EN / IT

AI · Agents · Automation · Governance

Artificial intelligence put to work in your processes

I take AI from proof of concept to a stable, measurable operational capability: agents, web apps and automations built into finance, audit, compliance and ESG processes.

Services

  1. 01

    AI agents and automation

    Agents and multi-step workflows for document analysis, structured data extraction, quality checks and report generation, integrated with your existing tools (document archives, CRM, collaboration platforms).

  2. 02

    Enterprise web apps

    Complete data-driven platforms (frontend, backend, APIs, database) with modular, containerised architectures: access control, roles, audit trail, versioning.

  3. 03

    AI for finance, audit and sustainability

    Applications for people who work with numbers and rules: from ESG data collection to audit working papers, from reconciliations to reporting.

  4. 04

    Governance and security by design

    Environment segregation, credential management, access policies, monitoring, incident management and compliance with privacy and data-governance requirements. Security is decided before any code is written.

  5. 05

    AI training

    Courses and workshops for professionals and companies with Wolters Kluwer, Namirial and Eutekne: using generative AI productively, safely and responsibly.

How I build solutions

Every solution I build has rules, controls, logs and points where a person approves the result. You can always reconstruct what the system did and on which data: it is how I work as an auditor, applied to software.

Case studies

Projects I designed and built with AI for real needs of accounting firms and companies. Client names and data never appear.

Practice platform for tax records, collections and state aid

Problem
Data and deadlines scattered across the tax records portal, the collection agency's reserved area, the State Aid Register and spreadsheets, checked by hand client by client.
Solution
One multi-user platform that, using the firm's Entratel credentials, gathers tax payments, tax credits, letters of intent, tax notices, collection-agency statements and instalment plans, state aid and land-registry searches; a single deadline calendar and bulk operations. OCR and AI to read documents, with a dedicated service that masks personal data before it reaches the models.
Technology
Python, Flask, PostgreSQL, Redis, Docker; multi-provider AI; per-module roles, encrypted credentials; 300+ automated tests.

State aid in natural language

Problem
Checking de minimis headroom and preparing the state-aid disclosure for the notes to the accounts (Law 124/2017) means long searches in the National State Aid Register.
Solution
An MCP connector to query the Register directly from Claude or ChatGPT ("how much de minimis headroom does this company have?"), generate the disclosure and export to Excel. Over 25 million aid records, read-only.
Technology
Python MCP server on Docker, dual database with automatic failover, nightly search index, automated tests.

Central Credit Register analysis

Problem
Reading Bank of Italy credit-register reports and spotting anomalies takes time and experience.
Solution
A web app that imports the PDFs, extracts the data, shows trends and anomalies in a dashboard and produces a report with AI-generated comments reviewed by the professional.
Technology
Next.js, Python (FastAPI) PDF parsing, hash-chained audit log, access protection, Docker.

Lease management

Problem
Leases, rent, ISTAT indexation, flat-rate tax regime, registration tax, renewals and terminations handled across different tools.
Solution
A management tool covering the whole lease lifecycle, with an automatic deadline calendar (registration within 20 days, renewals, terminations), payment forms and arrears control, for multiple companies and users.
Technology
FastAPI, React and TypeScript, roles and audit log.

In development A management tool for flat-rate taxpayers, from downloading invoices from the Revenue Agency to calculating taxes, contributions and the annual return.

Frequently asked questions

Does our data leave the company?

It depends on the architecture we choose together: cloud models with contractual safeguards, European providers, or models running on-premise. We decide before writing any code.

Do we need a big project to start?

No. The best way is to start from a concrete, repetitive process, measure the result, then extend.

Who maintains the solution?

I deliver it documented and versioned; I can maintain and evolve it, or train your team to do so.

A process to automate

Tell me how it works today: in 30 minutes we can see whether AI can help.

Book a callWrite to me →