AI automation
We build artificial intelligence into the processes you already have: email triage, enquiry processing, document reading and drafting quotes. The system does the repetitive work, while decisions stay with your employees.
How we introduce AI
Use cases
It works best for tasks that repeat every day and where employees mostly retype, sort or look up data.
AI reads incoming mail, recognises whether it is an enquiry, complaint, invoice or order, and routes the message to the right department.
Control: Unclear messages are reviewed by an employee.
It extracts the customer, items, quantities and deadlines from an email or web form and records the enquiry in your CRM or ERP.
Control: The salesperson sees the original and the extracted data.
Purchase orders, delivery notes and invoices as PDFs or scans. The data is transferred to your system without manual retyping.
Control: Uncertain values are flagged for checking.
Based on the enquiry, price list and stock, AI prepares a draft quote and a cover message.
Control: The salesperson reviews, corrects and sends the quote.
On your website it answers common questions about delivery, products and order status – based on your own content.
Control: More complex questions are handed to an employee.
Employees quickly find answers in manuals, price lists and technical documentation, together with a link to the source.
Control: Access only to documents the user is allowed to see.
Summaries, weekly reports, reminders and data transfers between systems run automatically, on a schedule or triggered by an event.
Control: Every run is recorded in a log.
Describe it. We will tell you whether it is suitable for automation and where we would start.
An end-to-end example
This is what one automated process looks like in practice. At every step it is clear who carries it out – AI, your system or an employee.
System
The enquiry arrives by email or through a web form. The system picks it up for processing automatically.
Novak Construction Ltd.
»Hello, we need 40 rolls of 50 mm sealing tape and 120 tubes of neutral silicone. Delivery by 14 November at the latest.«
AI
AI recognises that this is an enquiry and extracts the customer, items, quantities and deadline from free text.
All data recognised
System
The system checks stock, the partner's price list and lead times in the ERP. AI does not guess this data – it reads it from the source.
Prices with the partner's 12% discount
AI
AI prepares a draft quote and a cover message. It flags the partial stock and suggests delivery in two parts.
Quote Q-2026-0187 · draft
Total excl. VAT€509.20
Employee
The salesperson sees the original enquiry, the draft and the ERP data. If needed, they make corrections and approve the quote.
Added a note about delivering the remaining 24 tubes of silicone
Approved · sales
System
The quote is sent to the customer and recorded in the CRM and ERP, and the system sets a reminder in case the customer does not reply.
The data in this example is fictional.
An example from tourism
AI understands the enquiry, finds offers in supplier APIs, a local database, PDF catalogues and Excel price lists, and prepares a quote. With the customer's consent it emails the quote and records the contact in your CRM or ERP.
»I'm looking for an affordable holiday in Egypt for two adults and two children, half board, between 15 and 25 July, flying from a nearby airport.«
Enquiry
The customer describes the holiday in their own words.
Searching the sources
APIs, local database, PDF catalogues and Excel price lists.
PDF quote
A well-formatted quote sent to the customer's email.
CRM or ERP record
The contact and quote, ready for the agent's follow-up.
Control and security
In your processes, AI prepares proposals and handles routine work. How it decides, what data it can access and where it runs is something we define together with you.
Quotes, replies to customers and entries in business systems only go ahead once an employee approves them.
When AI cannot reliably recognise a piece of data, it flags the case and hands it to an employee instead of guessing.
We record what AI suggested, what the employee changed and who approved it.
We send AI only the data the task strictly requires. Sensitive data is never shared.
Processing is set up in line with the General Data Protection Regulation (GDPR), including a data processing agreement.
We use services and settings where your data is not used to train models.
We implement solutions through the APIs of established AI providers. For special requirements, a local deployment is also possible.
AI helps process data and prepare grammatically correct texts in your company's style. It receives only what the task strictly requires.
For companies that want their data never to leave their network, an open model can also be deployed on a local server. We check its suitability and quality on your own cases before you decide.
Rollout
We introduce AI gradually: first on one task and with a limited scope, so you can verify the benefits on your own data before committing to more.
Together we choose the task that takes the most time and review the cases, data and systems involved.
Result: Process selected
We introduce the solution in a limited scope. AI prepares proposals, and employees review and approve them.
Result: A working pilot
We check accuracy and time saved, and improve the instructions where AI missed.
Result: Assessment of benefits
Once the pilot works reliably, we extend it to more cases or new processes.
Result: Solution in use
We monitor operation and usage costs and adapt the solution to changes in your processes and in AI services.
Result: Reliable operation
FAQ
Have another question? Write to info@isaac.si.
Not by default. AI prepares a draft and an employee sends it after reviewing it. We introduce automatic sending only for simple, pre-agreed cases, such as confirming receipt of an enquiry – and only if you want it.
The employee sees the original and the proposal, so they notice and correct the error before approving. AI flags uncertain cases for review itself. We log errors and use them to improve the instructions.
Yes. Modern language models write grammatically correct texts in Slovenian, English, German, Croatian and other languages, and we adapt the style to your sample messages and quotes. An employee still reviews the text before it goes out.
We use AI through the APIs of established providers, with a data processing agreement and no use of data for model training. We send AI only the data the task strictly requires and never share sensitive data. For special requirements, a local deployment is also possible.
No. We connect AI to the systems you already use – email, ERP, CRM or your online store. We check which connections are possible during the process review.
It depends on the process and the systems that need to be connected. That is why we start with a pilot on one process – you get an estimate of the duration after the review, before any work begins.
Costs consist of developing the solution and using the AI service, which is billed by usage. We estimate both in advance, based on the scope of your process.
Let's start with one process
Describe it to us. We will tell you whether it is suitable for automation, how a pilot would run and what data would be involved.
Automated assistant · quotes are prepared by Igor personally