How one COO runs a €100M+ multi-business portfolio — with an AI assistant on Telegram

A multi-business portfolio — retail stores, sports facilities, parking lots, office buildings, manufacturing properties, residential estates — runs under a single COO. The portfolio spans 30+ buildings and locations, 100+ employees on payroll, 50+ suppliers and contractors in regular orbit, and over €100M in annual turnover. The COO is the buck-stops-here for hiring, opening and closing stores, vehicle fleets, cleaning and gardening across the estates, maintenance at the sports facilities, revenue tracking, store inspections, and vendor coordination. His direct reports number in the dozens; the staff under them, in the hundreds. The work was simply too large for one person to keep in their head — even a great one.

What we built

The COO had already organized his information well — but it lived across documents, messages, and his own memory. We did two things. First, we structured the entire portfolio into Supabase: people, locations, assets, vehicles, schedules, vendors, expenses, revenue. Then we built the intake pipelines so new information lands in the right place — direct reports, voice notes, forwarded messages, all parsed and routed. Once the data was clean and live, we gave the COO an OpenClaw autonomous AI assistant on Telegram that works against that data on his behalf.

The assistant does what a senior chief-of-staff would do: takes tasks the COO delegates ("get me the Q3 maintenance summary for the sports facilities by Friday"), pulls information from direct reports and structures their answers, follows up if anything’s missing, generates ad-hoc reports the moment he asks for one, runs scheduled ones quietly in the background, and manages notifications, calendar entries, meeting prep, and reminders. The COO stays the decision-maker; the assistant keeps everything moving between his decisions.

How we built it

Supabase is the backbone — a structured, queryable model of the entire portfolio: people, places, vehicles, assets, jobs, money. OpenClaw runs as the AI assistant, with full access to that data and tools to act on it: write to Supabase, schedule notifications, generate documents, send Telegram messages. The interface is Telegram, deliberately — the COO was already there all day. There is no separate dashboard to remember to open. Voice notes and forwarded messages from direct reports go straight into the assistant, get parsed, and land in the right place in the database. We own and maintain the data layer and the assistant; the COO owns the decisions.

The outcome

One COO, six asset classes, €100M+ in annual turnover — run from a chat window. Information that used to live across documents, inboxes, and one person’s memory is now structured, queryable, and worked on autonomously by an AI assistant in his service. The volume that was overwhelming is now a manageable conversation.