Nikola Kočić and his team are building an Agent Engine on which banks, telecoms and regulated industries run their own artificial intelligence, without sending a single piece of data to someone else's server. While the world chases the smartest model, one company from Serbia is solving a problem SaaS providers cannot even touch: how to bring AI into organizations whose data must not leave the building.
The same obstacle everywhere
The story goes that everyone who has ever tried to bring AI into a bank, a public institution or a government agency has hit the same wall. The problem is not that they do not want AI, they do. The problem is that data must not go to someone else's server. SaaS providers offer a cloud solution, nicely packaged, with all the features, but when you ask where the data goes, the answer is always the same: to our server. For a small company that is perfectly acceptable. For an organization that handles citizens' data, financial data or sensitive internal information, it is simply not an acceptable option.
Nikola Kočić recognized exactly that angle. He is an entrepreneur with more than a decade of experience in building brands and selling across Europe, the UAE and the USA. Through all those years of working in highly regulated sectors, one problem kept recurring: the organizations that need AI most are exactly the ones that cannot use it through the SaaS model. Not because they do not want to, but because they are not allowed to.
An engine, not a platform
From that need came Growww AI Engine, the first PaaS approach to sovereign AI in Serbia. And here is a distinction worth clarifying: Growww AI is not a platform in the classic sense, a finished product sold ready made to every client and into which the client has to fit. It is an Engine, an engineering core that is embedded directly into the client's infrastructure and on which chat, voice and multi-agent solutions are built for each organization separately, tailored to its processes. Instead of the client adapting its work to someone else's product, Growww Engine adapts to the client.
In practice, this means the whole Engine is installed on the client's own infrastructure. The model runs inside their firewall. Data never leaves their environment. The client holds the compute, the storage, the network and every byte that flows through the system. In addition, Growww AI is the first company in the region to have integrated AI voice into its own Engine, a voice AI that holds a full conversation with the user in real time, not just chat.
Why SaaS does not work where confidentiality matters most
Imagine the following situation. A public institution wants to automate responses to citizens' inquiries. They have thousands of requests a month, the team is overloaded, answers are late. AI would solve it overnight. But the question the director asks is not everything AI can do. The question is: where does the citizens' data go? If the answer is to a server in Germany or the USA, the conversation is over. The Law on Personal Data Protection, internal security policies, audit requirements, all of it stops at that question.
Most providers do not understand this problem. They build better AI, faster AI, smarter AI. But they do not build AI that can work where it is needed most. Nikola saw this first hand, working with organizations where every conversation about AI hit the same wall. The technology was ready. The budget was there. But the data could not leave the building.
How the Engine changes the rules of the game
Instead of sending data to the cloud, the Engine brings the cloud to you. The whole system is installed on your infrastructure or in a local data center certified to domestic standards. All data, all communication, all models, everything stays inside your environment. No external access, no sending to other people's servers, no dependence on connections to external services. The AI becomes part of your internal infrastructure, just like any other system you already use.
Models we have already built for clients
On the Engine we have so far built several different models for clients, each tailored to its own operational needs, from internal knowledge platforms, through multi-agent systems for large field teams, to AI support in regulated industries.
The first model is an internal knowledge platform that runs under the client's brand, on their own infrastructure. The system uses RAG for knowledge management: the team enters the company's institutional knowledge and then searches it through the agents we developed. One such agent guides new employees through the entire onboarding, gives them the company's accumulated knowledge and answers questions in real time. The Engine license and AI tokens go directly through the client's infrastructure, and their team uses the system with no additional per user cost. No data leaves their servers.
The second model is a multi-agent system built for an organization with more than 100 field workers, where internal communication had serious gaps. We deployed four agents working together: a lead generation chat on the website, with advanced mechanisms for fast acquisition and minimal data entry; an AI call center that takes calls outside working hours, from 10 p.m. to 7 a.m., and prepares leads for the human call center in the morning; a knowledge agent for the sales team, with instant access to information about services, procedures and terms; and an agent that monitors all call center communication, tracks completed and uncompleted tasks, and at the end of the day detects where communication breaks down between the call center, the organizers and the field. This is not a chatbot, this is operational AI infrastructure that sits inside the client's network.
The third model is AI support built for one of the largest German operators in its industry, which also operates in Serbia: chat and voice AI on the website and in the app, a knowledge agent that guides the counter worker through the fault resolution process using RAG based knowledge from the manufacturer's documentation, and a voice agent for customer support. If the AI cannot solve the problem, the system automatically opens a ticket and a technician arrives at the location the next morning. Every part of this system runs inside infrastructure the client controls.
Why now is the moment
The AI industry is entering its production phase. Experimentation is over. Companies no longer want pilots and proofs of concept, they want AI in production, integrated into operations, with measurable results. The obstacle is not the model. It is the infrastructure.
The organizations that will lead the next phase are not those with the best models. They are those with the infrastructure to deploy those models safely, locally and under their own control. Sovereign AI is not a niche. It is becoming the standard for every serious deployment, and the solutions that can deliver it will win the market.
Read the full article at BIZLife: https://bizlife.rs/growww-ai-engine-kako-srpska-kompanija-sve-podatke-zadrzava-unutar-zidova-velikih-organizacija/
Find out how Growww AI Engine works inside your organization at growww.ai.
