companies actually respond to. Case study generators optimize for speed, not depth. Neither closes the real gap — no tool is a reasoning partner that helps curate the best case studies.

02 — THE APPROACH

Building the Judgment Before the Product

I built the reasoning engine before the interface. Before any conversation flow existed, I designed a skill architecture — knowledge skills that define and hold all the knowledge, and capability skills which can execute a task by referencing a knowledge skill (or multiple knowledge skills).


On top of that, I designed three reasoning states: Support, Pushback, and Refine. So Trove could decide in real time whether a designer's answer deserved encouragement, a deeper question, or a challenge. The architecture came first because no interface would matter if the system underneath couldn't distinguish between strong and weak thinking.

03 — THE SOLUTION

A Reasoning Partner, Not just a Generator


Trove ships as two complementary modes. Guided Conversation asks one question at a time, responding in Support, Refine, or Pushback based on what the designer actually says - no forced flow, no fixed sequence. Brain Dump lets designers drop everything they already have: Files, Figma links, URLs, and notes onto a canvas, so Trove can synthesize that context and ask sharper questions from the start.

04 - The outcome

0 to 1 AI product

A good case study used to take a few weeks to put together — pulled together across scattered evenings, rewritten more than once as a designer second-guessed which details mattered. With Trove, that same depth can be curated in an hour or two — not because the standard dropped, but because the right questions get asked in the right order, the first time.

The product is live at trytrove.in

NEXT PROJECT

Stonecraft Group →

Let’s build something worth shipping.

Let’s build something worth shipping.

Let’s build something worth shipping.

Have a product that deserves clarity and craft? I’m always open to new collaborations.

Have a product that deserves clarity and craft? I’m always open to new collaborations.

Get in touch

Get in touch

Abhimanyu Singh Marwah

Abhimanyu Singh Marwah