VoxAI
Service-based businesses that rely on inbound consultations and appointment scheduling—specifically consulting firms, recruitment agencies, beauty & wellness studios, education centers, and repair services. Today, their front-desk workflow is fundamentally broken by operational fragmentation. Up to 40% of valuable customer inquiries arrive outside standard operating hours or during peak staff workload, leading directly to dropped leads and customer churn to faster competitors. In-house staff spend hours every day answering the exact same routine questions regarding pricing, location, and open slots instead of focusing on high-value clients. Meanwhile, bookings are coordinated manually across unstructured channels like WhatsApp messages, phone calls, spreadsheets, and paper agendas. This manual handling causes scheduling delays, double-booking conflicts, and administrative chaos. Existing basic FAQ chatbots fail because they only spit out canned text without executing actual transactions, leaving business owners trapped between rising administrative payroll and lost revenue.
Preliminary score
- Value for the user17.0 / 25
A specific problem and a clear outcome from solving it.
- Prototype and use of AI13.0 / 30
A working core scenario, and what AI actually contributes to the result.
- Quality testing12.5 / 20
Tests, examples of failures, and comparison with the current approach.
- Feasibility7.5 / 15
Data requirements, running costs and a clear next step.
- Originality4.0 / 10
A meaningful difference from common solutions.
What the judges wrote
- GPT-6 AstraPreliminary59/100
- Strengths
- Your described workflow connects service selection, availability checks, explicit confirmation, and appointment creation across two isolated demo businesses. Your testing account identifies concrete failures, including the company-header mismatch and serverless persistence limitations, rather than merely claiming success. Your clear separation of local capabilities, deployed placeholders, and unfinished integrations makes the prototype's boundaries understandable.
- To improve
- Your next demonstration should show live AI completing and cancelling a booking, backed by reproducible test results. Your evaluation should measure booking completion, scheduling errors, and staff time against the current manual workflow. Your deployment plan needs durable storage, a realistic next pilot, and operating costs per completed booking. Your Vercel setup can use server-side environment variables without requiring a paid plan; keep long-lived Gemini keys out of browser responses and use short-lived tokens for voice.
- Claude Opus 5.5Preliminary49/100
- Strengths
- Your problem statement names concrete business types and a real pain: after-hours inquiries, repetitive questions and manual booking across WhatsApp, phone and spreadsheets. The booking engine with multi-tenant isolation and double-booking prevention is a solid, transactional core rather than a canned FAQ bot. Your testing write-up is unusually candid: it lists specific test areas, a real header bug (X-Company-Id vs X-Company-Key) you found and fixed, and clear explanations of what does not work on the live demo.
- To improve
- The AI is the heart of your pitch, but on the live demo the chat returns a template reply and voice is disabled, and with no repository or video the 28 tests and the tool-calling flow cannot be verified. A short screen recording of the local Gemini agent booking a slot end to end would change this a lot. Vercel environment variables are available on the free plan, and your /api/voice/config endpoint sends the Gemini key to the browser, so look at ephemeral tokens or a small persistent host instead. Name one adoption metric, such as the share of after-hours inquiries converted to bookings or admin hours saved, and add running costs per conversation and per call minute.
The team
Vusal IsayevLead
developer
Ruhin Bayramlı Şahmar
developer
Nuran Heydərov Bəhruz
product