This article analyzes how Robotic Process Automation (RPA), integrated with Artificial Intelligence, is becoming a strategic lever for the operations of Italian manufacturing companies, which are still lagging behind in terms of digitalization. RPA enables the automation of high-volume, repetitive processes—from order management to reporting—by connecting fragmented systems and freeing up people for higher-value activities. The benefits are rapid and measurable, but implementation requires preliminary Lean mapping, robust governance, and an adequate Change Management program to avoid key risks: cultural resistance, process fragility, and compliance vulnerabilities. The challenge for Italian manufacturing is not simply adopting AI, but knowing how to put it to work in a scalable and secure manner.
Artificial Intelligence and Generative AI are redefining Robotic Process Automation, enabling companies to automate repetitive tasks and optimize operations through adaptive intelligence. RPA tools have been on the market for years, but with the consolidation of AI technologies, they are truly making their way into companies, including those within the manufacturing sector.
Italian manufacturing companies boast solid operational maturity in Operations and a well-structured Supply Chain, but lag behind in overall digitalization. According to our Benchmarking Study 2025 Edition | What’s Next in Operations?, conducted on a broad sample of companies across various sectors, many organizations still operate with manual and reactive processes rather than proactive ones.
Only a minority has achieved Smart Factory status, while most are defining a roadmap for digital transformation. AI and GenAI are recognized as priorities, with projects launched primarily for strategic analysis, increased productivity, and process automation. This landscape highlights the opportunity for Made in Italy to accelerate adoption to compete in terms of innovation and value.
Process Automation with AI goes beyond traditional automation: it leverages intelligent algorithms to manage routine tasks, predict anomalies, and optimize workflows in real time. Generative AI adds an advanced layer capable of generating complex content, code, or predictions from structured and unstructured data, drastically reducing the need for human intervention.
This approach frees up resources for high-value tasks, such as innovation and strategic problem-solving. In an industrial context, it transforms ad-hoc operations into predictive ecosystems, integrating data from across the value chain to support data-driven decisions.
Applications of AI-powered Process Automation range from transactional processes to production, offering scalable solutions for Italian manufacturing.
Transactional automation handles invoicing, order processing, and data entry, eliminating manual errors and accelerating cycles. In customer service, AI-powered chatbots and intelligent routing systems filter emails, categorize tickets, and suggest personalized responses, improving the customer experience.
In the supply chain, dynamic inventory optimization and demand forecasting reduce stockouts and overstocking. Quality control detects production defects with high precision, while automated compliance checks verify regulatory compliance at every stage. In IT systems, network monitoring identifies and resolves issues autonomously.
These examples illustrate how AI adapts to the dominant B2B contexts in the Italian sample, enhancing critical areas such as quality and logistics.
The adoption of RPA (Robotic Process Automation), especially when combined with AI, enables an immediate reduction in time spent on repetitive, high-volume tasks (data collection, updates across multiple systems, checks, reporting, ticket management, master data), with a direct impact on operating costs, quality, and service levels.
In Operations, RPA becomes the "executive engine" that connects often-fragmented tools (ERP, MES, CRM, portals, email, Excel), automating end-to-end workflows such as opening and updating work orders, finalizing and closing administrative tasks, collecting and consolidating production data, document control, managing internal requests, updating master data, and extracting and distributing KPIs. The result is a more scalable and robust operational model: it increases the proportion of touchless processes, reduces bottlenecks and “chasing” activities, and improves the timeliness of information flows.
Integration with Lean paradigms amplifies the benefits: RPA eliminates waste associated with unnecessary information flows, waiting times, rework, and redundant checks, making processes more stable and aligned with “standard work.” When paired with AI, the organization shifts from reactive to proactive management: AI detects anomalies and weak signals (performance deviations, downtime risks, quality deviations), while RPA automatically triggers standard corrective actions (escalations, task creation, system updates, targeted communications), reducing response times and inefficiencies.
ROI tends to materialize quickly in contexts with high repetitiveness and critical mass (industrial operations, service operations, centers of expertise, shared services), thanks to modular approaches: we start with high-volume, low-complexity “quick win” processes and scale up to end-to-end streams. In addition to efficiency, automation also contributes to operational sustainability: fewer reworks and data errors, fewer urgent issues caused by misalignments, greater process stability, and higher data quality for measuring and managing environmental KPIs (energy, waste, rework).
Measuring the impact of RPA is not just an academic exercise: it is the prerequisite for scaling with confidence. In highly repetitive industrial contexts, improvements manifest across multiple dimensions simultaneously—speed, quality, reliability—and become visible as early as the first weeks of operation. The key metrics to monitor before and after implementation include:
Despite the advantages, the introduction of RPA (and even more so when enabled by AI) entails organizational and technical risks that must be managed with a structured approach. The first is human and cultural risk: the perception of "replacement" can generate resistance, decreased engagement, and defensive behaviors (shadow processes, rule-bending). To avoid this, transparent communication is needed, along with involving people in use cases and reskilling/upskilling programs for “enhanced” roles (exception handling, data quality control, process ownership, supervision of bots and models).
On an operational level, RPA can introduce a risk of process fragility if it automates non-standardized or “exception-heavy” activities: a bot replicates what already exists, so if the process is unstable, there is a risk of automating waste instead of eliminating it. This is a common risk when starting without end-to-end mapping (Lean/process mining) and without clear rules for managing exceptions and responsibilities (who does what when the bot fails).
In industrial and regulatory contexts, alignment with compliance, security, and auditability is crucial. Bots access systems and data: if digital identities, segregation of duties (SoD), action traceability, and access control are not properly defined, legal and operational risks arise (errors involving sensitive data, procedural violations, audit difficulties). Cybersecurity is also an issue: hardcoded credentials, shared accounts, or unmonitored logs can become risk vectors.
From an IT perspective, integration with legacy systems or unstable interfaces can generate technical debt: many "UI-based" automations are sensitive to changes in screens, fields, or permissions, leading to increased maintenance and bot downtime. Without governance, a "jungle of scripts" can emerge that is difficult to maintain, impacting operational continuity and resulting in unexpected costs.
When AI is added (e.g., document classification, data extraction, intelligent routing), additional risks emerge: data quality, model drift, performance degradation over time, and the need for continuous monitoring. In these cases, it becomes essential to define confidence thresholds, sample checks, and "human-in-the-loop" processes for ambiguous cases to prevent systematic errors that spread rapidly.
Finally, there are economic and delivery risks: initial costs (licenses, setup, change management), excessive expectations ("RPA solves everything"), misaligned priorities, and a prolonged time-to-value if starting with overly complex processes. The most common barriers remain: poor data quality, integration complexity, lack of standardization, and absence of ownership and governance.
AI-based Process Automation implementation strategies rely on structured and phased approaches. Mapping high-volume, low-value-added processes—the so-called "low-hanging fruits" such as reporting or invoice management—serves as the starting point. A rigorous assessment of data quality and the selection of reliable vendors precede the launch of pilot projects on one or two specific use cases.
The scaling phase occurs through integration with ERP and MES systems, with constant monitoring of KPIs such as cycle times, error rates, and overall ROI. Change Management plays a central role: ongoing training programs on the potential and limitations of AI help reduce internal resistance, while the involvement of company leadership fosters the development of a data-driven culture.
Vendor management is based on clear Service Level Agreements (SLAs), strategies to avoid lock-in, and customized solutions. Ethics and governance—through dedicated AI committees, periodic bias audits, and transparency protocols—ensure responsible use, which is particularly crucial for black-box generative models.
The results of the 2025 edition of the Benchmarking Study | What’s Next in Operations? reveal that the potential remains largely untapped, with an implementation level of around 10% across a sample of over 100 cross-industry companies. The areas most heavily involved are manufacturing, quality, and customer experience. Based on our experience, cybersecurity—which is often reactive—must be integrated by design into IoT and AI to protect connected operations.
Generative AI is shifting from a "support tool" to an active component of processes: it not only produces content but enables agents capable of orchestrating tasks across multiple systems, managing exceptions, learning from feedback, and restoring workflows when they are interrupted (self-healing). For Made in Italy, this represents a paradigm shift: less time spent chasing information, aligning data, and coordinating micro-tasks; more time dedicated to decision-making, quality, service, and innovation. Value does not stem from AI "in and of itself," but from its integration into an operational architecture composed of standard processes, reliable data, clear rules, and defined responsibilities.
This transition is not automatic: it requires targeted investment in training, Knowledge Management, and new roles (Process Owner, Automation Lead, Bot/Agent Supervisor) to avoid the "eternal pilot" effect and transform technology into industrial capability. In practice, the following are needed: a Lean foundation to reduce variability and waste before automating; governance that ensures quality, security, and auditability; and a corporate knowledge base (standards, procedures, lessons learned) that is "readable" by machines as well, so that agents operate with consistency and continuity.
Bonfiglioli Consulting, through the Knowledge Office and the Lean Factory School®, supports companies in combining Operations Excellence and AI: from designing processes and the operating model, to selecting high-impact use cases, to building internal capabilities and mechanisms for continuous improvement.
Process Automation is no longer just an efficiency upgrade: it is the lever to make operations more resilient, faster, and more manageable. In 2026 and beyond, the competitive edge will not be "having AI," but knowing how to implement it in a scalable, measurable, and secure way.
RPA is a software technology that automates repetitive, rule-based tasks—such as data entry, order updates, or report generation—by replicating the actions a human operator would normally perform on corporate information systems (ERP, MES, CRM, email). Unlike traditional AI, RPA executes predefined instructions without "learning." When integrated with Artificial Intelligence, the system becomes adaptive: it not only executes but also interprets unstructured data, detects anomalies, and handles exceptions autonomously, enabling hyperautomation.
In the manufacturing sector, the processes best suited for RPA are those with high volume and low variability: opening and updating work orders, finalizing and closing administrative tasks, collecting and consolidating production data, document control, managing internal requests, updating master data, and extracting KPIs. The supply chain also benefits significantly: demand forecasting, inventory optimization, and supplier monitoring are areas where bots deliver rapid and measurable results. The guiding principle is simple: the more repetitive and standardized a process is, the more suitable it is for automation.
Return on investment tends to materialize quickly in highly repetitive and high-volume contexts, typical of industrial operations and centers of expertise. The most effective approach is modular: start with high-volume, low-complexity quick-win processes—such as reporting or invoice management—and progressively scale up to end-to-end workflows. In these scenarios, the lead time for administrative tasks can be reduced from hours to minutes, with a direct impact on operating costs, data quality, and service levels as early as the first few weeks of operation.
There are three main types of risks. The first is organizational and cultural: the perception of “replacement” can generate resistance and defensive behaviors; it is essential to accompany the change with transparent communication and reskilling programs. The second is technical: automating unstable or exception-heavy processes means “automating waste”—which is why preliminary Lean mapping is essential. The third is compliance and security: bots access sensitive systems and data, so digital identities, role segregation, action traceability, and cybersecurity must be defined by design, not added as an afterthought.
The starting point is mapping the highest-volume, lowest-value-added processes—the so-called “low-hanging fruits”—while first verifying the quality of the available data. One or two pilot use cases are selected, reliable vendors with clear SLAs are chosen, and the project is launched in a controlled manner. The scaling phase occurs through integration with existing systems (ERP, MES) and constant monitoring of KPIs such as cycle times, error rates, and touchless rates. According to Bonfiglioli Consulting’s 2025 Benchmarking Study, only 10% of the companies in the sample have already implemented structured solutions: the untapped potential is enormous, especially in the areas of manufacturing, quality, and customer experience.
ai translated