Innovation
Jul 2025

Agentic AI: a new driver of business process transformation

Time to read: 5 min

written by

Davide Gamba

Digital Business Innovation Partner

Davide Gamba currently serves as the Digital Business Innovation Partner at the Department of Digital Transformation, Information Systems, Security & Compliance for Gewiss SpA. His work involves applied research and innovation management in the digital business field. He holds a Ph.D. in Technology, Innovation, and Management from the University of Bergamo, focusing on servitization strategies, where he also teaches in digital innovation, strategic management, and information technology courses.

The evolution of Artificial Intelligence (AI) is providing manufacturing companies with new tools to transform their operational processes. Among these, Agentic AI emerges as a strategic turning point for companies like GEWISS that aim to adopt business models increasingly aligned with the principles of Industry 5.0.

 

 

What is Agentic AI: from response to autonomous action

The concept of Agentic AI refers to artificial intelligence systems equipped with agency, that is, the ability to act autonomously toward a goal by executing sequences of actions in dynamic and complex environments.

Unlike traditional AI models, which are limited to reactive responses to single prompts, autonomous AI agents are defined by three fundamental characteristics:

  • Autonomy: the ability to operate independently without constant user input.

  • Contextual memory: the retention of interaction history, states and relevant information to learn and adapt over time.

  • Exploration: the ability to handle uncertain or unknown situations by formulating autonomous response strategies.


This approach signals a radical shift in logic, moving from simple prompting (where the user manually queries a Large Language Model each time) to intelligent delegation, where the goal is assigned to the agent, which then pursues it in an iterative and adaptive manner.

Among the key frameworks in this domain is ReAct (reasoning and acting), which allows the agent to alternate between reasoning and action to solve a given task.


Applications of Agentic AI in industrial processes

Within the industrial context, operational AI agents are already being applied in various areas to support the development of more adaptive and personalized processes. Below are the main use cases.

Predictive maintenance

AI agents capable of processing data from IoT sensors, detecting abnormal patterns and suggesting preventive maintenance interventions.

Advanced customer support

Intelligent assistants that consult technical documentation and interaction history to provide personalized and contextual responses to customers.

Logistics optimization

AI agents that analyze complex scenarios to propose dynamic delivery schedules or more efficient procurement strategies.

A B2B use case: Agentic AI in the sales process

A concrete example helps clarify the operational potential of Agentic AI in industrial environments.
Let’s imagine assigning the following goal to an intelligent agent: optimize commercial offers for the “Top 10” Italian clients for the next quarter.

  • Autonomy | The AI agent independently manages multiple stages of the process. It accesses company databases to extract up-to-date customer data, cross-references this information with current price lists and ongoing promotions and then prepares preliminary sales proposals. Once ready, these proposals are automatically submitted to the relevant sales contact for validation or final approval, eliminating the need for manual intervention at every step.

  • Contextual memory | When drafting proposals, the agent reviews each customer’s negotiation history: purchase frequency, average volumes, previous margins, past issues and responses to earlier promotional campaigns. This enables it to suggest solutions that align with the customer’s expectations and the trajectory of the commercial relationship.

  • Exploration | In addition to replicating known logic, the agent can experiment with new offer configurations. For instance, it may propose targeted discounts when a drop in orders is detected or test different bundling and pricing strategies to determine which approaches are most effective for similar customer segments. All of this is done in a controlled and traceable way to optimize performance without jeopardizing the commercial relationship.
     

Agentic AI and the democratization of technology

The adoption of enterprise AI agents is no longer the exclusive domain of large digital players. Recently, Agentic AI has gained momentum thanks to the democratization of development tools, made possible by no-code and low-code platforms such as Microsoft Copilot Studio and Salesforce Einstein 1 Studio.


These design environments provide intuitive visual interfaces that enable even non-technical users to create and manage custom AI agents. Through these tools, it is possible to:

  • Design the agent’s behavior based on clear objectives.
  • Define the decision-making flows for different operational scenarios.
  • Connect the agent to business information sources such as databases, documentation and CRM systems.
  • Configure rules, constraints and intervention logic tailored to the context.


An effective and reliable AI agent is built upon a carefully orchestrated combination of Large Language Models, Retrieval Augmented Generation (RAG) techniques, contextual memory to ensure consistency and connectors capable of integrating with existing enterprise systems (ERP, business management software, CRM, etc.). Its evolution depends largely on three key factors: the quality of the engineered prompts, the clarity of the initial operational rules and the ability to learn iteratively from outcomes in the field.

Governance and responsibility: critical aspects of adoption

As with any significant technological transformation, the widespread adoption of autonomous AI agents raises crucial governance-related questions. Businesses must clearly define the operational boundaries of these agents, establishing who is accountable for their actions, how to intervene in case of errors and which oversight mechanisms are in place throughout the decision-making process.

It is, therefore, essential to ensure that decisions made autonomously by an agent are traceable, explainable and subject to oversight, especially when such systems are deeply integrated into core business processes.

The conscious governance of Agentic AI must be considered a prerequisite for minimizing ethical, legal and reputational risks while fostering a balanced relationship between automation and human intervention.

 

Agentic AI: an opportunity for transformative enterprises

Today, Agentic AI serves as a strategic lever for companies seeking to enhance their organizational intelligence and drive meaningful innovation within their workflows. Far from being mere execution tools, intelligent agents can become co-actors in the corporate system, capable of reducing repetitive workloads, enhancing decision accuracy and unlocking new levels of efficiency, adaptability and personalization.

Choosing to invest in the design and orchestration of intelligent agents, even through accessible platforms like the aforementioned Microsoft Copilot Studio, means building a smarter and more resilient business ecosystem that can evolve continuously in harmony with people and the transformations shaping today’s industry. To be truly effective, this transformation must combine technological autonomy with human awareness, paving the way for a future in which innovation empowers rather than replaces.

Share

Write to us

Do you need information on Gewiss products or services?