salesforce for netcom bw
Design of Self Service Portal for German ISP
Duration of project: January 2025 – January 2026
€0.0M+
per month mitigated based on error-related losses and manual process
0%
reduction in time used for research by utilising multi-agent AI
0%
increase in CSAT: positive reviews within the first 90 days.
0
times more efficient sales process. Signups done via the portal instead of agent visit
Executive summary
Netcom is a German Internet Services Provider delivering broadband and mobile services to both business and consumer customers. The company model was built on a agent-driven sales process. Data captured by hand during visits, re-entered manually in the office. The flow was prone to errors, and the error correction workflow was expensive and inefficient. The financial losses from this model were estimated at €1.2 million every month.
The objective was to eliminate the financial exposure caused by manual processes and replace the agent-dependent model with a scalable digital product that allowed customers to self-serve.
I was engaged as the lead UX designer with full-scope ownership, from strategic diagnosis through to delivery. Operating within a lean Agile team, I worked directly with C-suite stakeholders, the Salesforce product team, and engineering. My responsibility spanned two connected products. An agent-facing application and a customer-facing Self-Service Portal. Both required to share the same business logic and had to be shipped under significant time pressure.
Discovery & strategic insights
With no legacy research documentation available, I orchestrated a compressed but intense discovery sprint. I conducted direct interviews with field agents and end customers, embedded with the office teams absorbing the error correction workload, and mapped the complete order journey end-to-end. This hands-on approach surfaced some interesting findings.
Compressing the research cycle with AI-augmented simulation
A traditional three-week discovery and synthesis cycle was not viable. To reduce research risk without sacrificing insight quality, I designed and ran a multi-agent AI simulation . Four agents in total, two Claude and two Gemini instances, each prompt-engineered with distinct user personas, personality traits, and real customer pain points. Running these agents in parallel against proposed interaction models allowed me to surface usability friction points in a single day rather than weeks.
For the research I used 4 agents in total 2x Gemini and 2x Claude. I then multiply these agents on various tasks and enrich them with specific goals, personality traits and problems. Each agent represents a real client with unique requirements and pain points.
The discovery findings were used to align the business and the Salesforce product team around a phased delivery strategy. The plan was to ship a minimal agent tool immediately to stop the financial losses, and reduce manual errors. Then architect the larger self-service tool in parallel. This framing gave stakeholders a clear risk-reduction narrative and secured the rapid buy-in needed to begin production within days of discovery completing.
The solution
Phase 1: Agent iPad application design
The immediate design priority was eliminating data entry errors. I ideated a high-fidelity iPad application built around real-time validation at the moment of sale. Customer details such as names, IDs, contact information, were verified as the agent entered them, at the customer’s address, before the visit ended. The manual re-entry step was removed from the workflow entirely.
The MVP was deliberately minimal. Scoped, designed, tested, and shipped in under two weeks. The Lean validation loop was tight: prototype, test with agents in the field, iterate, ship. Over the following four months, I replaced the MVP with a fully featured, production-ready platform.
The initial design helped agents capture customer information faster and with fewer errors.
Phase 2: Design of an independent Self Service Portal
The problem was structural dependency since the business relied on field agents for routine account management tasks that customers were fully capable of handling independently. I architected a Self-Service Portal (SSP) that transferred control directly to the customer: account review, product upgrades and downgrades, add-on purchases, settings management, and competitor number porting, all completable by the user through a single, unified interface.
The SSP was designed as a full e-commerce journey, integrated with Salesforce Service Cloud, with automated order submission, legal documentation, and product logic built into the flow. The user experience was designed for independent completion, removing every dependency on human intervention for routine transactions.
Netcom had no existing design system or component library. I built one in Figma using design tokens and variables, establishing a shared style guide and component library that served both the agent iPad app and the SSP from a single source of truth. Because the two platforms shared identical business logic, the design system ensured behavioural and visual consistency across both without duplicating design or engineering work.
Complex business rules, hardware compatibility, cabling constraints, product eligibility conditions, were translated into clear interactive decision flows and annotated engineering specifications, with AI tooling used to structure and navigate rule complexity quickly under deadline pressure.

Impact & business Value
Results were measurable within the first month and confirmed at the 90-day mark.
Quantifiable outcomes
-
€1.2M monthly financial exposure eliminated - data entry errors at source removed; error correction workflow ceased entirely within weeks of MVP launch
-
30% increase in customer satisfaction scores within the first 3 months of Self-Service Portal launch
-
Call centre load reduced significantly as routine account management transactions migrated to the self-service channel
-
Full e-commerce automation achieved. Order submission, legal documentation, product management, and number porting now completed by the user.

Qualitative outcomes
- The two-week MVP delivery established design credibility with the business and the Salesforce product team, enabling faster collaboration on the SSP build
- The design system gave engineering the autonomy to ship consistently across both platforms without design as a bottleneck