Company

Oracle NetSuite

Time & Role

2025.08-2025.12 Lead UX Designer

From Code-First Deployment to Agentic Orchestration

Overview

Overview

For years, NetSuite’s developer-centric deployment flow left non-technical administrators without a safe way to move customizations across environments—pushing them toward risky workarounds and driving significant operational strain.

As UX Lead, I led the cross-functional transformation from a code-heavy flow to a guided, point-and-click experience, and later helped define its evolution into an AI-native, agentic workflow. This established a more scalable, confidence-driven model for enterprise deployment.

For years, NetSuite’s developer-centric deployment flow left non-technical administrators without a safe, intelligible way to move customizations across environments—pushing them toward risky workarounds and driving significant operational strain.

As UX Lead, I led the transition from code-heavy execution to a guided, point-and-click orchestration, and later evolved it into an AI-native, agentic workflow that proactively surfaces dependencies, explains risk, and guides resolution—transforming deployment from manual troubleshooting into a state of system-supported confidence.

Role

Role

End-to-end UX Design Lead

End-to-end UX Design Lead

Stakeholders

Stakeholders

Product Manager, UX Researcher, Developer, Tech Writer, Content Designer, Data Scientist, Consultants

Product Manager, UX Researcher, Developer, Tech Writer, Content Designer, Data Scientist, Consultants

Domain

Platform & Admin Experience, Developer Tools, Agentic Pattern, Human-in-the-loop,
Design System, Agent Orchestration

Platform & Admin Experience, Developer Tools, Agentic Pattern, Human-in-the-loop,
Design System, Agent Orchestration

Impact & Outcome

Impact & Outcome

Bridging the gap by delivering a new feature that empowers 100K+ admins to confidently manage deployments, aligning NetSuite’s functionality with user needs.

Bridging the gap by delivering a new feature that empowers 100K+ admins to confidently manage deployments, aligning NetSuite’s functionality with user needs.

70%

Adoption rate increased from 25% to 70% of the deployment tool.

↓80%

Deployment-related support tickets decreased at the root cause level.

400K+

Legacy bundles prevented from further misuse as deployment workarounds.

New Agentic Workflow

Established an agentic workflow pattern for NetSuite’s enterprise operations.

Context

Context

NetSuite is widely known for its extensibility. Customers tailor the platform by creating large volumes of custom objects—such as custom records, fields, workflows, and scripts—to fit their unique business processes.

To reduce risk, administrators and developers typically build and test these customizations in a sandbox environment, then deploy them to production once they’re ready. This sandbox → production movement—known as deployment—is a critical but high-risk operation: missing dependencies or configuration mismatches can block releases or cause production issues.

NetSuite is widely known for its extensibility. Customers tailor the platform by creating large volumes of custom objects—such as custom records, fields, workflows, and scripts—to fit their unique business processes.

To reduce risk, administrators and developers typically build and test these customizations in a sandbox environment, then deploy them to production once they’re ready. This sandbox → production movement—known as deployment—is a critical but high-risk operation: missing dependencies or configuration mismatches can block releases or cause production issues.

Target Users

Target Users

Historically, deployment was a developer’s game. But as we scaled into the SMB market, our primary users shifted from technical IT experts to 'Business-first' Administrators—often the company’s CEO or a operations manager. They are brilliant at their business but have zero desire to touch a Command Line Interface (CLI) or write code. They needed the power of an expert, but with an interface designed for a leader.

Historically, deployment was a developer’s game. But as we scaled into the SMB market, our primary users shifted from technical IT experts to 'Business-first' Administrators—often the company’s CEO or a operations manager. They are brilliant at their business but have zero desire to touch a Command Line Interface (CLI) or write code. They needed the power of an expert, but with an interface designed for a leader."

The initial design challenge (ambiguous)

HMWempowernon-technicaluserstomanagedeploymentsthroughanintuitivepoint-and-clickexperience?

Problem

Problem

Through research, we discovered that most non-tech users weren’t using NetSuite’s intended deployment flow at all. Instead, they hijacked an old packaging tool—Bundler—to push customizations between accounts. This workaround spawned 400 K+ “misused” bundles and drove 60% of deployment-related support tickets, flooding both data and support queues.


Why did a safer, intended deployment flow see low adoption—while a legacy tool became the default for real-world deployments?

Through research, we discovered that most non-tech users weren’t using NetSuite’s intended deployment flow at all. Instead, they hijacked an old packaging tool—Bundler—to push customizations between accounts. This workaround spawned 400 K+ “misused” bundles and drove 60% of deployment-related support tickets, flooding both data and support queues.


Why did a safer, intended deployment flow see low adoption—while a legacy tool became the default for real-world deployments?

User Research

User Research

To understand why users bypassed the intended Copy to Account flow and relied on Bundler instead, we conducted research with 30+ customers and internal NetSuite consultants （non-tech users + tech users) who regularly help customers resolve deployment issues.

To understand why users bypassed the intended Copy to Account flow and relied on Bundler instead, we conducted research with 30+ customers and internal NetSuite consultants （non-tech users + tech users) who regularly help customers resolve deployment issues.

Key Insights

Key Insights

Non-technical users didn’t just need a point-and-click flow—they needed a deployment system they could trust.


Research showed that low adoption wasn’t simply a usability problem. For many users, deployment felt like a black box: they couldn’t clearly see what was included, predict what might go wrong, or feel confident that following the workflow would lead to a safe outcome.

Non-technical users didn’t just need a point-and-click flow—they needed a deployment system they could trust.


Research showed that low adoption wasn’t simply a usability problem. For many users, deployment felt like a black box: they couldn’t clearly see what was included, predict what might go wrong, or feel confident that following the workflow would lead to a safe outcome.

Dependencies were nested, hidden, or locked

Dependencies were nested, hidden, or locked

Errors surfaced too late—often after deployment

Errors surfaced too late—often after deployment

Users lacked visibility into what would deploy, what might break, and why

Users lacked visibility into what would deploy, what might break, and why

The refined design challenge

HMWhelpnon-technicaladminsmanagedeploymentswithconfidencebymakingtheprocessvisible,predictable,andguided?

Design Principles

Design Principles

These insights shifted the challenge from creating a simpler UI to building trust in a complex workflow. I translated that reframing into three design principles.

These insights shifted the challenge from creating a simpler UI to building trust in a complex workflow. I translated that reframing into three design principles.

Make the system visible

Surface what the system is doing, why it matters —so complexity becomes understandable rather than opaque.

Reveal risk before commit

Help users anticipate uncertainty, consequences, and failure points before they commit—so they can act with greater confidence and control.

Guide users through complexity

Structure complex tasks into clear, manageable steps that reduce cognitive burden and help users move forward with clarity.

From Principles to Workflow

From Principles to Workflow

With these insights, I studied users’ common deployment workflows and began designing several deployment flows that better aligned with how admins plan, validate, and ship changes in practice.

With these insights, I studied users’ common deployment workflows and began designing several deployment flows that better aligned with how admins plan, validate, and ship changes in practice.

Guide users through complexity
Break complexity into manageable steps to reduce cognitive load
Reduce cognitive load and make progress feel safer.
Make the system visible
Review dependencies to understand impact
See ehat may be affected when selecting one object.
Reveal risk before commit
Validate early to avoid failure
Catch blockers before commit, not after deployment.

Final Solution
(Phase 1)
Point&Click Experience

Final Solution
(Phase 1)
Point&Click Experience

  1. A staged experience helps users move through complexity with confidence

  2. Catch blockers before deployment, not after failure

  3. Visible scope and status help users understand what they are deploying

The AI-First Shift:
Rethinking Deployment as an AI-Native Workflow

The AI-First Shift:
Rethinking Deployment as an AI-Native Workflow

In 2025, as the next generation of NetSuite embeds conversational AI and agentic workflows across the suite, I began rethinking how deployment could evolve from a guided UI into an AI-native, system-assisted experience.

we conducted research to understand administrators’ mental models of AI—how they perceive AI as an assistant versus an autonomous agent, where they expect control, and where they welcome system initiative. These insights shaped not only how AI shows up in the workflow, but when it should act and when humans should remain in the loop in our ERP.

In 2025, as the next generation of NetSuite embeds conversational AI and agentic workflows across the suite, I began rethinking how deployment could evolve from a guided UI into an AI-native, system-assisted experience.

we conducted research to understand administrators’ mental models of AI—how they perceive AI as an assistant versus an autonomous agent, where they expect control, and where they welcome system initiative. These insights shaped not only how AI shows up in the workflow, but when it should act and when humans should remain in the loop in our ERP.

Reel image

Defining AI-Assisted Interaction Model

Based on the workflow risks and what we learned from testing, I turned that into three interaction principles for this experience.

Interpretation, 
not just information

An agent should do more than surface raw system data—it should help users understand what matters, why it matters, and what deserves attention.

Trust through transparency

Agentic systems should make reasoning, system state, and potential consequences more visible, so users can build confidence instead of operating through a black box.

Human control at decision boundaries

The system can accelerate analysis and recommendation, but users should remain in control at high-impact moments—especially when committing, approving, or resolving risk.

I actively contributed to defining NetSuite’s emerging AI patterns—shaping how conversational interfaces and agent behaviors should work across complex enterprise workflows.

I actively contributed to defining NetSuite’s emerging AI patterns—shaping how conversational interfaces and agent behaviors should work across complex enterprise workflows.

Final Solution (Phase 2)
AI-first workflow

Final Solution (Phase 2)
AI-first workflow

To make deployment truly AI-first, I didn’t start by adding agents—I started by evaluating where agentic behavior actually made sense. Deployment is a high-risk workflow, so the goal wasn’t full automation by default, but intelligent collaboration between humans and the system.

By defining when agents should act, when they should assist, and when they should step back, the deployment flow shifts from manual execution to system-guided orchestration—reducing effort and uncertainty without sacrificing trust or accountability.

This pattern became a reference for NetSuite’s AI evolution and was presented at SuiteWorld 2025, with strong positive feedback from enterprise customers.

To make deployment truly AI-first, I didn’t start by adding agents—I started by evaluating where agentic behavior actually made sense. Deployment is a high-risk workflow, so the goal wasn’t full automation by default, but intelligent collaboration between humans and the system.

By defining when agents should act, when they should assist, and when they should step back, the deployment flow shifts from manual execution to system-guided orchestration—reducing effort and uncertainty without sacrificing trust or accountability.

This pattern became a reference for NetSuite’s AI evolution and was presented at SuiteWorld 2025, with strong positive feedback from enterprise customers.

Start deployment from intent, not manual setup

Start deployment from intent, not manual setup

After creating a custom object, users can simply express what they want to deploy through Ask Oracle. The deployment agent interprets intent, identifies the relevant objects and target account, and helps users begin the process without manual configuration.

After creating a custom object, users can simply express what they want to deploy through Ask Oracle. The deployment agent interprets intent, identifies the relevant objects and target account, and helps users begin the process without manual configuration.

Turn completed deployments into clear stakeholder updates

After deployment, AI generates a summary report that users can tailor for different audiences by selecting the relevant objects or details—making it easier to communicate what changed, what succeeded, and what needs follow-up.

After deployment, AI generates a summary report that users can tailor for different audiences by selecting the relevant objects or details—making it easier to communicate what changed, what succeeded, and what needs follow-up.

Resolve issues with guidance while keeping users in control

When errors or warnings occur, users stay in control while the AI provides contextual explanations and guided resolution paths—reducing trial-and-error without automating high-risk decisions.

When errors or warnings occur, users stay in control while the AI provides contextual explanations and guided resolution paths—reducing trial-and-error without automating high-risk decisions.

Moving Towards Agentic Workflow and Agent Orchestration

What began as an AI-first exploration gradually evolved into a broader vision for agentic workflow design. Deployment was a strong candidate because it is inherently multi-step, exception-prone, and risk-sensitive—making it a natural space for systems that can not only generate, but also reason, coordinate, and guide action over time.

My focus was not just on introducing AI into the flow, but on defining how users collaborate with it: what the system should surface, when it should act, how progress and issues become visible, and where human intervention remains essential.

Additional agentic workflow patterns are confidential—feel free to reach out to learn more.

Proactive Extension: Reimagining the IDE Experience

While my primary focus was the web-based deployment engine, I recognized a critical friction point for "power users" who spent their time in VS Code. I proactively proposed and led the UX optimization for the SuiteCloud IDE extension to ensure a unified deployment experience across the NetSuite ecosystem.

While my primary focus was the web-based deployment engine, I recognized a critical friction point for "power users" who spent their time in VS Code. I proactively proposed and led the UX optimization for the SuiteCloud IDE extension to ensure a unified deployment experience across the NetSuite ecosystem.

More Details Available Upon Request

Due to confidentiality, I’m unable to share further details publicly. I’d be happy to share more about the design approach, scope, and impact in a conversation.

Get in touch

More Details Available Upon Request

This page represents key highlights from my four years at Oracle.

Due to confidentiality, I’m unable to share further details publicly. I’d be happy to share more about the design approach, scope, and impact in a conversation.

Get in touch

Next project

Co-founding an AI-Native Fintech Venture

Defining Product & UX strategy for a multi-agent system