Software integrations: your programs, finally connected.
ERP, CRM, e-commerce, invoicing, WhatsApp and Teams: we connect them so they exchange data on their own. Data entered once is up to date everywhere, even in older systems without an API.
- Data entered only once
- Systems without APIs too
- EU servers
When software doesn't talk, people do the work
Every system keeps its own copy of the data. Keeping them in line by hand costs hours every week and creates errors that surface too late.
"I type it in twice."
The same order, customer or product is keyed into the online shop, the ERP and the CRM. Three entries, three chances to get it wrong.
"Which number is right?"
Stock, prices and customer records differ from one program to the next. Nobody knows which one is current any more.
"I export the file and import it again."
Every evening someone downloads a spreadsheet from one system and uploads it to another. One shifted column and everything stops.
What a software integration is, in plain words
A stable connection that moves data from one program to another, on its own and always the same way.
Every company uses several programs: the ERP for invoices and stock, the CRM for customers, the online shop for orders, email, WhatsApp and Teams to communicate. Each works well on its own, but none knows what happens in the others.
An integration connects them. When an order comes into the online shop, it lands in the ERP. When stock changes in the ERP, the shop updates. When a customer signs in the CRM, their record is ready for the invoice. Data is entered once and reaches everywhere it is needed.
The difference from an automation is subtle but important: an automation runs a process with a start and an end, an integration keeps data aligned between systems all the time. You often need both, and AI comes in when data arrives as text: PDFs, emails, messages.
Three ways to keep data aligned
| Aspect | Typing it in | Exporting and importing files | Integration |
|---|---|---|---|
| Data up to date in every program | When someone remembers | Once a day, at best | In real time or on a set schedule |
| Typing errors | Frequent | Few, but one bad file stops everything | None, with data checks |
| Work required from the team | Hours every week | Every day | None, except exceptions |
| Notices when something is off | Depends | Rarely | Yes, and alerts a person |
| History of what moved and when | No | Partial | Yes, every exchange is logged |
How systems get connected
There is no single kind of integration. We pick the connection based on how fresh the data must be and what the programs allow.
In real time
The program notifies the integration as soon as something happens (a webhook): a new order reaches the ERP within seconds.
Two-way sync
Customers and products stay identical in both systems, whoever edits them. For each piece of data we decide which system is the source of truth.
On a schedule
For large volumes or slow systems, data aligns at fixed intervals: every hour, every night, every morning at 7.
Custom connector
When no ready-made connection exists, we build one on the program's API, with tests and documentation.
Files and shared folders
For systems that can only export: we read CSV or Excel files from a folder or an SFTP server and turn them into clean data.
Database access
For ERPs without an API, we read straight from the database, read-only where possible and with dedicated credentials.
Systems we connect
Some of the programs most used by companies in Italy and the Czech Republic. The list is not complete.
ERP and invoicing in Italy
- Fatture in Cloud
- TeamSystem
- Zucchetti
- Danea Easyfatt
- E-invoicing (SDI)
ERP in the Czech Republic
- Pohoda
- Money S3
- ABRA Flexi
- Helios
E-commerce
- Shopify
- WooCommerce
- PrestaShop
CRM
- HubSpot
- Pipedrive
- Salesforce
Communication
- WhatsApp Business
- Microsoft Teams
- Slack
Office and data
- Google Workspace
- Microsoft 365
- Excel and Google Sheets
- SQL databases
Don't see your program? If it has an API, a database or at least a file export, it can usually be connected. We check in the Discovery.
Even if your program has no API
Many ERPs installed years ago have no modern connection. That's no reason to keep them isolated.
Reading from the database
We read data straight from the ERP's database, with a dedicated user and, where possible, read-only.
Scheduled exports
If the program can export to CSV, Excel or XML, we pick the files up from a folder or an SFTP server, check them and transform them.
Documents read by AI
When data only arrives as PDFs or emails, AI reads them and turns them into structured data, to be checked where needed.
An API built on top
For systems that stay central, we build a modern connection layer the other programs can use.
Writing into an ERP's database without going through its rules is risky: we only do it when it's safe, after checking with you and, if needed, with your ERP vendor.
MCP servers: AI connected to your systems
The standard way to give an AI assistant or agent controlled access to the data and functions of your programs.
MCP (Model Context Protocol) is an open standard that lets AI use external programs: read data, look up information, carry out actions. It is supported by the main AI assistants, such as Claude and ChatGPT, and by the agents we build.
An MCP server is the link between the AI and one of your systems, for example the ERP or the CRM. We decide together what it can do: look up orders and stock, find a customer, prepare a draft quote. Everything else stays closed.
That way your team can ask the assistant "where is the Smith order?" and get the answer from the ERP, without opening it and without pasting data into a chat.
Only what you decide
Each function is opened explicitly: for example, reading orders yes, deleting them no.
Per-person permissions
The AI only sees the data the person using it has access to.
Approval for actions
Actions that change data, such as creating an order, ask for confirmation before they run.
Every request logged
A record remains of who asked what and which data was used.
The MCP server runs on EU servers or on yours. ERP data isn't copied elsewhere: the AI looks it up when needed.
Integration examples
Typical situations, with the connection that solves them.
E-commerce and ERP
Today
Online shop orders are retyped into the ERP and stock is updated by hand at the end of the day.
The connection
Shopify or WooCommerce ↔ ERP: orders one way, stock and prices the other
What changes
No retyped orders and no products sold that aren't there.
CRM and ERP
Today
Sales closes the deal in the CRM, then accounts re-enters the customer in the ERP to invoice.
The connection
HubSpot or Pipedrive ↔ ERP: customer records, accepted quotes, invoices and payments
What changes
The customer is entered once and sales can see whether they paid.
E-invoicing
Today
Invoices are prepared in one program and uploaded by hand to the portal to be sent to SDI.
The connection
ERP → data check → sent to SDI → outcome logged in the ERP
What changes
Invoices sent with no manual steps, rejections flagged right away.
WhatsApp Business and CRM
Today
Requests arrive on WhatsApp on one person's phone and stay there.
The connection
WhatsApp Business ↔ CRM: every conversation linked to the contact, new contacts created automatically
What changes
The full customer history in one place, visible to the whole team.
Czech ERP and online shop
Today
Data in Pohoda or Money S3 is correct, but the online shop has to be updated by hand with products and prices.
The connection
Pohoda or Money S3 → scheduled export → check → online shop updated
What changes
Products, prices and availability always aligned, even with an ERP without a modern API.
Alerts in Teams or Slack
Today
Whoever needs to know about a large order or a late payment finds out by chance.
The connection
ERP and CRM → rules → message in the right Teams or Slack channel
What changes
The right people are told at the right moment.
These are typical scenarios. In the Discovery we check how your programs connect and which method fits.
How we build an integration, in 5 steps
From separate programs to data that moves on its own, without stopping everyday work.
We map systems and data
We list the programs, the data they hold, who changes it and how it can be read and written: API, database or files.
We pick the source of truth
For each piece of data we set which system is authoritative: stock in the ERP, contacts in the CRM. That way there are no conflicts.
We build the connection
We use ready-made connectors where they exist and custom-build the rest, with access limited to the data needed.
We test with real data
We try the integration in a test environment with real data, including the hard cases: duplicates, empty fields, special characters.
We align and launch
We clean up existing data, switch the connection on and keep it monitored: if something doesn't get through, we know right away.
Every integration comes with its documentation: which data moves, in which direction, how often and who is alerted on errors.
Reliable even when a system goes down
Programs go into maintenance, change version, respond slowly. An integration has to cope with all of that without losing data.
Automatic retries
If a system doesn't respond, the integration retries on its own and queues the data: no order gets lost.
Error alerts
If a piece of data can't go through, for example a wrong VAT number, the integration alerts the right person with the reason.
Log of every exchange
Every piece of data that moves is logged: what, when, from where to where. You can always reconstruct what happened.
No duplicates
Checks on codes and records prevent creating the same customer or order twice.
Minimal access
Dedicated, revocable credentials with access only to the data needed. Never a colleague's password.
Servers in the European Union
Integrations run on OVHcloud servers in European data centers, or in your private cloud or on your servers.
Start with one connection. You decide whether to continue.
We begin with the integration that saves the most work. After each phase you have something that works and you decide whether to go on.
1. Discovery
1 week · freeWe map the programs, the data and the ways to connect them, and choose the connection to start with.
What you get
System map, connection method for each, cost and time estimate
2. First integration
1-5 weeksWe build, test and launch the chosen connection. Timing depends on the systems: with ready APIs it's quick, older systems need more work.
What you get
A working integration, with documentation and monitoring
3. Extension
in 2-week blocksWe connect the other systems in order of priority, one at a time.
What you get
New connections every two weeks
4. Management
ongoingMonitoring, updates when programs change version or API, small changes.
What you get
Monthly report on data exchanged, errors and interventions
In total, from Discovery to the first integration running takes 2 to 6 weeks, depending on the systems to connect.
How pricing works
Cost depends on how many systems, how they connect and how much data moves. That's why we work like this.
The Discovery is free
We check how your programs can be connected and tell you where it makes sense to start. No commitment.
Fixed price for each integration
You get the price of each connection before we start. It doesn't change along the way.
Decide one connection at a time
We start with the one that saves the most work. After each integration you decide whether to continue.
Transparent recurring costs
Hosting, any connector licences and maintenance have a clear fee, stated in advance.
When an integration isn't worth it
We'd rather tell you upfront than sell you a project you don't need.
Little data, rarely
If only a few records move each month, an occasional export costs less than an integration to maintain.
A system you're about to replace
If you're switching ERP in a few months, it's better to connect the new one. We can help migrate the data.
Data that needs fixing first
If customer records are full of duplicates and errors, connecting them copies the mess everywhere. We clean first, then connect.
Completely closed programs
If a program allows no way at all to read its data, no API, no database, no export, it can't be connected. We check in the Discovery.
Frequently asked questions
Looking for something different?
How many times does your team enter the same data?
In the free Discovery we map your programs and tell you how to connect them and where to start. If it isn't worth it, we'll tell you.