From models to orchestration: How Moveworks built an enterprise AI platform that works

The opportunity: Bring consumer speed to enterprise AI

Shah was inspired to pivot after realizing that consumer technology was racing ahead while enterprise systems lagged far behind. “If I can hail a car in two minutes, why can’t I reset my multi-factor authentication just as fast at work?” he laughs.

In the era of Uber, DoorDash, and WhatsApp, Shah realized that enterprise employees expected the same fast, seamless experiences at work, but enterprise systems weren’t up to it. Reflecting on this broader shift toward short-form, real-time, convenient communications, Shah saw an opportunity to bring consumer-grade speed and simplicity to enterprise AI workflows.

“We became the front door for employees—one place to search, act, and get work done,” he says. The company’s AI agent assistant now integrates across Slack, Teams, and the web, using natural language understanding (NLU) to retrieve data, complete forms, reset passwords, or initiate workflows across enterprise systems such as SAP, ServiceNow, and Workday.

“We saw that gap and decided to close it,” he says.

The LLM inflection point

Before Moveworks pivoted, Shah and his team had already built domain-specific models for IT and HR that could resolve tickets and handle employee requests with precision. Then came ChatGPT in 2022, and large language models (LLMs) changed everything.

“Overnight, we could handle every question across every department,” he recalls. The LLMs gave Moveworks a single, generalized “AI brain” capable of understanding any function, any role, and any request. What started as an IT assistant was now becoming a unified intelligence layer that could access data and act across once-siloed systems using natural language. The Moveworks mission was no longer just about support, but transforming how work gets done.

It was the start of a new path for Shah, one with fresh challenges. “The LLMs became the brain—but we still needed to build the arms and legs that actually do the work across enterprise systems.”

The messy middle: AI infrastructure is the real “magic”

While many AI startups chase elegant use cases, Shah describes enterprise AI as “squarely in the messy category.” To him, success isn’t about building a single effective model, but making hundreds of systems talk to each other reliably.

In his view, this need for deep integration and orchestration is what separates enterprise AI from consumer AI. AI’s “magic” is rooted in the infrastructure that ensures consistent execution at scale, making systems run 24/7 across departments for thousands of users.

To achieve this, Moveworks built its platform from the inside out: custom AI agent-ready APIs, reliable data pipelines, and end-to-end AI integrations that reliably turn AI answers into trusted actions. Unlike plug-and-play AI tools suitable for consumer brands, enterprises demand consistency, reliability, and measurable outcomes, or procurement teams will call you out.

“If you can’t deliver value over and over again,” he says, “after a year or two, the procurement team comes to you and says, ‘Hey, you didn’t deliver what you promised.’” Initial delight only matters if the AI is reliable in production.

Product-market fit: Earning the right to scale

In the fast-moving world of AI, product-market fit is less of a milestone and more of a moving target for Shah. “You reach it for a moment, then everything from competitors, markets, expectations all changes,” he says.

He credits Moveworks’ traction to its consistent execution and clear metrics: win rates above 40%, sales reps closing at 110% of quota, and one in four new deals coming from referrals.

“Vision doesn’t grant scale, delivery does,” Shah says. “When your customers start referring you, that’s when you know you’re hitting real pain and solving for it.” It’s a reminder that in enterprise AI, success isn’t about hype or model size. It’s about having internal trust in results and employee uptake.

Weak vs. strong ROI

Shah draws a sharp line between incremental productivity gains and true AI transformation, or what he refers to as the difference between weak ROI and strong ROI.

“Weak ROI leaves when the employee leaves, taking their domain knowledge with them,” he explains. “Strong ROI changes how the business runs; it’s transformative at a deep level.” Research from McKinsey confirms that end-to-end workflow transformation is one of the strongest predictors of lasting ROI with enterprise technology, AI included.

In Shah’s view, most failed AI projects suffer from “DIY syndrome”—enterprises trying to piece together vector databases, custom APIs, and models without the necessary integrations or expertise. The result is brittle systems that don’t scale.

His advice is simple but pointed: build a strong foundation with trusted vendors and proven infrastructure, as is common in transformational periods. “You can’t Home Depot your way to enterprise AI,” he says. “You need the right platforms, the right partners, and the right level of execution.”

Frothy markets vs flexible foundations

In today’s AI economy, where startups can go from zero to $100M ARR in just 18 months, Shah acknowledges that speed matters, but so does staying power. “We’re in the era of popcorn growth: up fast, down faster, with few true winners.”

Some call it an AI bubble while others say it’s the new normal. For Shah, the strategy is the same: learn fast, move fast, and build for change ahead. “You learn in two years what used to take five,” he says. “The upside is, you find out a lot faster whether your idea is real.”

This mindset guided the Moveworks evolution. Rather than chasing every new trend, the company invested in flexible foundations for AI—prioritizing interchangeable foundation models, multimodal inputs, stable system interconnects, and consistent quality of service across integrations. In a volatile market, Shah argues, long-term winners are the ones who focus on durability, consistency, and compounding outcomes over short-term hype.

The importance of leadership, execution, and good sleep

Shah closes by saying that what truly scales AI isn’t just speed, but execution, reliability, and responsibility. The future of AI, he says, won’t be won by those who simply move and break things, but by companies that deliver reliably. His secret to being an effective leader and navigating Moveworks to a multibillion-dollar acquisition? Good sleep.

“If I sleep well, I’m a better father, husband, manager, and listener,” he says. It’s a metaphor for the philosophy that brought Moveworks to acquisition: in a world of complex and evolving technologies, the fundamentals of human systems—rest, clarity, execution—are still the backbone of progress and performance. The same focus on fundamentals is essential to building AI systems that are enterprise-ready.

Listen to the full episode of MindMakers to hear Shah’s and Moveworks’ playbook for building enterprise AI that performs—and what separates the short-lived AI tools from what truly sticks and scales.