A4IT
Services / Automation & data
03 / 08
Automation & data · Processes, integration and reporting

Type it once. Let the systems do the rest.

In many small companies the same order is typed into three systems, an invoice travels by mail and by hand, and the monthly report is a spreadsheet one person understands. A4IT maps the process, removes the steps nobody needs, and connects the systems you already have so data is entered once and arrives where it is used.

Plan an automation intake → What we deliver How an engagement runs

What you get, what stays yours, where it stops

Said before the first meeting
You get
  • A drawn process, as it is today and as it should be, that your team recognises
  • Working automation between your real systems, tested with your real cases
  • Error handling and logging, so a failed transfer is noticed and can be repeated
  • Documentation and a handover session for the people who run it
Stays yours
  • Your data, your process models, your mappings and your source code
  • The licences and accounts of every tool, in your company's name
  • The business rules: readable and changeable, not buried in a black box
  • The decision which steps keep a human check
Boundaries
  • No promised recognition or extraction accuracy before we have measured on your documents
  • No automating a process that is broken: we simplify it first
  • No accounting, tax or legal advice: we build what your accountant specifies
  • No replacement of your ERP or accounting package; we connect to it

What we deliver

Six building blocks · take one or all
01 · Start here

Process analysis and simplification

Before a step is automated, we ask whether the step should exist. The cheapest automation is the task you no longer do.

Follow real cases
We trace actual orders, invoices and requests from arrival to archive, with the people who handle them, including the exceptions.
Process model in BPMN
The current flow drawn in a standard notation: who does what, in which system, and where work waits. One page per process where possible.
Simplify first
Duplicate checks, copies for the sake of copies and approvals nobody remembers the reason for are removed before anything is built.
Ranked candidates
Each step that could be automated is scored on time saved, error risk and build effort, so you choose where to start.
02 · Build

Workflow automation

Recurring work that follows rules: routed, reminded, approved and recorded without anyone keeping a list.

Rules and routing
A request goes to the right person based on amount, customer, type or region. The rules are configuration your own staff can read.
Approvals and reminders
Approval steps with deadlines, reminders and a substitute when someone is absent, by mail or in the tool your team already uses.
Scheduled jobs
Nightly exports, periodic checks and recurring documents run on a schedule, with a notification when something fails.
Exceptions for people
The standard case runs by itself. Anything unusual stops and lands in a work list with the reason attached.
03 · Build

Document capture and data extraction

Invoices, orders and forms that arrive as PDF, scan, mail or structured file, turned into data your system can book.

One intake
Mailbox, upload, scanner and supplier portal lead to the same queue, so every document is handled the same way and none is lost.
Extraction
Fields are read with text recognition and language models where they help, and delivered in a fixed schema. See AI consulting.
Validation
Totals are recalculated, suppliers and order numbers matched against your master data. Doubtful documents go to a person, with the doubt marked.
Structured e-invoices
Invoices received as structured files, for example over the Peppol network, are read directly and follow the same checks without any recognition step.
04 · Connect

System integration

ERP, webshop, CRM, accounting, warehouse and the system written twenty years ago: connected so each one keeps doing what it is good at.

REST APIs
Integrations over the interfaces your software offers, and new, documented APIs where your own applications need to be reachable.
XML and XSLT
Message formats mapped and transformed between partners and systems, with schema validation so a malformed message is rejected at the door.
File exchanges
CSV, fixed-width and XML files over SFTP or shared folders, made reliable: checked for completeness, archived and never processed twice.
Legacy systems
Older applications, including COBOL programs and their file layouts, are given a clean interface so they can stay in service while the rest moves on.
05 · Clean

ETL and data quality

Extract, transform, load: moving data between systems and making sure it is still correct when it arrives.

Pipelines
Repeatable loads from source systems into a database built for reporting, each step logged and restartable.
Quality rules
Checks on duplicates, missing fields, impossible values and broken references, with a report of what was found and where.
Master data
One agreed version of customers, products and suppliers, with a clear rule about which system is the source for which field.
Migrations
Moving to a new system: mapping, trial runs, reconciliation of totals and a rehearsed cut-over with a way back.
06 · See

Reporting and dashboards

The figures you need to run the company, refreshed automatically and defined once.

Definitions first
What exactly is revenue, an open order, an active customer? Agreed and written down, so two reports no longer contradict each other.
A few good screens
Per role a small number of dashboards that answer the questions asked every week. Detail is one click away, not on the first page.
Scheduled reports
The monthly overview arrives in the mailbox as a document or spreadsheet, produced from the same data as the dashboard.
Access by role
People see the figures they are allowed to see. Personal and financial data are restricted from the design stage.

How an engagement runs

Small steps · a decision after each
  1. Intake
    A call of thirty minutes: the process that hurts, the systems involved.
    30 minutes
  2. Analysis
    Real cases followed, the process drawn, candidates ranked.
    Process model + plan
  3. First automation
    One flow, built against test copies of your systems, with a demo.
    Working software
  4. Go-live
    Old and new run side by side until the results match, then switch.
    In production
  5. Extend
    Monitoring in place, the next flow chosen from the ranked list.
    Next step agreed

Technology

Chosen per case · nothing is mandatory
ModellingBPMN process models, data mappings, interface descriptions in OpenAPI
IntegrationREST APIs, XML & XSLT, CSV and fixed-width files, SFTP, message queues where needed
ApplicationJava and Spring Boot, .NET, JavaScript, microservices, scheduled jobs
DataPostgreSQL, SQL Server, ETL pipelines, data quality checks
DocumentsText recognition, language models for extraction, structured e-invoice formats
LegacyCOBOL programs and file layouts, older databases, Linux and Windows Server environments

Good fit, and not a fit

Saves both of us a meeting
A good fit when
  • The same data is typed into more than one system
  • Reports are assembled by hand in spreadsheets every month
  • An older system must keep running but should talk to newer ones
  • Someone in the business knows the process and can decide about it
Not a fit when
  • The process changes every week and nobody can describe it
  • The real wish is a new ERP: then choose that first
  • No system involved has an interface, an export or a database we may reach
  • Exceptions may never be reviewed by a person

Questions about automation and data

All questions →
01Do we have to replace our current software?

Usually not. The starting point is the software you have. We connect to it through its interface, its exports or its database, and only advise replacement when a system blocks every sensible route.

02Our main system is very old. Is that a problem?

It is a constraint, not a stop. Older systems, including COBOL applications, can often exchange files or be wrapped in an interface. We look at what is possible before promising anything. See software development.

03How reliable is automatic reading of invoices?

That depends on your documents, and we only know after measuring on a sample of them. Whatever the result, validation rules and a human check for doubtful cases are part of the design.

04Can you help with e-invoicing and Peppol?

We can connect the intake and sending of structured invoices to your own systems and processes. Which obligations apply to your company and from when is a question for your accountant.

05What happens when an integration fails at night?

Failures are expected and designed for. Each transfer is logged, a failed one raises a notification and can be run again without creating duplicates. Who is notified and how fast someone reacts is agreed up front.

06Do we need a data warehouse for dashboards?

Often a modest reporting database is enough for an SME. We start with the questions you want answered and build the smallest setup that answers them reliably.

Bring the task everyone dreads at month end. We will show you what can go.

Reply within three business days.Plan an automation intake →How we work