Projects
From Paper to Power: Fuel Expansion with AI
Vega Carburanti
The ultimate goal was never just to build a document analysis tool. It was to build a growth engine. That’s exactly what we’ve got. We achieved a system where users can instantly verify the source of data, turning skepticism into the confidence of having a new superpower.
The collaboration with Vega Carburanti transformed a complex acquisition process from a manual barrier into an intelligent, scalable system. This evolution redefined how assets are evaluated, turning static documentation into a queryable knowledge base and providing legal and technical teams with the tools to execute high-speed expansion with precision.
Operational Results at a Glance - KPI Objective
Enable rapid ingestion of acquired networks without proportional headcount growth.
- 100% Verifiability of data sources to ensure decision integrity.
A Complex Mandate for Market Leadership - Scenario
Vega Carburanti operates a high-performance network with a turnover exceeding €700 million and over 120 stations across Northern Italy. The operational model is distinct: their stations dispense six times the national average value. The vision is clear: rapid, strategic expansion. However, this growth strategy relies on the continuous acquisition of new stations and networks.
We are operators of service stations. Our speciality is efficiency. But imagine my general manager calls tomorrow and says 'we are acquiring 200 stations'—doubling our size. Personnel, IT tools, the whole company must be adequate.
This ambition faced a reality common to high-growth sectors: a massive influx of legacy information essential for operations, yet difficult to access.
A Race Against Complexity - The Challenge
The challenge was not finding assets, but integrating them. Each acquisition involved inheriting a mountain of unstructured documentation: leases, staff agreements, land purchases, construction permits, and technical blueprints.
Acquiring all the contracts and being able to extract accurate data quickly is fundamental to fast growth. Currently, it requires manually sifting through thousands of pages.
The existing process presented evident criticalities:
- Manual Data Entry: Teams manually analyzed thousands of pages to extract critical data points (e.g., tank measurements, permit status).
- Unstructured Fragmentation: Information was locked in PDFs, Word files, Excel sheets, and scanned paper, often distributed across different formats.
- Strategic Risk: The inability to quickly answer complex questions—such as 'Which new stations have permits for EV chargers?' —threatened the speed of the growth strategy.
- Scalability Bottlenecks: The manual evaluation process was slow, prone to error, and could not scale to meet the target of doubling the network size.
The Objective - Expected Results
The goal was to overcome this manual bottleneck by introducing a system capable of:
- Drastically reducing the time required to analyze legal and technical documents.
- Structuring data from disparate formats into a unified, queryable database.
- Providing absolute trust through verifiable data sources, eliminating the risk of AI 'hallucination.'
- Building a 'growth engine' that integrates with the company's expansion goals without requiring a massive increase in staff.
Intelligent Asset Platform: Intelligence Activating Knowledge - The Solution
The project was built through co-creation. Working closely with Vega’s legal, technical, and business development teams, the solution was designed to address specific domain needs rather than applying a generic tool.
We approached this not as a development project, but as a co-creation initiative. AI isn't an add-on; it's the core of the system.
Efficiency, Control, and Strategic Transformation - Results
The project produced strategic results that altered the operational capacity of the organization: From Passive Archive to Active System The document archive has evolved from a storage problem into a source of untapped intelligence. Information that was previously locked in static files is now immediately accessible. Teams can interrogate the database in real-time, extracting precise data points without manual sifting. Simplification and Trust via Verification The 'fact-checking' feature turned skepticism into adoption. By allowing users to click and instantly verify the source of data, the system bridged the gap between automation and human responsibility.
This is what users appreciate most: with one click, it opens the document and takes them to the paragraph from which the data point was taken. This served tremendously for users to gain awareness and trust in the tool.
Governance and Scalability The platform acts as a strategic catalyst. By automating the ingestion and analysis phase, Vega transformed a manual barrier into an automated pipeline. The system now supports the ambition of doubling the network size, ensuring that infrastructure evolves at the same speed as the business strategy
Tech Stack & Implementation
The solution combines advanced technologies for optimal performance:
Lessons Learned: A Pragmatic AI Model
The journey yielded a clear operational model for deploying transformational AI. Co-designing to Solve Real Bottlenecks The ultimate goal was never technology for its own sake, but solving a high-stakes business problem. By targeting the acquisition bottleneck, the project built powerful momentum.
The ultimate goal was to build a growth engine. That’s exactly what we’ve got.
AI is a Dialogue, Not a Command Value emerges when users learn to interact with the system. Proactive training on how to ask clear questions was essential to eliminate ambiguity.
“When we started using the actual chat, that's when the learning began. The user needs to know how to ask for things to get the best results.
Trust is Earned Through Transparency The fear of errors is a major barrier. The ability for a user to instantly verify the source of any piece of data was the single most important factor in empowering teams to rely on the tool for mission-critical decisions.
Key Indications for Your Organization - Key Takeaways
The success achieved with Vega Carburanti offers a replicable blueprint:
- Unlock Data Value: View document archives not as storage, but as a source of queryable intelligence that can drive decision-making.
- Target High-Stakes Goals: Apply AI to solve critical, high-pressure bottlenecks. Fixing the acquisition pipeline created immediate, measurable value.
- Build Trust as a Core Feature: Lasting solutions are built on confidence. Verifiability and 'fact-checking' capabilities are non-negotiable for user adoption in professional settings.
- Invest in Human-AI Dialogue: Technology is only half the equation. Training users on how to structure queries is essential to retrieving reliable results.