// What you get
What does AI app development include?
You get an AI feature that retrieves from your own content before it answers, not a generic chat widget dropped on the page. The build covers: retrieval over your real data, a defined refusal policy so it says it does not know instead of inventing, audit logging you can read, a hard spend ceiling and abuse protection, and an admin panel with logs. The model, whether OpenAI, Anthropic or an open one, is chosen per project and stays swappable. The code, prompts and database are yours, hosted on your own account.
Retrieval over your data
The model answers from your real content and catalog, so replies stay grounded instead of made up.
Refusal policy
A defined boundary so the assistant says it does not know rather than inventing an answer.
Spend ceiling
A hard-coded cost limit, per-user rate limits and a kill switch, set before launch, not after the first bill.
Audit logging
Every prompt and reply is logged, so you can read what was asked and what the AI answered.
Admin panel
A review surface where you see usage, spend and logs, and can turn the assistant off in one click.
You own it
Code, prompts and database are yours, on your infrastructure, with no license fee to keep it running.
Models and stack we build on
// Pricing
How much does AI app development cost?
A working AI build with one use case, one model and an admin panel is $1,647, given as a fixed price before work begins. Optional yearly maintenance is $1,097 and covers model version testing, swapping and behavior monitoring. Model usage is billed by the provider directly to you, at cost, with no markup. That keeps the price of the build predictable and the running cost transparent.
The price does not drift mid project. If the scope changes, you get a new quote before anyone continues, not a line item on the invoice afterwards. Adding an agent that takes actions, or connecting more systems, is scoped and priced separately so you always know what you are paying for.
The build is $1,647 for one use case and one model with an admin panel. Yearly maintenance is $1,097 and is not required. Model usage is billed to you directly by the provider, at cost. The maintenance agreement runs for one year, is paid yearly in advance, has no cancellation period and does not renew by itself.
AI build
One use case, one model, an admin panel with logs.
$1,647 build
Fixed price set before work begins. Yearly maintenance is $1,097 and is not required. Model usage is billed to you by the provider at cost. The maintenance agreement runs for one year, is paid yearly in advance, has no cancellation period and does not renew by itself.
Retrieval over your own data with a defined refusal policy
Hard spend ceiling, per-user rate limits and a kill switch
Audit logs and an admin panel that you control
First usable version in 14 days, code and prompts are yours
All prices are in US dollars, with no hidden items. The price covers the scope stated in the quote. If the scope changes, you get a new quote before anyone continues, not a line item on the invoice.
// Timeline and process
How long does AI app development take?
A first usable version ships in 14 days from the moment we have your data and a clear task. The most common cause of delay is not the model, it is the content: which data the AI may use and what it must never say. We ask for that on day one, not the last day. If the data is not ready, the timeline moves and we tell you right away, not on the delivery date.
Step 1
Intro call
Thirty minutes, no charge. We define the task the AI should do, the data it may use and what it must never say.
Step 2
Quote and sign-off
Fixed price, exact scope and timeline, in writing, before anyone starts to build.
Step 3
Build on a test version
You see a live test version from the first days and review answers and logs while the work is running.
Step 4
Launch and monitoring
The AI goes live, access moves to you, and spend, usage and behavior are watched from day one.
// Selected work
What have we shipped so far?
Real projects, still in production: Xaslarbet, sports analytics with ML models and an AI chatbot that answers from a knowledge base, plus a mobile app; Srpske novine, a news portal in seven languages with PWA and web push; Afia Marketplace, a multi vendor platform with Stripe payments, an admin console and a mobile app. The same people who built these would build yours.
Xaslarbet
Sports analytics with ML models and an AI chatbot
Srpske novine
News portal in seven languages, PWA and web push
Afia Marketplace
Multi vendor platform with Stripe payments and mobile app
// Questions
What do clients ask most before ordering an AI app?
Five questions come up on almost every first call: will the AI make things up, who owns the system when it is done, can you add AI to an existing site, why hire a small EU team over the cheapest offshore quote, and what you need to start. The answers are here, in full, and they do not change once you become a client.
Will the AI make things up about my business?
+Not if it is built to retrieve before it answers. We connect the model to your actual content and instruct it to answer only from what it retrieved, and to say it does not know when the answer is not there. We also log what it was asked and what it replied, so you can audit it. That refusal policy is part of the build, not an add-on.
Who owns the AI system when it is finished?
+You do. The code, the prompts, the retrieval setup and the database are yours, hosted on your own infrastructure or an account in your name. There is no license you keep paying to keep the assistant online. If you ever want to move it or hand it to another team, everything is yours to take.
Can you add AI to a website I already have?
+Usually yes, and it is often the cheaper path. If the site is WordPress or WooCommerce, we integrate against what is already there rather than rebuilding it. The AI reads your existing content and catalog, and the admin panel sits alongside your current dashboard, so you do not start over.
Why hire a small EU team instead of the cheapest offshore quote?
+Because the cheapest quote rarely matches the final invoice, and AI projects punish that gap. You talk to the developer building it, not a layer of account managers. Decisions are documented, timezones overlap for real work hours, and the spend controls that keep model costs sane are set up by the people who understand them. Fixed price is fixed here.
What do you need from me to start?
+Three things: the task the AI should do, the data it may use, and what it must never say or do. With those, we can scope the build and give you a fixed price. You do not need a technical spec or a data science team, just a clear picture of the job and access to your content.