Case Study - AI Assistant for Business Operations
Queried connects to a company's systems so teams and customers can query data, create offers and place orders in plain language.
- Client
- Queried
- Year
- Service
- AI, Web Development, Cloud Services
One of the biggest importers and distributors of natural stone in Romania.
algabeth.ro- Algabeth
- Launch partner and first deployment. Their catalogue, customers and daily questions shaped the product.
- XCODES
- Designed and built Queried end to end. The architecture, the application, the AI, the channels and the mobile apps.
Every screen on this page is Queried running for Algabeth, so the examples are related to marble and granite: slabs and finishes, prices per m², kitchen studios and architects. The product itself is not tied to stone — it connects to any company.
All screenshots show demo data. Names, phone numbers, conversations, prices and stock figures are fictitious and for illustration only.
No real customer or personal data is displayed, and any resemblance to actual people or companies is coincidental.
Problems
to solve
Most operational work is answering the same questions all day: is it in stock, at what price, can we offer it, can we deliver by Friday. The answers live in a database, an ERP, a price list, a folder of documents and, mostly, in the heads of a few experienced people — so every question needs one of them, and turning an answer into an offer or an order is a second job.
The questions also arrive everywhere — the website, WhatsApp, the phone — and each channel gets answered in a different place, by whoever is free, with no shared record of what was promised.
At Algabeth that looked like a stone yard fielding "which granite for a kitchen worktop, in 30 mm, delivered to Cluj?" dozens of times a day. The shape of the problem is the same in any business that sells from a catalogue.
Solutions
Queried is an AI assistant that works directly on the company's own systems. It connects to the databases, APIs and tools already in place, learns the business rules — price lists, who approves what — and then takes requests in natural language: an answer, a ready-to-send offer, or a placed order, in one conversation.
The same assistant serves customers on the website widget, on WhatsApp and on AI voice calls, with the same knowledge and the same tone, and a human inbox behind it so a person can take over at any point.
Each company runs in its own isolated database and workspace, with role-based access, real-time updates over WebSockets, and native iOS and Android apps for the team on the move.
Technologies
- Tailwind
- Vue.js
- Inertia.js
- Laravel
- PHP
- Docker
- Cloud Services
Presentation
Queried is an AI assistant for business operations. It connects to the systems a company already runs on — databases, ERPs, price lists, documents — and lets teams and customers ask for what they need in plain language: an answer, a ready-to-send offer or a placed order, in one conversation.
The same assistant works on the website, on WhatsApp and on AI voice calls, with a human inbox behind it so a person can step in at any moment. Each company gets its own isolated workspace, native iOS and Android apps, and an assistant that only answers from its own data and cites what it used.
queried.ai
Query
Ask a question. Get the work done.
Query is the heart of the platform: a conversation with the company's live data — products, stock, prices, customers, documents. A request in plain language comes back as an answer, an offer or an order, and every answer is grounded in the company's own data and cites what it used.
For Algabeth: which granites are in stock with a light finish, what 20 m² of Kashmir White costs delivered to Cluj, whether a marble is safe for a bathroom. Thumbs up or down and notes on each answer feed back into the system.
Customers
Every contact, every channel.
Companies and the people in them, WhatsApp-verified numbers, customer groups and broadcasts — the records offers and orders are created against.
Each customer's conversations — website, WhatsApp, phone — sit on their record, so the whole team sees what was promised. Algabeth groups its architects and kitchen studios and tells them when a new container of slabs arrives.
Inbox
The human behind the AI.
Every channel lands in one inbox: website chats, WhatsApp 1:1 and group conversations, and phone calls with their full transcript. An agent can take a conversation over, transfer it, leave an internal note or set themselves unavailable — the AI hands off, it never dead-ends.
WhatsApp groups, in the same inbox
Many businesses run on WhatsApp groups with their customers and partners.
Queried joins them: announcements, questions, reservations and replies are all in the inbox, linked to the customers they come from, and answered from the same account the customers already know — for Algabeth, the groups with its kitchen studios and fitters.
Support
Tickets, without the tool sprawl.
When a conversation needs follow-up — a delivery date, a sample request, an invoice correction — it becomes a ticket with a number, an owner, a priority and a thread, escalated straight from chat or logged by hand.
Queried in your pocket
Wherever work happens.
The native iOS and Android apps carry the Dashboard, Query and the Inbox, with push notifications when a customer is waiting — the same data, the same assistant, from the warehouse, the showroom or the road.
And more
Automated offer creation from the company's own pricing, order placement by command, workflows with the company's own approval rules, usage analytics on what the team asks and what gets automated, Stripe billing per company, WhatsApp broadcasts and groups, a CMS for the public site, and roles and permissions. For Algabeth, a reference library of stone families, finishes, shapes and suppliers keeps the catalogue clean.
Challenges
Three things made this project demanding. Isolation: every company gets its own database and workspace, so one customer's data and conversations can never leak into another's — while the assistant still answers in seconds. Channels: the website, WhatsApp and the phone each have their own protocol, and all three had to stream — text on the widget, messages through the WhatsApp gateway, and speech on the phone, where sentence-by-sentence synthesis brought the first spoken word from 10–20 seconds down to about two. Trust: the AI only answers from the tenant's own data, cites it, and hands off to a person the moment it should.