The starting problem, what we built, what changed.
Grid-connection applications filled in by hand on the operator's portal, one at a time.
An AI system fills them in on its own: the operator types the client's name and the system does the rest.
Ad creatives wear out fast, and producing new ones took days.
Each week an AI measures, proposes and produces new creatives; nothing goes live without the client's OK.
Review replies written by hand, one by one, across languages, on three different platforms: Booking, Google and Tripadvisor.
An AI pipeline downloads and checks reviews from Booking, Google and Tripadvisor every day, and drafts replies in the house tone: a human reviews and publishes.
Group buying clubs coordinate shared orders by hand, between spreadsheets and WhatsApp threads, with mistakes and wasted time on every cycle.
WeGrocery: a single open-source (MIT) platform that every organisation activates under its own brand via environment variables, with no forks or dedicated code. Order cycles, automatic notifications and two native languages.
Information, decisions and context scattered across chats, emails and documents: every new AI conversation started from scratch.
A structured second brain (the PARA method) that any AI assistant can read and update, with ready-made templates and clear rules on what goes where.
People who submit the form with their tattoo idea go cold before the studio follows up, and the discounts sent out go unused.
As soon as the form comes in, an automatic WhatsApp goes out with a randomly drawn discount; then an AI assistant carries on the chat in 3 languages, collects the details and keeps the lead warm until the studio takes over.
GPT ChatbotWordPressMake.comMeta Business Management API+8The team was using AI by instinct, everyone their own way, with no shared method and no time to learn it properly.
Hands-on classroom training, not slides: a shared session on the fundamentals, then modules tailored to each role, built on the team's real work.
The skin report had to be read and rewritten by hand into a patient presentation, one report at a time.
A Telegram bot takes the report link: a scraper pulls the data, AI turns it into a summary and treatment plan, and the system builds a ready-to-share Google Slides deck on its own.
Several booking sites, inquiries at all hours and in many languages, all handled by hand. And someone who wants a bike today won't wait.
An AI assistant on each site, trained on its own content (bikes, sizes, locations, trails, prices), with the same voice everywhere and a handoff to a human for urgent cases.
GPT ChatbotWordPressMeta Business Management APIClaude+6If a task repeats and eats your time, it can almost always be automated. Let's start with yours.
