AI Automation — Stock & Sales Bot

AI AUTOMATION

Stock & sales via Telegram · TypeScript · freelance

Freelance integration for shops and small entrepreneurs: stock and sales control through a Telegram bot with voice-note support, integrated with Google Sheets, digital stores and Gmail (capturing events via Google Workflows). Designed to be simple and cheap — real automation for those who can't afford 'enterprise' automation.

SERVERLESS + IaC

Lambda · API Gateway · Terraform

Events on AWS Lambda behind API Gateway, with Terraform as infrastructure as code to deploy each instance automatically. A modular, dynamic architecture: the same system re-adapts to each business model, with a set of agent skills to give it context and reconfigure it fast. LLMs integrated through open-source, multi-provider libraries.

'enterprise' flows → too pricey for a small shop

Discarded path

Discarded path: complex automation suites are out of reach for small producers — who already have their stock and sales control built in spreadsheets anyway. The system adapts to what the business already uses, not the other way around.

exhibit A: this site's robot

Bito, this portfolio's guide, is living evidence: an agent with streaming, guardrails and a bilingual persona that I built. You're using it right now.

the bot on duty

You send the bot a voice note — 'sold three bags of feed' — and the business's spreadsheet updates itself. It also captures Gmail events via Google Workflows and integrates with digital stores. Useful AI is the kind that runs at 3 AM with nobody watching.

cached responses → fewer requests

Cached responses to avoid redundancy in user prompts and save requests to the model. In a product that's cheap by design, every token counts.

20+ SHOPS, TWO YEARS WITHOUT FAILURES

automation at small-business prices

Live at over 20 clients, running correctly and without issues for more than two years, with regular updates. The best metric for a cheap, simple system: nobody ever has to think about it.

STACK

TypeScript Lambdas on AWS with API Gateway for events, Terraform to deploy each instance as code, Telegram as the interface and several interchangeable LLM providers behind a single layer. Serverless: if nobody uses it, it costs nothing.