Treat

Treat lets anyone send a real-world surprise, could be a meal, gift, or experience from a local business near someone they love, anywhere in the world.

Comment

Inspiration

Gifting is one of the most natural expressions of human connection. People love surprising those they care about with food, gifts, services, and experiences. Yet in practice the experience is fragmented — it takes multiple apps and coordination, or people default to plain money transfers that strip away the emotional intent behind the gesture.

We saw a gap: the desire to give is universal, but the infrastructure to support meaningful, real-world gifting at scale simply didn't exist in a simple, unified way.

At the same time, millions of local businesses — food vendors, salons, phone shops, experience providers — struggle to reach customers consistently without spending heavily on marketing or digital storefronts.

Treat bridges these two worlds: it preserves the spirit of gifting while building a system that empowers local businesses everywhere.

What it does

Treat is a global gifting platform that lets anyone, anywhere, send meaningful surprises to people they care about.

Instead of sending generic value or money, senders send real-world experiences — meals, gifts, services, and moments — from local businesses near the recipient. Recipients are notified, verify and claim their Treat, and enjoy it in person or by delivery, depending on the merchant.

Treat is free for merchants: they sign up, get verified, and list their products at no cost. The sender pays a small service charge; the business pays nothing and gains a customer it didn't have to chase.

How AI powers Treat

AI isn't a feature bolted onto Treat — it runs one of our core operational workflows: merchant and organization verification.

When a business applies to join Treat, a dedicated review service calls the Gemini API via Google Cloud Vertex AI, with live Google Search grounding, to research the applicant against the current web and judge whether it looks like a genuine, operating business. It returns a structured verdict — findings, observation, and an approve / decline / request-more-info recommendation — that appears in the admin approval screen.

The AI is deliberately advisory: it informs a human reviewer and never auto-approves or declines. It's also built to be fair to small and local businesses — a thin online footprint is treated as "insufficient information," never as evidence of fraud, so the vendors we exist to serve aren't penalized for being small.

This lets a lean team verify businesses across multiple countries quickly and consistently, without a large manual review team — AI doing real operational work in production, exactly where a small startup needs the leverage most.

How we built it

Treat is a distributed, cloud-native platform built for global scale and real-world gifting workflows.

The frontend is a React app deployed on Vercel. The backend API runs as Ruby on Rails in containers on AWS ECS Fargate, with data in Amazon Aurora PostgreSQL for durable, highly available storage of users, merchants, payments, and claim flows. Authentication supports email/password, MFA, and Google OAuth.

The AI review service is a separate, cross-cloud component: Treat runs on AWS, while the verification service runs on Google Cloud Run, calling Gemini through Vertex AI. The two clouds communicate over signed HTTPS requests webhooks (HMAC SHA-256) with no shared cross-cloud credentials and no PII — only business fields are ever sent for review.

Real-time SMS and email power the gifting and claim experience, and payments run through publicly available payment gateways.

Challenges we faced

Treat meant rethinking traditional e-commerce flows, because every transaction introduces a third participant — the recipient. That forced us to carefully design claim and acceptance flows, refund and cancellation logic, merchant fulfillment, identity verification at pickup or delivery, and dispute handling.

Delivery was the hardest problem, because delivery systems work completely differently from country to country. We're rolling it out per-country to fit each market rather than forcing one global model.

Verifying foreign businesses at a distance was another challenge — one we turned into an opportunity by building the Gemini-powered review service instead of relying on slow, inconsistent manual checks.

What we learned

Gifting isn't just a transaction — it's an emotional experience. The real challenge isn't enabling payments; it's preserving intent, making sure what's sent feels personal, meaningful, and easy to receive.

We also learned that local businesses thrive when they're built into moments of personal connection, not just handed another advertising platform. And we learned that AI, used carefully as an advisor rather than an autopilot, lets a tiny team operate trustworthy processes — like business verification — at a scale that would otherwise need a whole department.

What's next for Treat

We launched into production and are now onboarding real businesses and users, starting in Ghana and expanding to Nigeria and the Ghanaian diaspora in the UK.

Next:

  • Android and mobile app rollout
  • Expand to more countries as we finalize payment-vendor agreements
  • Roll out country-specific delivery
  • Add more communication channels, including WhatsApp
  • Deepen AI operations — extending the Gemini review agent toward richer, multi-step verification
  • Improve resilience with multi-cloud (GCP) integration

Ultimately, Treat aims to become the global infrastructure for intentional gifting — connecting people and local businesses through everyday moments of generosity.

Share this project:

Updates