AI Solutions
BIGBANG ITS designs and builds AI agents, automations, websites and apps on Claude, the model family we know best. You pay for capacity in two units, engineering hours and a token allowance, so every month shows what was used and why.
AI delivery you can measure, billed in hours and tokens
Our work is practical AI: agents that act inside your systems, automations that remove manual steps, and AI-assisted websites and apps that reach production. Claude is our specialism. Rather than an open-ended retainer, you buy a set number of engineering hours and a pool of tokens, and both can be spent on any AI task you hand us.
A monthly statement shows each meter separately. No item is charged twice, and no hour or token is used unless an approved work item sits behind it.
The eight requests we hear most often
These are not trend words. They are the projects clients actually bring us, listed from the most frequent down.
Websites and apps built with AI
Complete products taken from specification to live release. AI shortens the build, but code review, testing and a named person accountable for the outcome all remain.
Agents tailored to your business
Agents that work with your real tools, such as the CRM, the database and the file store, across multi-step tasks, and that stop for human approval before any action that cannot be undone.
Assistants that read your documents
Retrieval-based assistants that answer from your own contracts, policies and manuals, and point to the exact page each answer came from.
Paperwork turned into clean data
Invoices, purchase orders, commercial registrations and trade licences, ID cards and delivery notes are read, checked and passed straight into the system that uses them.
Conversational AI that starts in Arabic
Assistants on WhatsApp, the web and the phone that talk in Gulf Arabic and move into English in the middle of a sentence, just as real customers do.
AI connected to the systems you already use
ERPNext, Odoo, WHMCS, WordPress, Microsoft 365 and in-house back-ends. Most of the value comes from the integration rather than from the model itself.
Testing and guardrails
Before launch, a scored test set replaces the feeling that it seems fine with a measurable result you can hold us to.
Hosting and where data lives
We operate our own servers. If your data must remain under a particular jurisdiction, raise it during scoping and the design will respect it.
Claude is our platform, inside our own company too
Many agencies list three model vendors and switch to whichever is cheapest this week. We chose depth in one, because depth is what lets an agent run unattended and still be trusted.
Agents that act, not only chat
Tool calling, the Model Context Protocol, structured output and long-running agent sessions: the layer that turns a chat window into a worker that finishes tasks.
Context of one million tokens
An entire codebase, a complete tender file or years of correspondence can be read at once, without being cut into pieces that lose the connection between parts.
Our own sites are the proof
Our UAE site, its.ae, is bilingual, dark by default, maintained with AI agents and linked to our billing portal through tools we wrote ourselves. The method we sell is the one we run on.
One invoice, two meters
Engineering hours
Time from our engineers for scoping, prompt and context design, building, integration, testing, review, deployment and handover, recorded in half-hour steps against a work item you signed off.
Token allowance
The model usage your work generates, both in our development runs and once your system is live. A single allowance covers both, so no second bill arrives from another vendor.
Keeping usage low is our job
We cache prompts, send simple steps to smaller models and batch anything that can wait. Since tokens are part of the package, saving them benefits us as much as you.
Capacity you can count, not promises
Each package includes the scoping call, the evaluation set, the full source code and complete handover documents. Current package prices are shown in the client portal.
Pilot
Test one workflow and learn whether it works before committing to anything bigger. One workflow scoped and built Compared with how you work today A written recommendation to proceed or stop Reply within 2 business days
Build
A working production system for one department and one defined job. Deployed on our infrastructure or yours Connected to one existing system An evaluation set that stays with you Two training sessions for your staff Reply within 1 business day
Scale
Several workflows on live traffic, with someone tracking the figures. Up to four workflows in production A monitoring dashboard and cost report Monthly tuning on real data Arabic and English coverage Same-day reply during office hours
Embedded
A named engineer working inside your team every month. A named engineer who joins your stand-ups 25% of unused capacity carries over Private or on-premise deployment A roadmap review every quarter Priority queue with 24/7 escalation
Extra capacity and rates
Every package can be topped up with extra blocks of 10 engineering hours or 10 million tokens. The larger the package, the lower the rate per block.
Rush delivery within 48 hours carries a surcharge on Pilot, Build and Scale; it is not offered on Embedded, which already has priority.
Work outside a package is charged per hour. Ask us for the current rate card, or check the prices shown in the client portal.
Private or on-premise model hosting, and anything needing a dedicated GPU server, is priced separately. On one-off packages, tokens left unused do not carry over; unused hours can be extended once, by 30 days, if you ask.
How far a million tokens goes
A token count only becomes useful when you can match it to a real task. The ranges below come from projects we have already delivered. Read them as guidance, not as a quote.
Task — Usual token range — Package that covers it
Questions and answers on a 100-page tender file — 0.1 – 0.3M — Pilot, with room to repeat it many times
Pulling fields out of 1,000 supplier invoices — 2 – 3M — Pilot
Support assistant across 1,000 customer chats — 4 – 6M — Build
Single agent pass over a mid-sized codebase (read, plan, change, test) — 0.5 – 1.5M — Build, about 20 passes
Translation and localisation of a 50-page site in two languages — 1 – 2M — Pilot
Internal knowledge assistant, 40 daily users, one month — 15 – 25M — Scale
WhatsApp sales assistant at 300 chats a day, one month — 35 – 50M — Embedded
The same content in Arabic needs about twice as many tokens as in English. Every quote we send already includes this; it is a known feature of bilingual work, not a cost that appears later.
Four stages, always in this sequence
Scoping call
A free 45-minute session. We trace the workflow and say openly when AI is not the right answer, and that advice costs nothing.
Pilot
A single workflow is built and measured against your current way of working, so you get a figure rather than an opinion.
Build
The live system, connected to the tools you already run, with the evaluation set ready before the launch date.
Run
Monitoring, tuning every month and a cost report written to be read. Models keep changing, so someone keeps checking that the system stays right.
AI projects for organisations in Qatar
How we work with Qatari clients, stated plainly so there are no surprises later.
Bilingual and right-to-left from the start
Assistants and interfaces are built in Arabic and English together, with proper right-to-left layouts, to suit staff and customers in Qatar who switch between both languages.
Personal data mapped against Law No. 13 of 2016
Where a workflow touches personal data, scoping records what is processed, where it is kept and who can access it, so it can be reviewed under Qatar's personal data protection law.
Honest about hosting location
We have no office or data centre in Qatar. Our servers are in Germany and Finland, AWS options cover the UAE, Germany and the USA, and Claude runs as a hosted API. You hear this at scoping, before you commit. Other locations are available on request.
Billing and support that fit Qatar
Pay in Qatari riyals through the client portal and plan support around the Friday–Saturday weekend. Reach the team at info@its.qa or on WhatsApp at +1 (667) 222-1616.
Questions buyers ask before they sign
What exactly is a token?
It is the unit an AI model reads and writes, and so the unit it is charged in. One token is about three quarters of an English word, which makes a million tokens roughly 750,000 English words, or close to half that amount of Arabic text.
What happens when a meter is about to run out?
You get a warning at 75% and another at 90%, always before a meter empties. You then choose to add a top-up block or to pause. Nothing beyond your package is used without your written approval.
Who owns the code, the prompts and the evaluation set?
You own all three once the final payment is made. Your data is never used to train any model, and if you move on, you take a system that another team can pick up.
Can the model run on our own servers?
Open-weight models can, either on our servers or on yours. Frontier models such as Claude are only available as a hosted API, so if a regulator insists that nothing leaves your premises, we tell you during scoping, before any payment.
Do you deliver in Arabic?
Yes. Interfaces, prompts, evaluation sets and documents come in Modern Standard Arabic and Gulf Arabic as well as English, and everything we hand over exists in both languages from day one, this page included.
Why do so many AI projects fail?
Mostly because nobody agreed what success looks like. The evaluation set fixes that: a shared, scored test that runs before launch and again every month, so a falling score reaches us before it reaches your customers.
Can a Qatari company pay in riyals?
Yes. Packages and top-ups are invoiced through our client portal, where you can choose Qatari riyals as your billing currency.
How do you handle personal data under Qatar's law?
When a workflow involves personal data, we document what is processed, where it is stored and who can reach it, so your team can check it against Law No. 13 of 2016 on personal data protection. The legal sign-off stays with you or your adviser.
Tell us which workflow frustrates you most
Forty-five minutes, free of charge and without slides. You walk us through how the task runs today, and we explain what AI would change, what it would cost in hours and tokens, and where it would add nothing.
Talk to our team
Tell us what you need and we will reply by email or WhatsApp within one business day.