The power move is to show that the user journey is not just a user path. It is an ownership system spanning sales, pricing, legal, design, and platform constraints — and the scroll below walks all seven beats of it, from Einstein-ranked lead to live account hub.
Every cell is a real handoff. The tinted cells are where the old process lost the deal thread — the redesign made each one an explicit, recoverable state.
I design the system lane as deliberately as the seller lane — the system is an actor with its own failure modes, and pretending otherwise is how dual-list pickers happen.
Lead quality and seller confidence establish urgency.
Locations, devices, SIM/eSIM needs, service type, and equipment constraints.
Catalog items proceed cleanly; non-catalog needs exception handling.
Custom terms, vendor dependencies, and government/enterprise constraints surface.
The output becomes traceable, reviewable, and handoff-safe.
Einstein ranks every inbound lead by predicted conversion; Pardot grades engagement history. My design problem was trust: sellers ignore a black-box score. The queue exposes why a lead ranks — top factors, engagement trail, recency — so the ranking earns the right to order someone's morning.
The intake replaced a Salesforce dual-list-picker maze with nine explicit sections — vertical, network, hardware, security, bandwidth, international, resourcing, compliance, timeline. Each section is independently saveable and recoverable; a seller can stop mid-deal and another can resume with full context.
Everything in the intake routes one of two ways: catalog items flow straight to quote logic; non-catalog needs become structured exceptions instead of email threads. Making the fork explicit is what lets the clean path stay fast.
Connected-vehicle and SASE bundles, volume discounts, multi-year terms — the pricing surface composes them as visible deal components, each with its own approval chain. The seller sees what's standard, what's exceptional, and who owns the exception.
Legal sits with contracting — its job is reviewing the pricing team's work. A linear flow under-describes the real problem: dependencies, exception routes, ownership boundaries. The fix wasn't moving legal earlier; it was handing legal structured pricing output — typed components with approval chains — instead of email threads, so review starts from state, not archaeology.
Every clause traces back to the intake section and pricing component that produced it. The package is traceable, reviewable, and handoff-safe — when negotiation reopens, the team edits state, not documents.
The signed contract instantiates the account hub: orders, SIM activation, fleet status, billing. Post-sale telemetry feeds the next renewal's lead score — the journey is a loop, and the design closes it.
I use the network view because a linear flow under-describes the real problem: dependencies, exception routes, and ownership boundaries.
A senior designer can clean up an intake. A principal designer shows how intake, pricing, governance, and contract readiness become one shared operating model.
Written with engineering before pixels. If a screen can't satisfy these, the screen is wrong — not the criteria.
AC-01Every intake state is resumable by a different seller with zero verbal handoff.AC-02Every exception has an owner and a visible status the customer-facing seller can read.AC-03Every price component shows its approval chain before submission, not after rejection.AC-04Every contract clause traces to the intake section and pricing component that produced it.AC-05Reopening a deal edits state, not documents — no re-keying on renegotiation.Modeling routing, scoring, and clause-tracking as a lane with its own failure modes is why the exception paths got designed instead of discovered in production.
Legal reviewed pricing's work from email threads and PDFs for three quarters before we shipped typed components. The journey map flagged that handoff as the bottleneck in week two. I should have trusted the map.
The complete narrative: research, tiers & scale module, decision log.
Read the case study → 02 / 2022–2024Parent program: discovery, design system, applied AI, operations.
Read parent → Child · 03A decision surface, not a chatbot — typed blocks over federated LLMs.
Read child →