Back to work

Case Study

Dell Premier, Workflow & Access Management

Dell Premier is the portal where large companies buy Dell products. Inside it, different people have different levels of access. Nobody really understood how it worked, so they called Dell instead. Over 120,000 calls a month. My job was to reduce it by half or more.

Client
Dell Technologies
Role
Lead Designer
Duration
6 months
Team
Product, Sales, PAM, Helpdesk, OBM
Platform
Web · B2B Enterprise
Dell Premier, Workflow & Access Management hero

The problem

The system already had everything people needed. The problem was that nobody felt confident using it. Years ago, a site admin deleted something they didn't understand and broke every order on the account. After that, Dell locked parts of the system down and people stopped touching things they weren't sure about. Once someone did get access to Premier, there were no limits. They could buy anything, ship anywhere. So gatekeepers were extra careful about who they approved, and most of them just picked up the phone instead.

Approach

  1. 01Interviewed end users who needed access, site admins who were meant to give it, and Dell reps who kept getting pulled in to help.
  2. 02Ran two workshop sessions with 8+ stakeholders across product, sales, helpdesk and PAMs, spread across the US, Europe and Asia, using Menti for live anonymous polls so room politics didn't drown out honest opinions.
  3. 03Separated the four reasons people were actually calling, then designed a dedicated path for each one.
  4. 04Redesigned the role language around what people said they wanted to do (place orders, create quotes) instead of internal role names.

Context

Calls were the workaround, not the workflow

Dell Premier had a permissions model that had grown for 15 years. Roles, access groups, account hierarchies, all of it technically self-serve. In practice almost nobody used it that way. They called Dell. The volume was structural, not occasional.

120K+
calls per month
3,500
daily access requests handled manually
30–40K
helpdesk cases a year
−30%
target call reduction
The Premier account hub, the surface every site admin lands on before they decide whether to self-serve or call Dell.

Diagnosis

It wasn't a missing feature, it was missing confidence

Site admins could already manage access. They just didn't trust themselves to do it without breaking something. And because Premier had no spending or shipping limits once you were in, the people approving access were carrying a real risk they hadn't signed up for.

I don't want to do it. I don't understand how to do it. So I need help.

Site admin

Research

Three groups, three very different problems

I talked to the people on every side of the request. The same workflow looked completely different depending on where you were standing in it.

There is no way for a customer to track unless they keep track of the email communications. We don't have a customer-facing ticket system.

Dell helpdesk

End users

Didn't know why a feature was blocked. Couldn't find anywhere in Premier to ask. Once a request was sent, they had no idea what happened next.

Site admins

Didn't feel safe making changes. Requests came in through email, Teams, hallway conversations. Nothing was tracked anywhere.

Reps and PAMs

Handling 3,500 access requests a day, most of which the customer could have done themselves. Stuck using the same clunky interface as the customer.

Reframe

One call volume, four different reasons

As I sorted the transcripts, the 120K calls weren't one problem. They were four. Each one needed its own path, and the existing portal funnelled all of them through the same dead end.

Self-serve via admin

Someone needed access and their own site admin could sort it.

Routed through sales

Someone needed access but had to go through their Dell sales rep.

Direct to Dell

Someone needed access and had to contact Dell directly.

Broken access

Someone already had access but something wasn't working.

Workshops

Getting stakeholders to agree on the problem

Two sessions with 8+ people across product, sales, helpdesk and PAMs in three regions. Live anonymous Menti polls let people say what they actually thought without the room politics. The disagreement was real and useful, and it forced a clearer answer at the end.

Stakeholder workshop boards across EMEA/US and APJC, mapping outcomes, edge cases and admin pain points before any screens were drawn.

Don't make site admins feel like they're just a helpdesk.

Susanne, workshop
Some people wanted
vs
Others worried about
Customers to manage their own access
·
Admins being overwhelmed
Fewer calls to Dell
·
Some customers not wanting the responsibility
Less work for sales reps
·
Fraud if the wrong person gets access

Simple requests go to the admin

Adding a user, changing a role. The site admin owns it.

Complex requests stay with Dell

Billing, SSO, payments. Too sensitive to delegate.

Self-service is opt-out

On by default, but a page can turn it off if the customer prefers Dell to handle it.

Insight

The real shift was making people feel safe

This wasn't about building a new feature. It was about making people feel safe enough to do something they technically could already do. Once we framed it that way, half the design decisions became obvious.

Language

Roles that match how people actually talk

Nobody said 'assign the Buyer role.' They said 'I just need them to be able to place orders.' Same system underneath, different surface. I rewrote the role language around outcomes people recognised, and added the consequences of each choice inline so admins could see what would change before confirming.

Ticket detail view, request context, role, and timeline kept together so admins see the why before the decision.

In order to understand access groups and roles, you need to be in it all the time. Most customers just try to log in and start nosing around.

Dell internal

Design

Two flows, deliberately separate

One flow for the person requesting access, one for the admin approving it. Linked together behind the scenes but never collapsed into a single screen. They have completely different needs and fears. Combining them would have made it worse for both.

Admin queue in motion: pending, approved and denied requests live in one searchable place, with approve/deny where the eye lands.
Requester
vs
Site admin
Sees a locked feature with a clear reason
·
Gets a structured notification, not a forwarded email
Submits a request with details pre-filled
·
Reviews queued requests in one dashboard
Gets notified when approved or rejected, with a reason
·
Sees exactly what will change before confirming
Can check request history anytime
·
Access applied automatically, nothing manual

Decisions

Research mapped to design moves

Open queue, the pending state where the admin's decision actually happens. Approve and Deny live where the eye lands.
What people told me
vs
What I did about it
Admins were scared of making mistakes
·
Made the consequences visible before they acted
Users didn't know why they were blocked
·
Added a clear reason to every locked state
Requests came in from everywhere
·
Built one structured place to send and track them
Nobody learned anything from the process
·
Added education at the moment it's needed

Audit

A history people can actually point at

Closed requests stop being lost email threads. Every approval and denial is timestamped, attributable, and searchable, which matters as much for the auditor six months later as for the admin tomorrow.

Closed tab, approved and denied requests stay searchable so nothing has to be reconstructed from memory.

Reflection

What I took away

The hardest part wasn't designing the screens. It was getting people who all knew different pieces of a complicated system to agree on what the actual problem was. Keeping the two flows separate felt counterintuitive at first, but the more we tested it, the more obvious it became that one shared screen would have failed both groups.

Activity thread, the back-and-forth between admin and requester lives inside the ticket, not in someone's inbox.

Outcomes

What changed after the work shipped

120K → <84K
calls per month target
Days → <24h
time to get access
0% → 60%+
people self-serving
−30%
support call reduction target

Other work

All work