Atradius
Trade credit insurance
AI, SOFTWARE & CLOUD
AI takes shape. Inside your systems.
We connect models, data and software to remove friction from your processes. From the first architecture to production, with the people writing the code by your side.
The company and the track recordTransparency note
Deexma Labs s.r.l. was founded in 2026. The ten years of projects predate its incorporation: they are the experience the company carries with it, not its own record. The distinction matters, and we state it in full here.§
Explore the system
- 10+ years on critical systems
- Insurance, fintech, banking
THE FIRST CLIENTS OF DEEXMA LABS
Deda.next
Italian IT group — public sector and utilities
Only the company's own clientsTransparency note
Only clients of Deexma Labs s.r.l. appear here. The work of the deexma project that preceded it and the experience the founder built before deexma are not added to this list: they sit in the other notes and in “Who's behind it”, declared for what they are. Trust is built by not blurring them.§
OUR VISION
AI creates value when it becomes part of the process. When a document becomes an answer. Data becomes action. And people get back time to create.
From data to action. All the way to production.
We integrate AI into your processes and build the software behind them. Three areas of expertise, one purpose: making things work, every day.
01 AI integrations
AI, inside your systems.
We connect LLMs and agents to your existing ERP, CRM and portals. Your processes evolve; your data stays where it belongs.
From an order email to data ready for your ERP: customer, items and deadlines. See it in the example below.
Explore the service02 Assistants and automation
Put your knowledge to work.
Assistants that search your documents and answer with sources. Agents that handle the next steps, with human oversight where it matters.
A salesperson asks your technical catalogue a question and gets an answer, together with the source document.
Explore the service03 Development, cloud and data
Software built for real work.
Applications, APIs and microservices. Cloud infrastructure and data architecture. Enterprise experience, including projects that need no AI.
From deexma’s work before the limited company: production microservices for the insurance division of a national group.
The enterprise work behind usdeexma project (before Deexma Labs)
Cloud-native microservices for the insurance arm of a large Italian group, delivered by deexma before the limited company was incorporated. Of that client we describe the sector and the work, not the name: a name appears here only when the client agrees to be cited.§
THE PROOF
What happens when AI works on your documents?
An email comes in and work moves forward: an order, a quote request, a payment reminder become structured data, ready for your ERP with no re-typing. Not a product to try — an example of what we build in your systems.
Transparency notice, Article 50 EU AI Act: You are about to interact with an AI system: it is not a person and it can make mistakes. The data you enter is not stored.
Input
From: [email protected] Subject: Order no. 2041 — confirmation Hello, we confirm the order for 12 Fjord desks (code FJ-120) and 24 Anni chairs (code AN-40), delivery by 30 September to the Pomezia warehouse. Payment within 60 days as agreed. Best regards, Bruni Arredamenti
OutputPreloaded example
- Type
- Order
- Customer
- Bruni Arredamenti
- Order
- no. 2041
- Items
- 12 × Fjord desk (FJ-120)24 × Anni chair (AN-40)
- Delivery
- by 30 Sep 2026, Pomezia warehouse
- Payment
- 60 days
Suggested action: Record order no. 2041 in the ERP and prepare the order confirmation for Bruni Arredamenti.
How your ERP receives it
{
"type": "order",
"customer": "Bruni Arredamenti",
"order_number": "2041",
"items": [
{
"code": "FJ-120",
"description": "Fjord desk",
"quantity": 12
},
{
"code": "AN-40",
"description": "Anni chair",
"quantity": 24
}
],
"delivery": {
"by": "2026-09-30",
"location": "Pomezia warehouse"
},
"payment": "60 days"
}Input
From: [email protected] Subject: Quotation request — systems maintenance Hello, we are a restaurant chain with 4 locations and we are looking for an annual maintenance contract for our refrigeration systems, with a response within 24 hours in case of failure. Could you let us know costs and lead times? Thank you, L'Officina del Gusto
OutputPreloaded example
- Type
- Quotation request
- Requester
- L'Officina del Gusto
- Need
- Annual maintenance of refrigeration systems
- Locations
- 4
- Constraints
- response within 24 hours of failure
- To ask
- location of the sitesmake of the systems
Suggested action: Draft a quotation for 4 locations and ask for the site locations and the make of the systems before the site visit.
How your ERP receives it
{
"type": "quotation_request",
"requester": "L'Officina del Gusto",
"need": "Annual maintenance of refrigeration systems",
"locations": 4,
"constraints": [
"response within 24 hours of failure"
],
"missing_information": [
"location of the sites",
"make of the systems"
]
}Input
From: [email protected] Subject: OVERDUE NOTICE — invoice 2026/118 past due Hello, we are writing to remind you about invoice no. 2026/118 dated 15/06/2026, which fell due on 15/07/2026, for € 4,280.00. A copy of the invoice is attached. We look forward to hearing from you. Tessiture Valli S.p.A.
OutputPreloaded example
- Type
- Payment reminder
- Supplier
- Tessiture Valli S.p.A.
- Invoice
- no. 2026/118 of 15 Jun 2026
- Due date
- 15 Jul 2026
- Amount
- €4,280.00
- Attachments
- invoice copy
- Priority
- high
Suggested action: Check the status of invoice 2026/118 with accounts and schedule the payment or the reply to the supplier.
How your ERP receives it
{
"type": "payment_reminder",
"supplier": "Tessiture Valli S.p.A.",
"invoice": {
"number": "2026/118",
"date": "2026-06-15",
"due_date": "2026-07-15",
"amount_eur": 4280.00
},
"attachments": ["invoice copy"],
"priority": "high"
}If the model is unreachable, the preloaded examples still show the full output.
Does this look like your process? Tell us about the input, systems and intended outcome, and we will reply to discuss whether an audit is the right first step.
Talk to usThe architecture
And inside your systems — where does it run?
There is no new platform to buy and no archives to move: we add a thin layer between what you already have and what needs to happen.
What you already have
- ERP or management system
- CRM
- Email and certified mail
- Document archive
The layer we add
- A connector on your own APIs
- Extraction and verification of the data
- Rules and thresholds agreed with you
- A human in the loop where it matters
What comes back
- Rows ready in your system
- Drafts to approve, not actions already taken
- A trace of every decision
- EU data residency: it is our default in every project.
- Your archives stay put; the model only receives what a single request needs.
- Code and documentation remain yours.
The time calculator
Your process, in numbers
You have seen a document become data. Now enter the three values from your process: the comparison updates instantly and stays in your browser, without sending anything.
The maths assumes 5 working days and that extraction succeeds: documents handed back because the model is not confident are retyped by hand and counted separately. All three values are yours, the checking time included — raise it and the margin shrinks, which you should see before rather than after.
THE METHOD
Find the direction. Prove it. Then scale.
Audit, pilot, production. Each stage delivers something tangible and the evidence to decide what comes next.
Audit
01 · 1 week
We map your processes, data and systems. We identify the opportunities with the most potential and define how to measure their value.
What you get: Tailored roadmap — the first draft within 24 hours of the kick-off meeting.
Pilot
02 · 4–6 weeks
We build a 4–6 week pilot around one real process. We compare the results with the agreed criteria: moving forward depends on meeting them.
What you get: Pilot measured against the criteria agreed at the start.
Production
03 · ongoing
We bring the system into production. Integration, monitoring and human oversight become part of the architecture, with the ability to intervene and roll back.
What you get: System in operation, with controls and documentation.
Where the 24-hour promise comes fromTransparency note
The draft roadmap within 24 hours of the kick-off meeting is not a slogan: it is the practice already in use at deexma, before Deexma Labs s.r.l. was founded. It contains the candidate processes, the recommended order and the first step — enough to decide, not to replace the audit.§
One week to identify where to act. A roadmap to decide how.
WHO'S BEHIND IT
You talk to those who build
From the first conversation to release, you work with the people designing the architecture and writing the code. Responsibility stays in the same hands.
Cristian Mariano — founder. Ten years of development and technical leadership on critical systems in insurance, fintech and banking.
The timeline in fullFounder's prior experience
The track record in full: J2EE development on insurance systems, then full-stack, then technical team leadership, then application ownership of enterprise products. Every step on projects in production, never on prototypes.§
In the ancient Greek theatre, the deus ex machina was the stage machine that lowered an external, unexpected and seemingly impossible solution into the scene — one capable of resolving a very complex problem with simplicity. Twenty-five centuries later the machine has changed — today it is software, AI — but the principle is the same.
FAQ
Answers before you get started.
What does it cost to start?
You start with the audit: one week, a price fixed before we begin, a written roadmap at the end. From there you decide whether to continue with the pilot. No flat-fee quotes on unmapped projects: first we measure, then we estimate.
Where does our data end up?
In your systems, or with providers chosen together and disclosed up front — our default is data residency in the EU. You always know which providers are involved and under which safeguards, DPAs included. Your data is never used to train models.
How long before we see something concrete?
The first draft roadmap within 24 hours of the kick-off meeting; the pilot in 4–6 weeks. If the audit shows that AI does not pay off, we tell you: it is a legitimate outcome, and it costs you a week instead of a project.
Will we be locked in?
No. Code, prompts, configurations and documentation are yours, on open standards. If tomorrow you want to take everything in-house or to another supplier, you can. Lock-in is a business model, not ours.
What if the model gets it wrong?
It will, and we plan for it: human checkpoints where errors are costly, continuous quality measures, the ability to roll back. The AI we sell is the AI you know how to switch off.
Do you also do software without AI?
Yes, and it is half of our work: Java and Quarkus development, microservices and APIs, Kubernetes and OpenShift, databases and data governance. They are services in their own right, not a hook to sell AI: if the problem is solved with good software, that is what we propose.
Your process is the starting point.
Tell us what slows your work down or what you want to build. That holds even if the problem is not AI but software, integration or infrastructure.
What happens next
- The reply comes from whoever will then carry out the work.
- If there is an honest way to help, our reply will suggest the smallest useful step: a call, or the audit directly.
- If there is not, we will tell you in our first reply — not after three calls.
Contact request
Request sent.
To add anything, write to [email protected].
Prefer to write? [email protected]