A portal for e-mobility – developed with AI support

Customer

Services

  • Customer portal
  • Web Development
  • User Experience
  • User interface
  • Requirements Engineering

One portal for charging infrastructure, fleets, partners and billing. Built in around six months. Developed with AI support, owned by humans.

For JOYA, Switzerland's largest charging network and the new mobility brand of Swisscharge, GOFAST, MOVE and Energie 360°, we showed how AI speed and solid engineering go together – with clear guardrails.

Visit JOYA
A look inside the JOYA B2B portal: login, driver groups, sites and charging sessions

One portal instead of two apps

  • For Switzerland's largest charging network with more than 25,000 charging points
  • Charging infrastructure, fleets, partners and billing in one portal for B2B customers and admins
  • More than 250 user stories delivered in around six months
  • Multilingual, multi-tenant and with white-label branding per operator
One portal, four roles: platform operator, operator, partner and fleet manager each see only their own areas

One network, one app. And a lot of complexity behind it.

With JOYA, Swisscharge, GOFAST, MOVE and the mobility division of Energie 360° join forces to form Switzerland's largest charging network. The promise to customers: “One network, one app. All across Switzerland.” Behind it, JOYA plans, installs and operates charging infrastructure for businesses – from a single charging point to an entire charging hub.

Operators, partners, site owners and fleet managers need a portal to run their entire charging business: live fast, reliable in operation and extensible by JOYA itself. Roles and permissions have to apply across the whole system, not just in the UI. And the portal has to keep pace while backend and requirements evolve in parallel.

One portal that knows every role

Everyone sees and does exactly what their permissions allow.

  • Fleet management: invite drivers by email or CSV/Excel, manage driver groups, order and block RFID cards.
  • Charging infrastructure: sites, charge points and EVSEs, access and tariff groups, start and stop charging sessions remotely.
  • Partners & billing: onboarding, contracts with versioned money flows, billing periods per document type, monthly close for accounting.
  • Identity & access: central identity management with SSO.
More than 250 user stories delivered in around six months

Speed that stays under control

Guardrails instead of vibe coding

AI speeds up delivery; architecture, quality and accountability stay with humans. That is why we first built an AI harness, the guardrails for the AI: repo setup, architecture rules, pipelines, code owners, a testing strategy and clear rules. Only then did we scale up.

From Jira straight into the repo

Requirements flow automatically from Jira into the repository. Every story is classified – ready, needs a design prototype or needs a decision – and re-checked after every change to the API contract. So it is always clear what can be built and what is blocked.

A feature factory with quality gates

Screens, forms, tables and workflows are built with AI support along clear patterns. Human in the loop applies throughout: developers review every change under the four-eyes principle, backed by comprehensive automated tests.

Design system page of the JOYA portal with colours, typography, badges and components

Designing in code instead of Figma

The foundation is a design system built on shadcn/ui in the new brand design. A dedicated design system page in the portal shows every component live – from colours and tokens to tables, forms, toasts and tooltips.

  • A designer in the code: Our designer works directly in code instead of Figma, using the same technologies as the developers. They set guidelines and prompt design improvements themselves, which show up in the real UI right away.
  • Built for agentic engineering: shadcn/ui lives in the repository as readable source code. The AI builds new screens from the same building blocks instead of inventing variants. Recurring patterns such as multi-select or group pickers become shared components.
  • Prototypes as the spec: New flows are built as clickable prototypes in code and serve as the visual spec, with no handover gap to delivery.

From foundation to feature factory

  • Foundation: architecture, pipelines and AI rules.
  • Vertical slice: a first end-to-end flow with login, API integration, permissions, i18n and tests.
  • Feature factory: fleet, infrastructure, partners and billing, delivered with AI support.
  • Production setup: acceptance testing, monitoring and go-live readiness.

Weekly updates, a regular jour fixe and close collaboration with the backend team kept everyone on the same page.

Zwei Männer sitzen drinnen an einem Tisch; einer im weissen Hemd spricht und hält eine Schüssel, während ein anderer mit Brille im Hintergrund zusieht.

Want to get there faster with AI? Peter is there for you.