Impact map in the world of AI: a new framework for service evaluation

In 2022, Ammagamma—an innovative Italian company specialized in AI-based solutions —approached us with the idea of defining a clear and structured method to measure the impact of AI. The ambition was to look beyond efficiency metrics, exploring how technology could reshape daily practices, culture, and stakeholder relations, while anticipating improvements as well as risks, unintended effects and long-term changes.

One of Ammagamma most recent and successfull implementations, an Inventory Optimizer for a company in the food logistics, became the first case study. That company manages thousands of meals every day, coordinating suppliers, warehouses and canteens in a complex supply chain where even small forecasting errors can lead to significant food waste and financial losses. To address this, the Inventory Optimizer uses historic data to make predictions for supplies restock, supporting the warehouse and procurement offices in placing orders to the exterrnal providers. The first goal of the AI implementation was obviously to reduce any waste, with a clear benefit both in terms of environmental footprint and economic savings. Yet the adoption raised a crucial question: what other impacts—environmental, social, economic, and governance-related—would such an innovation create within the organisation, across its supply chain, and even in the broader ecosystem of institutions and communities it interacts with?

Before approaching the impact measurement, we built a framework that served as the backbone of the entire project. The first step was a full reconstruction of the supply chain system, identifying not only the technical functions—such as IT, Logistics, CSR, Communication, Purchase, and HR—but also the wider network of external partners including suppliers, canteen operators, and institutional stakeholders. In this phase, it was essential to work closely with the food warehouse and logistics company to understand the entire ecosystem, as well as Ammagamma, the technology provider, to reconstruct the development lifecycle and journey of AI generation. Next, we started to identify and prioritise impact indicators, thanks to the collaboration with Tiresia (Politecnico di Milano) and their deep expertise in service evaluation. This meant agreeing on what exactly should be measured across the four chosen dimensions—economic, environmental, social, and governance—at every stakeholder level, and defining clear criteria for evaluation. Indicators ranged from tangible metrics, like reductions in waste volumes or improvements in delivery accuracy, to less immediately visible aspects, such as employee perception of their work, collaboration across departments, and the degree of transparency in governance processes. The impact measurement methodology combined qualitative and quantitative approaches. On the qualitative side, it relied on in‑depth interviews with staff at all levels—from warehouse operators to senior management—along with on‑site observations and digital ethnography to assess internal and external communication. On the quantitative side, it analysed company data, algorithm outputs and management system metrics in order to track changes in efficiency, stock management, waste reduction and costs.

The results were summarised in a visual map displaying all the types of impact occurred, distinguishing among positive and negative outcomes, unkown or missed opportunities as well as risks to consider moving forward. The findings showed that the AI solution’s benefits went well beyond environmental goals. The most significant effects were governance and social impacts: employees saw the tool as a supportive “assistant” that streamlined workflows and freed time for quality control. Fears of job loss gave way to requests for broader automation. At the organisational level, structured real‑time data management fostered a culture of data literacy, helping staff make better decisions. Beyond internal gains, the project also influenced regional R&D and debates on AI, showing how its impact can spread across multiple layers of the ecosystem. In addition, the results of the mapping could be used to demonstrate the value and return of investment of an AI solution in relation to ESG metrics.

Mapping the impact of AI implementation is absolutely critical and companies need to consider using monitoring and evaluation frameworks that could accompany the design and adoption of any AI-based solution. Based on our experience in this project, we reccomend all companies we work with on similar implementations to start brainstorming upfront the expected (or unexpected) impact that a new AI solution could bring, and define objectives and metrics that could first lead to a more conscious design and execution of the project, and later be monitored over time.