SaaS, Developer platform, AI agents
From months to minutes: one self-service platform for Dell's engineers. A SaaS platform where Dell's engineers find what they need to start a new initiative.

- Role
- Senior Product Designer
- Sector
- IT Industry / Manufacturing
- Timeline
- 3 years
- Platform
- Web
Where every initiative starts
Engineers at Dell lost weeks or months before a new initiative could start. Information was spread across several portals, and every answer began with a request. I was the main designer of a SaaS platform that became their single starting point. It brought one experience and one login to the business unit's 7 applications, search across all of them, capabilities other teams use instead of building their own, and an AI assistant for access and setup. Engineers now do 80% of provisioning themselves, in minutes instead of months.
Dell invests heavily in new technology, so engineers often need new services, tools and infrastructure to test an idea. Before writing any code, they first had to find out what existed, who owned it and how to get access.
- For engineers, a new initiative started with weeks or months of waiting: for information, for the right people and for tickets.
- Information about resources, capabilities and infrastructure was outdated and spread across several portals. Every answer started with a request, and every request slowed an initiative down.
- Each application had its own login, so engineers signed in again and again, and every change to login had to be made in each application separately.
We interviewed 30 engineers, architects, directors and managers, and analysed support records to find the paths that cost the most time.

One trusted place to start
- User need
- Find what exists, where it lives and how to get it, in one place I can trust.
- Business requirement
- Stop every answer from starting with a request to another team.
Capabilities teams don't have to build
- User need
- Get workflows, content and notifications for my product without building them.
- Business requirement
- Stop teams duplicating the same features.
Help with access and setup
- User need
- Get help to access services and provision correctly, without waiting on someone.
- Business requirement
- Increase the pace at which engineers deliver.
Every application had its own look, navigation and login, so engineers relearned the interface each time. I designed one experience: a shared header, look and feel, search and a single sign-in that carries across connected products.
My first attempt was a shelf of links to other applications. Engineers barely used it and asked for it in the header. Teams then resisted losing control of their headers, so I designed one with an area each team controls.
It runs on Dell's existing design system. I worked with its team to improve it, including a dark mode for long sessions.
In the first phase, all 7 applications in the business unit adopted the shared header, look and feel, search and login.

Answers lived in many sources and in many types of content. We studied every type of content the search would return and grouped them under labels. I designed tabs for the core types, so engineers could narrow results, with results ranked by how closely they matched the query.
Adobe Analytics showed engineers reaching applications they hadn't visited before.

Instead of letting each team build its own, we offered the platform's features as a service: a workflow builder, a CMS and a notification centre that other teams use in their own products.
The workflow builder had to fit many teams, so I researched their needs with them. Teams create their own templates and choose which roles approve each step, but they can't name a specific person. Approvals follow roles, not people, so a workflow keeps working when someone changes team or leaves.
In the first phase, two teams adopted the workflow builder, two the CMS and one the notification centre.

Engineers still needed help to access services, provision and set up what a new initiative required. I designed the experience of a GenAI assistant for that, backed by agents that connect to Dell's capabilities through MCP.
In interviews, engineers expected a chat that returns code snippets. I added visual diagrams of each step and its status, so engineers got feedback and guidance while the assistant worked. The assistant asks for input when it needs it and helps solve errors when a step fails.
In tests, engineers could follow what was happening after every input they gave.

Minutes
Provisioning time, down from months
80%
Of provisioning done by engineers themselves
13.8K
Active users
With one place to start, engineers stopped waiting on tickets, searching for information and looking for the right people to ask. They still provisioned in Dell's own tools, but now they knew where to go and how.
Measured on engineers' performance and deliveries after adopting the platform: 94K active projects, 63M pipeline runs and 500K user stories.
