A-Stats Online: a full-stack SaaS for wellness practitioners, built solo and shipped to production

Psychologists, therapists, coaches, bodywork practitioners run solo businesses with the operational complexity of a small agency. Every function has a separate SaaS provider, priced per-seat, none integrated. Intake forms in one tool. Session notes in another. Invoicing in a third. Scheduling in a fourth. Follow-up sequences in a fifth. The practitioner manages the gaps manually.

The hypothesis: could one integrated platform replace six or seven subscriptions, with AI workflows woven throughout, built solo by a single developer on the same stack recommended to clients?

Six-month solo build on the production stack I recommend to clients: Next.js 14, FastAPI, PostgreSQL, Railway + Vercel, Cloudflare. Claude Code orchestration throughout the development process — not as a gimmick, as genuine leverage on a solo timeline.

Forest green (#627862) as the primary color — warm, earthy, grounded, distinct from the blue-sterile enterprise SaaS aesthetic that dominates the space. The brand needs to feel like a practitioner built it, not a VC-funded startup.

AI layer woven into practitioner workflows throughout: intake form parsing to extract key client history, session note drafting from practitioner bullet points, follow-up composition with practitioner voice, invoice memo generation. AI as workflow acceleration, not AI as the product.

Client management (lightweight CRM scoped to practitioner needs), intake form builder with AI-assisted question generation, session scheduling with Google Calendar integration, session notes with AI-assisted drafting, invoicing with Romanian/EU tax handling, automated follow-up sequences, analytics dashboard.

Multi-tenant architecture from day one — each practitioner's data is isolated. AI orchestration respects practitioner voice rather than replacing it. GDPR-first data model — designed for EU compliance, not bolted on after the fact.

Underscoped the design system. I started building UI components before locking a Figma design system and paid for it in rework around week six. Should have spent a full week on design foundations before writing a single component.

I shipped four AI workflows at launch. Two are heavily used by practitioners. Two are barely touched. Lesson: ship one AI feature that genuinely works, measure usage, then iterate. Shipping four half-built AI features is worse than shipping one that practitioners actually reach for.