Connected the sources
U-ON Travel, Wazzup (WhatsApp and Instagram), Telegram and payment statements flow into one loop with hourly reconciliation.
The operator that dropped manual checking
Leads arrived from Instagram, WhatsApp and Telegram, deals lived in U-ON, payments sat in bank statements, and the overall picture was assembled by hand in a spreadsheet once a week.
Lilo TourTravel, KazakhstanAI command center, 2026
Lilo Tour
Before
A founder opens the CRM, sees forty charts and closes it again. The data is there and the decision is still a guess.
Before
- Four lead sources and nowhere to see them together.
- The founder learned about a stalled deal when the client had already left for a competitor.
- Funnel drops were investigated manually, by rereading conversations.
- The stand-up report took half a day to build and was stale by the meeting.
Under the hood
- Node.js and PostgreSQL, deployed on Railway
- U-ON Travel API — deals, tours, contacts
- Wazzup — WhatsApp and Instagram Direct
- Telegram Bot API and web push — notifications
- Claude — conversation reviews and findings
Case studies
What we built
Leads arrived from Instagram, WhatsApp and Telegram, deals lived in U-ON, and the overall picture was assembled by hand in a spreadsheet. We replaced that with one screen that answers “what is on fire”.
Built the “on fire” screen
The home page shows stalled deals, silent conversations and unpaid invoices, sorted by amount.
Taught an agent to review conversations
At night the agent rereads the day's chats and in the morning brings a list of reasons clients did not reach payment, with a link to every conversation.
Rebuilt analytics around decisions
Funnel, sources, manager performance and money by day. Any number opens up into the list of deals it came from.
Moved digests to the phone
The morning digest and urgent alerts arrive in Telegram and push, so the founder does not have to log in and search.
Redesigned the interface for long shifts
White cards on a light background, Inter with no bold weights, color only on data. The dark theme wore people out on large tables.
Running with clients
What it connects to
We plug into what you already run. If a system is not on the list, we read its API and write a connector.
AI command center, 2026
What changed
We are counting the quantitative results together with the client over the first quarter of operation and will publish them once both sides confirm the numbers.
- The stand-up starts from a ready screen instead of collecting exports.
- A stalled deal shows up within the hour, not at the end of the week.
- A conversion drop is explained by the agent's review rather than manual digging.
- Managers work in one window instead of four tabs.
MoonAI 1.0 · mooonai.com
Do we have to replace our CRM?
At night the agent rereads the day's chats and in the morning brings a list of reasons clients did not reach payment, with a link to every conversation.
Data refreshed a minute ago
MoonAI 1.0 · mooonai.com
How long does rollout take?
Funnel, sources, manager performance and money by day. Any number opens up into the list of deals it came from.
Data refreshed a minute ago
Let us look at your process
Send the basics — industry, channels, CRM, lead volume. We come back with a rollout plan and a quote.