SERHANT.'s playbook for rapid AI iteration
Not a pilot. SERHANT. scaled from 200 to 900+ agents without touching infrastructure—here's how they orchestrated Claude, GPT, and Gemini to stay flexible as the AI landscape shifts.
Why it matters
Real-world case study showing how production AI teams balance multi-model deployments, cost optimization, and infrastructure flexibility without vendor lock-in. Demonstrates the shift from single-model bets to task-specific model selection and conversational agents.
The key facts
15 to knowScaled from 200-agent internal pilot to 900+ users with zero API-layer changes
Orchestrates Claude Sonnet, Claude Haiku, OpenAI, and Gemini by task to optimize cost vs. output quality
Used Vercel AI SDK and AI Gateway to abstract model provider complexity and consolidate visibility across multi-model deployments
Generates 35% more content than top five brokerages combined—content generation workload at scale
Shifting from linear workflows (single action → result) to conversational multi-agent workflows (human steers agents mid-flight, chains multiple tasks)
Founded 2020, AI-native real estate company; 800–900+ real estate agents now using S.MPLE platform
Stack: Next.js/Vercel PWA + React Native iOS app; Vercel Fluid Compute for auto-scaling
Scaled from 200-agent pilot to 900+ agents without backend rewrite
Uses Claude Sonnet for reasoning (market analysis), Claude Haiku for speed (intent/field-filling), OpenAI for conversation, Gemini for image generation and computer-use
SERHANT. generates 35% more content than top 5 brokerages combined
API layer required zero changes during 900-user scale-up
Implemented model abstraction layer (Vercel AI SDK) + consolidated observability (AI Gateway) to reduce cognitive load and iteration friction
Experimenting with 'models as guardrails' and caching strategies to optimize token spend
Moving from linear single-task workflows to conversational multi-agent orchestration
Stack: Next.js on Vercel, React Native iOS app, Vercel AI SDK, AI Gateway, Fluid Compute
Go to the source
Vercel Blogvercel.com
Publisher excerpt: Impact at a glance Using multiple models to balance cost, speed, and complexity Worry-free scale: Adding users and assets Future-proofing an unpredictable landscape Started with Next.js on Vercel, which made it easier to expand to a React Native iOS app without rebuilding their backend Engineers…
