Suppo · 2026

Eight channels, one AI front desk

Suppo is an AI receptionist for small businesses in Georgia, built to answer calls and chats on eight channels from one set of facts. I founded, designed and built it.

Category: AI receptionist · Web + Mobile
Role: Founder · Designer · Builder
Timeline: Aug to Sep 2026 · pre-launch
Surfaces: Landing · Portal · iOS app
Year: 2026
Status: Pre-launch
Client: Own product

Visit the live productsuppo.ge(opens in a new tab)

Problem: A small business gets the same price, hours and free-slot questions on eight channels. Whoever is free answers, or nobody does, and two hires to cover phones and social cost 2,100 to 3,350 GEL a month.
What I did: I founded, designed and built Suppo: the landing at suppo.ge, a web portal where the owner teaches and watches the agent, and an owner app on TestFlight.
Result: Pre-launch, so no customer, usage or revenue numbers exist yet. What is measured is build quality: 1,901 of 1,901 keys in Georgian and English, and 363 of 363 engine tests passing at the 11 September deploy.

Suppo, the product shown in a styled scene

01 Problem
The same questions arrive on eight different doors.For a salon, clinic or car service, price, hours and free slot questions arrive on calls, chats, social and email. Whoever is free answers, or nobody does, and the other option is a salaried hire.
Landing page heroEight channels, on a phoneOwner app home
02 Constraints01Eight channels, each with its own app, inbox and connection rules. The Meta app waits for App Review and Viber needs a paid bot application.
02Georgian first with English complete, and the brand font Gilroy has zero Georgian glyphs and no lari sign.
03Wrong answers cost money, so the owner sees every chat, can step in, and approves every post by default.
03 Approach
One receptionist, taught once, watched by the owner.Three surfaces: a landing that sells it, a portal that trains it, an owner app. Portal and app shots show demo data, not customers.
01Teach it oncePaste the website and Facebook page; the agent drafts its knowledge, the owner ticks each line.
02The owner decidesPosts wait for approval. Step into a chat and the agent goes quiet.
03Priced like a menuMessages, calls, posts and ads priced apart, so a shop buys only what it needs.
Teaching the agent its factsPosts waiting for approval
The pricing builder
1 h
the agent goes quiet once the owner steps into a chat
Since 8 September the owner can answer from the portal. The reply is sent first, so a failed send records nothing, then logged, then the agent stays silent in that thread for an hour.
04 The turn
I deleted the orb. Then I wanted it back the same day.The home screen spent its whole view on a decorative orb with zero numbers, so I swapped in a shift report. Within hours I wanted it back: the orb is the agent. It came back with the report under it.
Never cut the brand to make room for the data. Put the data under it.
The orb in full, the shift report below it
05 Results
Build quality16.7 msLanding frame time, meanAfter the 4 September performance round, 17.8 ms worst, down from spikes of 300 ms
−31%Channels page height2,104 px → 1,460 px, after the Connect buttons were rebuilt real
1,901 / 1,901Georgian and English keysPortal dictionary, Georgian first with English complete
363 / 363Engine tests passingAt the 11 September deploy
In contextIn contextsuppo.ge on a MacBook Air, three-quarter view, pale studioIn contextsuppo.ge on a MacBook Air, frontal, pale studioIn contextsuppo.ge on a MacBook Air, close on the screenIn contextsuppo.ge on a MacBook Air, tilted close-up over the keyboardIn contextsuppo.ge on a MacBook Air, low three-quarter view
06 Role and teamOwnedEverything: I founded it, designed it and built it, from the landing at suppo.ge to the web portal and the owner app.
Suppo is pre-launch: the landing and portal are live, real phone numbers wait on a carrier. No customer numbers exist yet, so none appear here.