RK.

Ramamoorthy Kumar · Product Design Leadership · Sprinto

Design systems are decision systems.

I build them for products, for AI, and for the teams that ship both. Eighteen years in, I still build them myself.

Receipts

At Sprinto, "done" only counts once it's verified. I hold my own claims to the same rule, so every number here carries a label.

Shipped live in the productMeasured tracked against a baselineIn progress being builtTarget a goal, not a result yet
5–6 hrs → 40 min
Setup time per customer, SprintoX.Every sub-25-employee onboarding now runs through it instead of a technical account manager.
Shipped
92%
Design-system coverage against a 90% target.NeoGrid across the product, with 51+ components and a shared token architecture.
Shipped
25%
Faster implementation at Sprinto with NeoGrid.Engineers build from shared components and tokens instead of one-off screens.
Measured
15+
Product pages running Ask AI.An entity-aware action layer, reused across the product instead of rebuilt page by page.
Shipped
12+
AI UX patterns with guardrails.Provenance, confidence and an escape path are required fields, not nice-to-haves.
Shipped
3–4 → 1
Steps to create an audit.Part of the simplification stream: small, visible fixes customers notice.
Shipped

Selected work

Each story is named for what I believe it proves. Each ends with what didn't work yet, because that's usually where the real decisions were.

SprintoX · AI-native onboarding

Turning compliance from a checklist into a state

Customers needed a person to walk them to audit readiness. We designed a system that sets itself up, keeps a living compliance state, and only interrupts people when its confidence drops. Setup went from 5–6 hours to 40 minutes.

Not yetEnterprise adoption hasn't followed. Accuracy and trust gaps are the reason, and they're named in the story.

Read the case study →

Ask AI · AI UX patterns

Conversation as the control plane

Most B2B AI is a chatbot bolted onto a dashboard. Ask AI is an action layer that knows which record you're on. It runs on 15+ pages and sits on 12+ patterns with built-in guardrails.

Not yetOutput quality is still being improved. The story shows how we measure it and what changes each round.

Read the case study →

Constitution · NeoGrid · RedPill · Datum · Forge

A design system is an organizational decision system

Five layers, from brand promise to executable code, with one rule between them. 92% coverage, 51+ components, 25% faster implementation, and the People area cut from seven tabs to three.

Not yetThe real test is an independent team reusing a pattern without us. That review runs in January 2027.

Read the case study →

Three streams · Ownership ladder · Builder OS

Designing a design org that works without me in the room

Eight designers, three streams with different mandates, and a ladder from owning a solution to owning a business outcome. Builder OS, which I co-created and orchestrated, let designers ship small fixes with AI coding tools.

Not yetBuilder OS first read as activity, not outcome. Some designers felt the ladder made the job "basically a PM role." Both changed how I frame the work.

Read the case study →

Earlier work: Sprouts.ai, Birdeye, Yubi, Chargebee, Ellucian →

How I lead

Most design leaders can show an org model or a design system. To me they're the same argument at two altitudes.

Read how I lead →

The promise

Product Design Constitution

Fourteen sections that turn brand into product law, including a written order for settling conflicts.

The material

NeoGrid

Tokens, components and accessibility contracts. The shipped foundation.

The work

RedPill

Patterns for agentic work, with a trust model as a required field.

The proof

Datum

A reference build. One source file generates five artifacts and fails the build on a bad value.

The lab

Forge

Where ideas get tested safely. Four promotion gates, and the default at each one is no.

One rule holds the stack together: a lower layer may specialize a higher one, but never silently contradict it.

The team

New products and AIBets that need room to fail
Core productCustomer commitments
Simplification and design systemsTrust, rebuilt in small wins

What each designer owns

L1A solution
L2A workflow
L3A problem
L4A business outcome

Before any significant work starts: what metric moves, what's the baseline today, and how will we measure it.

Workbench

I still build. These are things I make with my own hands, mostly so the team has something better to design against.

Open the workbench →

Prototypes in real code

Working React, not mockups

Multi-screen app flows, data tables with grouping and infinite scroll, and interactive prototypes for leadership reviews.

Systems you can run

Datum

A reference build that turns design-system rules into checks. Raw hex values and failed contrast break the build instead of slipping through.

AI that knows our rules

Skills and project instructions

A design-system skill so AI agents build with NeoGrid conventions, and project rituals that keep facts, guesses and decisions apart.

What people say

From people who managed me, worked beside me, and grew on my team.

An exceptional design leader who brings together strong creative vision, user empathy, and a deep understanding of how design drives business outcomes.

Goutham JoshiManaged RK at Sprouts.ai

He enabled me to move from IC work to leading pods. That space to grow was created by him.

Utkarsh BhandariDesigner on RK's team at Sprinto

The most skilled yet most humble designer I have known since 2017.

Prashant Kothari JainFirst PM hire at Chargebee

Field Notes

Essays that started as problems inside a build and turned out to travel.

All Field Notes →