A mobile app is always in the user's hand. It has a camera, a microphone, and it knows where the person is. That opens more room for AI than a website does. The rule is still the same: AI must solve a specific user problem, not serve as a marketing label. Below are the features we have actually built or recommend most often.
An AI assistant inside the app
The assistant answers user questions right inside the app, based on your services, orders and rules. It can also take actions: find a free slot, file a request, show the status of an order.
For example, Consulo (iOS, Android and Web) connects doctors and patients, includes an AI assistant and serves several clinics, each working in its own time zone.
Personal recommendations
The app remembers what a user has viewed, bought or saved, and suggests relevant products, content or services. Notifications become personal too: people get what actually interests them and switch notifications off less often.
Size and product advice
Returns caused by the wrong size are one of the biggest problems in online fashion. A size advisor takes the user's data and the product's specifics into account and recommends the right size. We built one in a fashion store app that also has Apple Pay, Google Pay and real-time order tracking.
The same idea works elsewhere: furniture, cosmetics, electronics. The user answers a few questions and gets the option that fits.
Real-time translation
When two people speak different languages, AI can translate their speech during the conversation. We built an app that translates calls in real time, and the client mentions this work in their review. Such a solution is useful in tourism, healthcare, logistics and international sales.
Document and photo recognition
The user photographs a passport, an invoice, a receipt or a contract, and the app fills in the required fields by itself. Manual entry errors drop, and a process that took minutes takes seconds.
The same technology works on the server side. CargoFlow reads PDF orders from email (OCR), extracts the data, links it to drivers and vehicles, and calculates the route and pay automatically.
On the device or in the cloud?
AI can run directly on the device or on a server ("in the cloud"). The choice depends on the task.
| On the device | In the cloud |
|---|
| Speed | Instant, no network delay | Depends on the connection |
| Works offline | Yes | No |
| Privacy | Data never leaves the phone | Data is sent to a server |
| Model power | Limited, small models | The most capable models |
| Updates | With a new app version | On the server, no app update needed |
In practice we often combine both. Reading text from a photo or pre-processing voice happens on the device, while complex answers and recommendations come from the server. The app stays fast, and as little personal data as possible is sent out.
Price ranges in 2026
| Type of work | Example | Timeline | Approximate cost |
|---|
| AI feature in an existing app | assistant, recommendations or document recognition | 2–4 weeks | 400,000 – 1,200,000 AMD |
| Simple app with AI | catalogue or booking with an AI assistant | 4–6 weeks | 1,800,000 – 3,500,000 AMD |
| Medium complexity | orders, payments, personal recommendations, recognition | 6–10 weeks | 3,500,000 – 7,500,000 AMD |
| Complex system | several roles, real-time translation, integrations | from 2 months | from 7,500,000 AMD (from about $19,500) |
The figures are approximate. We give the exact price and deadline after a free estimate in 2–3 days; they are fixed in writing and do not change. You can pay after you receive the complete working system and confirm that everything matches your requirements. Larger projects can also be split into stages, each paid after acceptance.
What drives the price
- Where the AI runs. On-device models need more testing across different phones.
- Number of features. Payments, notifications and integrations add work.
- Backend and admin panel. Almost every AI app needs them.
- Languages. Translation and recognition must be tested in each language.
- Amount of design. Ready-made components or a unique design.
Projects like this usually take longer on the market. Our speed comes from experience: the native iOS and Android app for ART Smart Repair went from first design to App Store approval in about 2.5 weeks. The ProMobile app is used daily by 200+ employees across 140 stores, with 19 roles and chat messages delivered in under 1 second. And according to one client's review, the mobile app we built for their online shop increased sales by 40%.
How we work
We sign an NDA before discussing details. You pay only after you receive the complete working system and confirm that everything matches your requirements. On larger projects payment can follow accepted stages. Design revisions are unlimited at the design stage. At handover you receive technical documentation listing the languages and versions used. The source code belongs to you. We follow international standards and provide a warranty and long-term support.
More on the AI and automation page.
Frequently asked questions
Can AI work in an app without an internet connection?
Yes, if the model runs on the device. That suits text recognition, simple classification and voice processing. More complex tasks need a server.
Can AI be added to an app that already exists?
Yes. In most cases AI connects as a separate module and the app does not need to be rewritten. It usually takes 2–4 weeks.
Will AI make the app heavier or drain the battery?
Cloud AI has almost no effect on app size or battery. On-device models are chosen to be small and are used only when needed.
How fast can a new app with AI be launched?
A simple app takes 4–6 weeks, a medium one 6–10 weeks, a complex system from 2 months. The final timeline depends on features and integrations.
Who owns the app's code?
You do. At handover you receive the source code and the technical documentation.