Next.js & React
Next.js is our default for new product builds and modernisation projects. The React Server Components model maps well to AI features — server-side calls to model APIs without exposing keys to the client, edge functions for low-latency chat, ISR for content that needs to revalidate against a database. We've shipped a dozen Next.js apps to production across financial services, healthcare, and SaaS, and the platform you're reading this on is one of them. We commit to Next.js for a specific shape of project: AI-powered B2B web apps where shipping speed matters and the team will extend it after we hand over.
What you get
Real examples
AI-powered B2B SaaS platform
Illustrative scenario: an Australian fintech ships an AI-assisted underwriting product. Next.js handles the front-end + AI orchestration in one codebase. RAG over policy documents runs in API routes; document upload + extraction uses streaming server-sent events for live status. Single repo, single deployment, full audit trail.
Multi-site marketing platform with shared data
Illustrative scenario: a consultancy operating multiple brand sites needs one CMS, one analytics pipeline, and brand-specific theming. Next.js + Turborepo monorepo handles the shared packages; each site deploys independently to its own Vercel project. (This is the architecture the Horizon Labs website itself runs on.)
Internal admin tool with embedded AI assistants
Illustrative scenario: a logistics company's operations team gets a custom admin UI built in Next.js with Claude-powered assistants embedded for triage, search, and recommendations. Server Components render the data fast; AI calls stay server-side; the UI feels like a native app.
Common questions
Why Next.js over SvelteKit, Remix, or Astro?
Three reasons. One, the React ecosystem is the deepest pool of AI/ML libraries and components. Two, Vercel's deployment + preview-env story is genuinely best-in-class for iteration speed. Three, when we hand a project to an in-house team, hiring React developers in Australia is easier than hiring Svelte or Remix devs. The other frameworks are technically strong; the operational story for in-house handover favours Next.js.
Can we self-host Next.js instead of using Vercel?
Yes. Next.js runs anywhere Node runs — AWS, GCP, Azure, on-prem. We've deployed Next.js to ECS, Cloud Run, and Kubernetes. Vercel is our default for the fastest shipping experience, but it's not a lock-in.
What about Server Components and edge functions for AI?
These are the killer features for AI products. Server Components let you call model APIs directly in the component tree with the API key server-side and stream the response to the client. Edge functions run those calls from the nearest region to the user, which on the Sydney + Melbourne edges keeps latency under 100ms for round-trip metadata calls.
Can you modernise an existing React app to Next.js?
Yes — this is a common engagement. We've migrated from Create React App, Vite, and even legacy Webpack setups. The pattern is incremental: stand up a Next.js shell, port routes one at a time, deprecate the old build. Typical timeline is 6-12 weeks for a mid-sized SPA.
How do you handle authentication in Next.js?
Default to Supabase Auth or Clerk for new builds — both have first-class Next.js integration with Server Component-aware sessions. For enterprises with existing SSO (Okta, Azure AD), we use NextAuth.js or Auth.js. Either way, sessions flow through middleware so AI endpoints can check auth before hitting the model.
Ready to get started?
Tell us about your project and we'll tell you honestly how we can help.
Get in Touch