AI Receptionist for SMEs by Akhil YadavAI Receptionist for SMEs by Akhil Yadav

AI Receptionist for SMEs

Akhil Yadav

Akhil Yadav

Handwerk40 — AI receptionist for German trades
Handwerk40 — AI receptionist for German trades
Handwerk40 owner dashboard
Handwerk40 owner dashboard
Self-initiated prototype for German trades businesses. The application and API were run locally, with 65 automated tests passing. External services use mock fallbacks; no live calls, bookings, payments, or messages were made.

Overview

Handwerk40 is a German-language SaaS prototype for skilled-trade businesses such as plumbing, heating, and electrical contractors. It is designed to handle incoming enquiries, identify urgency, record call details, and help owners manage follow-up requests and quotes.

The problem

Tradespeople are often on site and unable to answer every call. Missed enquiries can mean delayed responses and potentially lost work, while owners have limited time to manage requests and prepare quotes.

Product workflow

Receive and normalize: accept call events through a signed voice-provider webhook.
Qualify: identify the trade, urgency, and possible emergency, then assign a disposition such as qualified, partial, or failed.
Review: present call details, open requests, emergency flags, and activity statistics in the German-language owner dashboard.
Prepare quotes: create quotes with line items, sequential numbers, and 19% VAT.
Manage the service: expose settings and billing pages, including subscription tiers and usage-based overage logic.

Privacy-conscious design

The prototype includes an AI disclosure in the greeting, configurable retention windows, a data-processing agreement flow, and an erasure workflow. These are implemented software features, not a claim of independently verified GDPR compliance.

Engineering and verification

Built as a pnpm monorepo using TypeScript, Fastify, Next.js, Drizzle ORM, and embedded PGlite or PostgreSQL. A ports-and-adapters approach gives external integrations—telephony, voice, calendar, payments, email, SMS, storage, and LLM services—mock fallbacks so the stack can run locally without live accounts.
The API test suite passed 65 tests across seven files, covering qualification, compliance-related rules, pricing, webhooks, cryptography, and authentication-token rotation.

Verification and limitations

External integrations were run in mock mode; no real calls, calendar bookings, payments, or messages were made.
Dashboard records are seeded demo data with fictional customers.
Privacy and compliance flows have not undergone an independent legal review.

Source code

Like this project

Posted Oct 18, 2026

A SaaS for SMEs that qualifies call enquiries, flags urgency, manages requests and prepares quotes.