A reference is a project that can be checked: what the client needed, what we built and how it works in day-to-day business. This page covers the client project for Instar informatika and our own products in production. We do not give estimates of time or money saved; instead, we describe what each system actually does.
Instar informatika: an AI layer on the Luceed ERP
Instar informatika is a Croatian retailer and distributor of IT equipment: computers, laptops, components, gaming equipment and peripherals. It sells through its own stores and wholesale, and runs its business on Luceed, the ERP by Tomsoft. We connected Business Booster, our business platform, to the data from Luceed and built a layer on it for reports, business questions and search.
What work the solution removes
Brand suppliers regularly ask the retailer for reports on sales and stock levels, each in its own Excel format. Without automation, that means exporting data from the ERP, copying it into each supplier's template and sending it by email, week after week and separately for each brand. Business questions are similar: to find out which goods are not moving or are running out, someone has to export the data and put a spreadsheet together.
What we built
- Nightly synchronisation. Every night, in an agreed time window, Business Booster takes from Luceed the retail and wholesale invoices with their line items, business partners, products and other master data, and the daily stock levels per warehouse. Failed steps are retried automatically, every run is logged, and the synchronisation is monitored. The connection to Luceed is read-only: nothing is written to the ERP.
- Reports by brand. Sales and stock reports are produced in the Excel format each supplier asks for, on a schedule, for example on Mondays for the previous week, and sent by email. The brands include Acer, Asus, Canon, Dyson, HP, Lenovo, Logitech and Samsung.
- Business questions in Croatian. A user can ask which products sell best, which stores have the highest sales, who the biggest business partners are, how the categories are doing, what is low in stock but still selling, which goods are not moving and what the stock looks like in each warehouse. The answer comes with figures, a chart and a short explanation.
- An optional AI commentary on the brand report: a summary, trends, suggestions for moving goods between stores, ordering priorities and risks.
- Semantic search across products, business partners and invoices, built on PostgreSQL with the pgvector extension, which finds a term even when it is not written in the same words.
Business Booster runs on servers in the European Union, in Frankfurt.
What is AI and what is not
That is a question worth asking every supplier, so we answer it directly:
- Not AI: the nightly synchronisation and the scheduled brand reports. The figures in the reports come straight from the ERP data, and the scheduled reports are not written by a language model.
- Not AI: the figures in the answers to business questions. They are calculated by database queries.
- AI: a language model classifies the question asked in plain language and writes a short explanation of the result.
- AI: the commentary on the brand report, when the user switches it on.
- AI: semantic search, which compares meaning, not just words.
What we learned
- The ERP remains the source of truth. The AI layer works on a copy of the data and writes nothing to the ERP, so an error in the AI layer cannot change anything in Luceed.
- The database calculates the figures, and the language model explains them. That is why every answer can be checked, and the model does not calculate amounts.
- Predictable work does not need AI. Reports in the supplier's format are produced by rules, and AI makes sense where an explanation or search is needed.
- The value lies in removing manual work. The suppliers' formats are fixed, so the benefit is not in how the reports look but in no longer having to export and copy data by hand.
- An integration has to report by itself when something fails. That is why the synchronisation has retries, a run log and monitoring.
How we build a layer like this on other systems is described on our AI integration with CRM and ERP systems page.
Our products in production
Alongside client projects, we develop and maintain our own products. All three are in production.
CallVista
CallVista is a platform with an AI voice agent for call centres. It handles inbound and outbound calls in Croatian and other languages and answers from a knowledge base filled from PDF, Word and CSV documents. After every call, it saves a summary, an assessment of the caller's mood and tags. When the person shows interest, it sends them a quote by email, it schedules callbacks itself, and it builds a client profile across all conversations.
AdReady
AdReady is an AI platform for brand management and the automatic creation of banner variants. It keeps the brand guidelines in one place and creates variants that follow them, so the team does not put together every version by hand.
Business Booster
Business Booster is a business platform that brings together communication, tasks, analytics and AI on ERP data. The project for Instar informatika described above is built on it.
How to check a reference
- On the first call, ask how a similar approach would look with your data and systems.
- If you would like to talk to the client behind a reference, tell us; we arrange contact on request, with the client's consent.
- For each product, see its page with a description of its features.
If you want something similar
The process is the same as for any project: a free first call, then a scope, a cost estimate and a phased plan, a first working version in 4–8 weeks, and maintenance. You follow progress in the client portal, and we host the applications we maintain on servers in the EU.
The price depends on a few things:
- which system we connect to and how its data can be accessed;
- how many reports are needed and in which formats;
- whether you need plain-language questions, AI commentary or search;
- hosting, maintenance and upgrades.
For the first call, prepare the name of the ERP or CRM you use, an example of a report you produce by hand today and a list of questions you want answered faster. If you are still looking for where your data could help you, see data and analytics strategy or custom software development.