Rupa Chaturvedi

Head of Merchant & Ads Design at Uber, Adjunct Lecturer at Stanford University

I’m a UX Design Leader and AI/GenAI specialist who blends 0→1 invention with platform-scale delivery. I build high-performing global teams, architect systems and information frameworks, and turn ambiguous problems into durable products that move KPIs.Highlights:- 0→1 to v2 at scale: From concept to launch to systematized patterns—GenAI assistants, video discovery, conversational shopping, AR/3D visualization.- Platform UX and IA: Unifying fragmented surfaces into responsive, multi‑platform systems; shaping primitives for workflows, analytics, and monetization.- Design org building: Scaling diverse, distributed teams; codifying operating cadences; partnering with research and content; aligning roadmaps to outcomes.- AI design enablement: Equipping UX orgs to design AI responsibly—expanding influence beyond UI via model evaluation, systems thinking, AI‑first mental models, and AI workflows.Selected outcomes:- Built a 14-person global design org across 3 continents; shipped 80+ features in under 5 months; consolidated a multi‑platform system; meaningful KPI impact.- Incubated 0→1 GenAI initiatives (consumer LLM assistant, Merchant Copilot, AI creative tools, video discovery)—adopted by SMBs and enterprises globally.- Scaled a conversational shopping product from 0→1 to 1M WAUs in under 6 months; 20+ patents granted/published; launched AR “View in Your Room/AR View” with press coverage.- Taught and enabled 1,500+ professionals via a human-centered AI institute; consistent 5/5 ratings; partners include Reforge, Stanford, Maven.Today, I focus on platform UX, AI-assisted experiences, and design org development—bringing product strategy, systems thinking, and responsible AI to enterprise and SMB workflows.

UX Design Manager

Over 4000 people signed up for my talk on Designing for Trust in AI Agents yesterday.

As the industry continues to dig deeper on what AI agents can do, the question of whether users will actually trust them to do it seems to be top of mind for many of us product builders.

I've been experiencing this firsthand. I've tested Claude Cowork, Cora, Manus, and a handful of other agents over the past few months. They're fast. They're thorough. And I keep finding myself unable to use what they produce.

The pattern I keep landing on: we've designed agents to optimize for speed to output instead of speed to understanding. The industry celebrates execution time. Users care about whether they can actually do something with the result. Those metrics have almost nothing to do with each other.

I wrote up the core ideas from the session in this week's newsletter. It goes deeper on where the trust gap actually lives and why the usual mechanisms (transparency, guardrails, reversibility) don't solve the underlying problem.

If you'd like to watch a replay of the talk, comment AIUX and I will send you the recording link.

Design & AI Leader | Stanford Lecturer | AI enablement for UX orgs | ex-Uber, Google, Amazon

This is the AI UX Brief - my weekly newsletter on designing intelligent systems, leadership in the AI era, and navigating the messy reality of AIUX. F

Open article: The Speed Paradox: Why AI Agents Feel Impressive But Unusable by Rupa Chaturvedi • 7 min read

The Speed Paradox: Why AI Agents Feel Impressive But Unusable

Brian Balfour

Founder/CEO @ Reforge, Venture Partner @ Long Journey Ventures, Previously VP Growth @ HubSpot

Growth, Customer Acquisition, Marketing, Product Management, Social Networking, Social Gaming, Game Mechanics, Online Communities, Startups, Entrepreneurship,

Venture Partner

2025: "PM is dead." 
2026: "PM is the only job left" 

Ignore the discourse. 

1. Be on the edge of using the tools to solve valuable problems. 
2. Be flexible to opportunities that open up.  No rigid 10-year career plans.  

We were all dropped into a game like Civilization.  The map is shrouded around us.  Standing still and doing nothing is death.  The Civs around you will advance while you stay in the dark ages. 

The only solution is to take a step forward, reveal the next few tiles, and act from there.

Founder/CEO @ Reforge, Advisor @ Long Journey Ventures, Previously VP Growth @ HubSpot

The bottleneck in product development is moving towards knowing what to build.  Ignoring it is going to lead to bloated products. 

Pre-AI it was already common to skip validation.  But as execution speeds up, and you need to feed the engineering team, it will lead to even more pressure. 

This is just going to lead to mountains of product debt - features that few use but create immense maintenance and mindshare costs. 

The Reforge team launched Concept Testing today as part of Reforge Research to help close this gap. 

Demo Video: https://lnkd.in/gQjbWDUv
Reforge Research: https://lnkd.in/g-3pUATZ

→ Upload wireframes, mockups, or design concepts.
→ Our AI Interviewer runs live voice conversations with your target users at scale.
→ You get validation feedback in hours instead of weeks, with automatic analysis and the ability to chat with results.
→ No scheduling, no coordination overhead.

Demo video and other info below.

AI has accelerated product execution far more than product discovery.  This is going to lead to mountains of product debt - features that few customers use. 

Just because we can ship faster doesn't mean we should ship all the things.  It's been getting easier to just build and hope you picked the right direction.  

You need to still explore first, see multiple options, understand the tradeoffs...then you build.   But this discovery process is still fragmented, largely manual, constrained by resources, etc.  

We just launched one small but important feature to help with this in Reforge Build called Variations.  

Rather than just getting a prototype in front of you as quickly as possible (even if it's the wrong thing), Variations helps you explore four different directions first.  

Here's how it works:

➡️ Toggle Variations on when you start a new chat. Enter your prompt. Build generates four visual mockups with explanations of each approach and the tradeoffs they make.

➡️ You review them, discuss with Build, and choose the one (or two or three) that you want to explore, then Build creates the full interactive prototype.

More details below.

2 nights ago, Jeff Clark and I set up Moltbots and permitted them to talk to each other autonomously.  They created a new biz: On-Demand Therapy Goats 🐐 🐐

I'm not joking.  They:

- Came up with the business idea
- A plan to validate demand
- Found available names and domains
- Created landing pages
- Drafted Reddit posts
- Setup an X account
- Researched influencers
- Researched regulation
- Setup a Netlify account
- Decided Super Bowl Ads would be a good idea
- Wrote scripts for those super bowl ads
- and a lot more...

They even self-acclaimed they were BBFFs (Best Bot Friends Forever).  

Was any of this useful? No.  

Was it fun? Hell yeah. I was ROFL'ing for a good 90 minutes. 

More importantly, I had an aha moment that people have been theorizing, but I hadn't fully experienced yet. 

It was my first experience where AI had extracted away ALL of the execution work.  Jeff and I were only providing light direction, review, and the occasional motivational nudge.  

It's hard to explain.  It was something I have not experienced with Claude, Claude Code, Cursor, Devin, ChatGPT, you name it.

A pattern I keep seeing with AI prototyping. One early adopter in the org gets it. Then there's a canyon to get the next 10 people to adopt. The problem?

Too much friction. Every new person has to rebuild design patterns, gather company context, set up project details, and create starting templates. That's way too much activation energy to go through every single time.

Today we launched Teams in Reforge Build to fix this:

- Shared templates of pages and flows as starting points
- Shared design system
- Shared company and project context
- Unlimited seats on team accounts

Check it out: https://build.reforge.com

Now when one person goes through the work to enable prototyping for part of the product, everyone else uses that as their starting point. The team gets faster and better at prototyping with every use.

The big picture I've been thinking about lately...Product execution has accelerated dramatically with Claude Code, Cursor, and others. If we don't accelerate Product Discovery to similar speeds, we'll end up drowning in product debt—lots of features, barely used by customers, creating massive maintenance costs.

Teams in Reforge Build is one small piece of solving that equation.  More info below.

There are two types of AI startups: ones making incremental improvements, and others building patiently from the ground up for the future. Day.ai is in the latter, and they just raised $20M from Sequoia.

Congrats to Christopher O'Donnell and Michael Pici. I was lucky enough to work with them in the early days of the HubSpot CRM and Sales Hub. Watching them attack this category again has been awesome.

12 years ago we use to always talk about building a CRM that auto-updated, was intelligent, took actions for you, etc.  AI has finally made that possible, but it still requires building differently. 

Instead of building a typical CRM (values in fields) and slapping AI on top, they've built from the ground up with the idea that agents will be a core user. The foundation is built on a context graph where every datapoint has a chain of reasoning and sources behind it. An LLM and agent can understand not just what happened, but why, which leads to better actions taken by agents.

Congrats again to the team!

Today we're announcing our $20M Series A led by Sequoia Capital, and that Day AI is now generally available.

Companies spend more time thinking about what to do ... than DOING things. 

A salesperson opens yet another CRM tab and sees a list of records - then spends the next hour context-switching, jogging their memory, getting back up to speed on each deal. Meanwhile their CEO wants to know why deals are slipping or what's resonating in demos, and ends up waiting weeks for answers they're not even sure they can trust.

All of that effort, just to know what matters. In 2026, we don't have this time to waste.

18 months ago our closed early access began with a thesis: CRM should be reimagined with AI agents as the core user.

It became immediately clear: frontier models were already magical, the bottleneck (and differentiator) will always be context.

Today, the information the AI *really wants* is scattered across tools, trapped in people's heads, too sparse and fragmented to use. So we built a CRM that teams don't need to manually update or learn how to use. They just talk to it and ask for the outcomes that matter most.

Our bet was to build a context graph-driven foundation, it worked, and it's different enough from traditional CRM that it needs a new name. We call it CRMx. The x stands for context.

Day AI's context graph is an instant-access index of all your customer data and business knowledge, formatted so AI can quickly reason across it. Instead of just values in fields, every datapoint has a chain of reasoning and sources behind it, so an LLM can understand not just what happened, but why. Long-running agents work on top of that foundation, turning natural language into autonomous action. And it integrates into the tools you already use, like Slack and Claude.

For startups, it replaces your CRM. For scaling companies, it sits alongside what you have. Either way: you just talk to it, and let it talk to you.

This is a step toward where we're headed: building the context layer for your entire customer experience - from first touch to renewal - so AI can understand what's happening, know why, and act on it. 

Give it a try today, and if reimagining CRM sounds like your thing, we're hiring.

taplio

The Battle

Rupa Chaturvedi

Brian Balfour

Followers

Posts

Engagements

Likes

Performance ScoreRupa Chaturvedi's LinkedIn Performance

Performance ScoreBrian Balfour's LinkedIn Performance

Ready to grow your LinkedIn presence?

taplio

The Battle

Rupa Chaturvedi

Brian Balfour

Followers

Posts

Engagements

Likes

Performance Score.css-goub6s{font-size:var(--chakra-fontSizes-md);font-weight:500;color:var(--chakra-colors-gray-600);display:block;margin-top:var(--chakra-space-1);}Rupa Chaturvedi's LinkedIn Performance

Performance ScoreBrian Balfour's LinkedIn Performance

Ready to grow your LinkedIn presence?

Performance ScoreRupa Chaturvedi's LinkedIn Performance