People copy data from one screen to another
Orders, invoices, addresses, stock levels — the same information entered two or three times, in different systems.
One hour lost per day per person is more than 250 hours a year, each.
Automation · AI agents
AI automation that takes repetitive work out of your company
We apply AI where it saves hours every day: request triage, document processing and workflows that connect your systems.
AI is not the product. It is the decision layer inside a workflow that otherwise runs on integrations and clear rules. We build automation measured in hours recovered, not in impressive demos.
We reply the same working day. No obligation.
12 years
building digital systems
120+
projects delivered
4–8 weeks
from analysis to the first workflow in production
3+ years
average client relationship
Situations
Three signs your company needs automation.
Requests get lost or reach the right person late
Emails, forms, chat messages. Someone reads them, sorts them and forwards them by hand, whenever they have a moment.
Your staff already use AI, but on their own
Everyone with their own account, no policy, no standard, no measurement. Inconsistent results and company data going into public tools.
Benefits
Repetitive decisions handled by AI
Interpreting a request, routing it, and extracting data from documents.
Workflows that connect your systems
Site, stock, invoicing, CRM and email — joined up, with no human in the middle.
Documents processed automatically
Invoices, orders and contracts: the data is extracted and lands straight in the system.
A result measured in hours
Every workflow has an estimated time saving before we build it and a verified one after.
What you get
What an automation project includes.
We do not deliver an abstract AI solution. We deliver a workflow that runs on its own and that you can stop, change or extend.
Types of automation
Where AI delivers a real result.
Request triage
Emails, forms and messages interpreted automatically and routed to the right person or department, with a summary attached.
Document processing
Invoices, orders and contracts with data extracted and entered into the system automatically, with validation on exceptions.
An AI assistant on your site
Trained on your company data: services, prices, opening hours, policies. It answers, qualifies and books.
Agents for operations
Order validation, flagging anything unusual, updates between systems, reports that write themselves.
Sales workflows
Leads qualified automatically, proposals prepared, follow-ups sent at the right moment.
Automation audit
We analyse your current processes and tell you what is worth automating, in what order and with what expected gain.
Process
Five steps, no experiments at your expense.
We map the process
We follow one real workflow from input to outcome and measure the time spent at each stage.
We propose and estimate
What gets automated, with what technology, how many hours and how much time it saves. All written, by module.
We build the first workflow
We start with one — the one with the best ratio of effort to gain. You see it running in production.
We validate on real data
It runs in parallel with the manual process until the results are consistently correct.
We extend and monitor
The second workflow costs less: the infrastructure already exists. Monitoring stays on.
How we build
AI where there is a decision. Clear rules everywhere else.
A good workflow is not AI from end to end. It is a combination of classic automation, which is cheap and predictable, and AI used exactly where judgement is required.
Our approach
Automation with an AI decision layer
Most steps in a process are deterministic: if X arrives, do Y. Those are handled with classic automation, which costs little and does not make mistakes. AI comes in only where there is a decision a person makes today: what does this request mean, where should it go, what data does this document contain.
- We start with a single workflowWe do not transform the company overnight. We pick the process with the best effort-to-gain ratio and take it to production.
- Every workflow has a numberTime saved, estimated before and measured after. If it cannot be measured, we do not build it.
- Human control where it countsThresholds decide what runs on its own and what goes for approval. The AI does not get a free hand.
- Built on open infrastructuren8n, Make or custom code. The workflows stay yours, not locked into a closed platform.
- The second workflow costs lessInfrastructure, connectors and credentials already exist. You scale organically.
- Error handling from the startRetry, alert, hand over to a person. A workflow that fails silently is worse than no workflow at all.
On the Cabinero project we built an AI assistant that interprets the patient request, routes it to the right specialism and books the slot in the calendar — with no phone call and no operator.
Worth knowing: not every process is worth automating. On a recent project we advised the client to solve half the problem with a better designed form before spending anything on a conversational assistant.
Process automation
Deterministic workflows between applications: new order to invoice to shipping label to notification to stock update. Cheap, fast, predictable.
Suited where the steps are clear and no interpretation is needed. Most quick wins come from here.
Document processing
Extracting data from invoices, orders or contracts arriving in different formats, with automatic validation and exception flagging.
Suited when you receive tens or hundreds of documents a month and somebody keys them in by hand.
Conversational AI agents
Assistants trained on your company data that answer, qualify the request and take action: book, quote, escalate.
Suited when the volume of repetitive questions eats into sales or support time.
| Classic automation | Automation + AI | AI agent | |
|---|---|---|---|
| Suited to | fixed steps, structured data | interpretation plus execution | conversation and decision |
| Initial cost | low | medium | medium to high |
| Running cost | very low | low | variable, usage based |
| Predictability | total | high | needs clear limits |
| Time to production | 1–3 weeks | 3–6 weeks | 4–10 weeks |
| Human oversight | not required | on exceptions | required at thresholds |
Our rule: We always start from classic automation, because it is cheap and does not make mistakes. We add AI only at the points where the process needs judgement. An honest supplier also tells you when AI is not the answer — and we do that in writing, in the proposal.
Measurement
Automation without numbers is just a modern-looking expense.
Every workflow we build has an indicator agreed before implementation and verified afterwards. Otherwise there is no way to know whether it was worth it.
A concrete example: on the Hypericum project, automatic synchronisation between stock software and the online store covers hundreds of product lines, with stock and prices updating without human intervention.
What is included
What you get in an automation project
Projects
View full portfolioWe do not promise. We demonstrate.
Systems delivered, in production, running every day.
Frequently asked questions
How is the price set?
We work on an hourly rate: €50 per hour, excluding VAT. The price comes from the number of estimated hours rather than from a fixed package. You receive a proposal in which each module carries its estimated hours and cost. If something extra comes up, we estimate it separately and you approve it first.
What does an AI agent actually do in a company?
An AI agent takes over a repetitive decision a person makes today: it interprets a request arriving by email, form or chat and routes it to the right person; it extracts data from a document and enters it into the system; it validates an order and flags it if it looks unusual. It does not replace the process, only the judgement step inside it.
Where do we start if we do not know what to automate?
With an automation audit: we follow your current processes for a week or two and give you a prioritised list — what is worth automating, in what order and with what expected time saving. You can implement it with us or with anyone else.
How soon do we see a result?
A simple workflow reaches production in 3–6 weeks from approval. We recommend starting with one, the one with the best ratio of effort to gain. The second then costs significantly less, because the infrastructure already exists.
What happens to our company data?
We establish from the start what leaves the company and what stays internal. For sensitive information we use local processing or providers with contractual terms preventing the use of your data for training. The policy is written into the project documentation rather than assumed.
Can the AI make mistakes? What happens then?
Yes, it can. That is why every workflow has thresholds: what runs automatically and what goes to a person for approval. Exceptions are flagged, not hidden. A workflow that fails silently is more dangerous than no automation at all.
What technologies do you use?
n8n and Make for workflows, custom code in PHP or JavaScript where specific logic is required, plus AI APIs for the interpretation and processing components. Integrations with existing systems run over REST or SOAP.
Do we have to replace the systems we already have?
No. Automation is built on top of them. If your stock software exposes an API or a web service, it can be integrated. See the ERP, CRM and API integrations page.
Can this be funded through a grant?
Yes. Process automation and AI components frequently qualify under national and EU digitalisation programmes. We build the technical side and prepare the documentation alongside your funding consultant.
What happens after implementation?
We stay on for monitoring and adjustment. Processes change, and the workflows have to keep up. Our client relationships average more than three years.
Estimate
Tell us what you need. We reply with an estimate.
Five steps, under two minutes. The more context you give us, the closer the estimate will be to reality.
ProjectStep 1 din 5
What would you like us to build?
You can pick more than one.
Tell us what is eating your time.
You do not need to know which technology suits you or whether AI is required. Describe the process that hurts most. We reply with what can be automated, in what order and with the estimated hours.
Get an estimateOr email us directly: contact@divasweb.ro+40 755 336 514