Run your business from WhatsApp with AI agents.

The problem

Fleet, repairs, inspections, shifts and timesheets lived in separate spreadsheets for each area. Nobody could see in real time which vehicles were off the road or who could cover each route, while the customer expected flawless service.

What we built

• A web control centre: fleet, vehicles off the road, incidents, inspections and staff in one place
• Timesheets synced every night, with no manual copying
• An automatic shift planner that respects consecutive days and free weekends
• Shifts sent to each driver by WhatsApp and email
• An AI copilot to ask about the operation, on the web, by messaging or by voice

How it runs day to day

The dispatch team opens the control centre and sees the day's fleet and incidents. The planner proposes the week's shifts and every driver gets theirs on their phone. Whenever someone needs a figure they ask the copilot, and if they ask it to change something it confirms before writing.

What changed

• One single source of truth for the whole operation instead of a spreadsheet per area
• In the week of 14–20 September, the planner covered every route the available staff could cover
• The team spends its time keeping the customer happy, not reconciling spreadsheets

The problem

Shoppers hesitated right when deciding — what goes with this, which size, when will it arrive — and the team kept answering the same questions about orders and returns.

What we built

• An AI shopping assistant on the website and on every product page
• Suggestions of what to pair each item with, with photos and links
• The shop's WhatsApp handled by the same agent
• A loyalty club with a welcome discount
• WhatsApp campaigns with a personal code for each shopper (in preparation)

How it runs day to day

A shopper opens an item and asks. The assistant suggests what to pair it with, shows those items with photo and link, and answers shipping and returns questions. If she writes on WhatsApp, the same agent picks it up with the same information.

What changed

• Every shopper has a stylist on hand right when she's deciding
• Repetitive order and returns questions no longer land on the team

The problem

Travellers from all over the world writing and calling from every time zone, in high season, with questions about their booking that only someone with access to the system could answer.

What we built

• Chat and voice agents that know every destination
• A connection to their booking system
• Booking lookup by email or phone: status, dates and payments
• Support in the traveller's own language

How it runs day to day

A traveller asks about their stay at three in the morning. The agent identifies them by email or phone, looks up the booking and answers with their dates, status and what they've paid.

What changed

• Travellers looked after at any hour and in their own language
• Booking questions no longer wait for the office to open

The problem

Orders came in by phone, by WhatsApp and on paper; the catalogue was huge and hard to maintain, the website didn't sell and support ended when the office closed.

What we built

• A new B2B website with the whole catalogue published and a quote basket
• An AI agent in the website chat and on WhatsApp that knows the catalogue
• It recognises customers already in their management system
• The client's own customer-service rules, applied and tested with real messages
• A custom orders, routes and customer-portal system (in development)

How it runs day to day

During office hours the team answers, as always. After hours, a baker writes on WhatsApp or through the website, the agent recognises them, answers catalogue questions and points them towards the order or quote. Next day, the team picks up whatever needs a person.

What changed

• Trade customers served after hours too
• The whole catalogue published and findable on the website
• A sales team focused on advising, not on taking down orders

The problem

Information was scattered across markets, news and indicators, and the tools handed out numbers without explaining what they meant or why they mattered.

What we built

• An app with portfolios and strategies tested on paper, without moving money
• An AI agent by chat and voice that analyses and explains its reasoning
• A fair-price calculation for hundreds of assets
• A private dashboard for each client

How it runs day to day

The client opens their dashboard or talks to the agent: “how is my portfolio doing?”, “why did this drop?”. The agent cross-checks prices, news and indicators and answers explaining why. Strategies are tested on paper; nothing moves real money.

What changed

• An assistant on call at any hour that doesn't just hand over numbers: it explains what they mean
• All the information and tools in one place

The problem

End-of-day tills arrived as photos on WhatsApp, supplier invoices piled up, and cash and bank figures only showed up at month end, too late to act on.

What we built

• A multilingual website with bookings, confirmations and reminders
• Order and pay at the table with a QR code
• A floor and till app that reconciles every close with the card terminal
• Supplier invoices read automatically and a shopping list with estimated prices
• Bank transactions imported without duplicates
• A daily WhatsApp message with the day's cash and the bank balance

How it runs day to day

At closing time the team does the close in the app and the system checks it against the terminal. In the morning the owner gets the day's cash and bank balance on WhatsApp. Bookings and table orders come in on their own, and the shopping list is already priced.

What changed

• Tills check themselves, no more reconciling by eye
• Purchases are priced before anyone goes shopping
• Cash is followed from a phone, without waiting for month end

The problem

Supplier invoices piled up in two inboxes, income was reconciled by hand in a spreadsheet per venue, and staff costs were copied from the payroll PDF.

What we built

• Inboxes read automatically: every invoice saved and filed by venue
• AI extraction of issuer, tax ID, net, VAT and total
• Till closes read straight from the card terminal
• The usual spreadsheet, filled in automatically every morning
• A daily report with sales and new invoices
• A filter that archives email that isn't work

How it runs day to day

Every night the system reads both inboxes, extracts the new invoices and fills in the income spreadsheet. In the morning the daily report arrives. The team keeps working in the spreadsheet it knows, and can ask the agent about a supplier, a month or a duplicate invoice.

What changed

• The first run sorted close to 900 invoices from dozens of suppliers
• The inbox was left with no pending invoices
• No more typing up the till every night