Last weekend I spent two days at the Workpeg Lagos Hospitality Hackathon with my teammates Faruk Amoo and Emmanuel Falode. Ten teams pitched. We finished 4th, 2.2 points off the podium.
We didn't win the cheque. But the product we built, and the decisions we made building it, taught me more than most quarters of regular work do. This post covers what we built, how it works, the pivot we made on day one, and what I'm taking into the next one.
Think about how people actually book things in Lagos. You find a hotel or lounge on Instagram. You send a DM. You wait. Someone sends a price list as an image. You ask whether a room is free on Friday. You wait again. Then you're told to "send the transfer and the receipt", and you hope the booking is real when you show up.
Everybody is already on WhatsApp. Guests are comfortable there and so are the businesses. What's missing is something on the other end that can answer straight away, check real availability, take payment, and confirm the booking without a human copying details between five places.
That was the gap we went after.
Kaabo (Yoruba for "welcome") is an AI hospitality concierge that lives on WhatsApp.
For guests, there is nothing to download and no menu of numbered options. You just talk to it: Natural conversation, any language. You write the way you'd text a friend, in whatever language you speak, and Kaabo replies in that same language.
Voice in, voice out. Send a voice note and you get a voice note back (plus the text). Type a message and you get text. A lot of people would rather talk than type, and the reply should match how they spoke.
Find, check, book. Hotels and short-lets near you, with rooms and prices. Lounges and restaurants. Car and luxury car hire.
Pay upfront, with a hold. When you pick something, Kaabo holds it for 30 minutes while you pay. When the payment lands, the booking is confirmed. If it doesn't, the hold expires and the slot goes back on sale.
Order once you're there. At a partner restaurant or lounge, you can order from its menu in the same chat, for dine-in or takeout.
No hotel stay required. Someone who lives in Lekki can use Kaabo just to book a lounge or a car for the night.
For venues, every hotel, lounge, restaurant or car-hire partner is its own business on Workpeg, with a guest desk where its bookings, guest contacts and conversations live. That was important to us: the guest relationship belongs to the venue, not to us. Kaabo is the channel, and the venue owns the customer.
For the platform, payments are split automatically. The guest pays once, each venue receives its share, and Kaabo's fee is deducted at that point. Nobody has to reconcile it by hand afterwards.
At a high level there are three layers:
The conversation layer (WhatsApp). This layer takes text or voice, works out the language and what the guest wants, and decides how to reply: a voice note for a voice note, text for text. It never shows the guest a rigid menu. The guest says what they want and the system works out which action to take.
One generic booking engine. This was the most important design decision of the weekend. Rooms, tables, lounge sections and cars all go through a single booking model: something bookable, belonging to a venue, for a time window, at a price, with a hold and a payment. We didn't build a separate flow for each venue type.
The venue's own workspace on Workpeg. Each partner venue is a separate business built on Workpeg's building blocks ("pegs"). Confirmed bookings, guest contacts and follow-ups land in that venue's guest desk, where its staff already work.
The hold-then-pay flow sits in the middle. A booking starts as a hold with an expiry. Payment turns it into a confirmed booking and triggers the split. If the expiry passes first, the hold is released. That one state machine handles a hotel room on Friday night and a car at 7pm on Saturday the same way.
We came in with Kaabo as a guest-desk product for hotels and short-lets: a WhatsApp concierge for booking rooms, plus a front-desk dashboard for the hotel.
On the first morning we realised at least two other teams were also building hotel front-desk products. Three teams pitching the same idea to the same judges is a bad position, however good your execution is.
So we widened the scope. Kaabo stopped being "book a hotel room on WhatsApp" and became "book hospitality in Lagos": everything a guest needs around a stay (eating out, lounges, nightlife, car hire), from the hotel itself or from places nearby.
Two things made that possible within a hackathon:
The generic booking engine. A widened scope only works if every new venue type doesn't need its own code path. Once a lounge table and a car were just "bookable things" like a room, adding them meant adding data, not rewriting flows.
Saying no out loud. We explicitly cut flights and timeshare. Both fit the "around a stay" story, but neither would have been solid by demo time. A smaller scope that works beats a bigger one that breaks on stage.
We also added one small but high-value touch: after a guest pays for a room, Kaabo makes a single offer of extras (a table, a car, a lounge) from the hotel or nearby partners. It's one offer, not a stream of upsells.
Look at the field early. The most valuable thing we did on day one wasn't writing code. It was seeing what everyone else was building and changing course before it was too late.
Generic building blocks pay for themselves. A booking engine that doesn't care whether it's booking a room or a car is the reason the pivot fit into a weekend. Build the primitive, then configure it.
Build deep on the host's platform. Workpeg's pitch is "AI that doesn't just answer, it takes action." Making every venue a real business on Workpeg, with real guest desks, meant our demo was also a demo of their platform. Sponsors want to see their product used well.
Cuts are features. Every item we cut gave the remaining ones more time to actually work.
Fourth is data, not a verdict. We were 2.2 points off third. That gap is small enough to learn from and close next time.
A big thank you to the Workpeg team for a really well-run event. They looked after every participant, from the food to the mentoring to the energy in the room, and I learnt a lot about building on a platform designed for AI that takes action.
Congratulations to ClearTab on the win, and to Roomay and HotelOp for second and third. You earned it.
And to Faruk and Emmanuel: we built something real in a week. We're not done with Kaabo.
If you're working on hospitality, conversational commerce or AI agents in Nigeria, I'd love to compare notes. Find me as @abdulahmed_dev.