The adoption of AI Agents has established itself as one of the main areas of technological investment in 2025, and all indications suggest that this will remain the case in 2026 as well.
In this article, we put ourselves in the shoes of organizations that are just beginning their journey with AI agents or, more often, are using them in a sporadic way, supporting individual processes or activities. The goal is to understand the challenges that need to be addressed and how to design a solid and realistic path to enhancing business productivity.
AI Agents: a Key Investment also in 2026
Although estimates vary from source to source, the underlying signal is clear: according to PwC, only 4% of companies have not launched initiatives related to AI agents and do not plan to do so in the near future. All the others fall along a maturity spectrum ranging from full enterprise-wide adoption (17%), to broad but not yet pervasive use (35%), to more limited and experimental applications (27%).
Return on investment is still a variable in the process of consolidation, but one certainty remains: companies around the world see agentic AI as a key factor for improving operational efficiency, optimizing costs, and increasing productivity, especially in an increasingly competitive and complex market environment.
AI Agents and Productivity: the Real Challenges to Address
For many companies, adopting AI Agents aims to boost productivity, as a lever that cuts across the entire organization and, at the same time, as a concrete area in which to start investing in AI. In practical terms, most organizations are asking how agents can help people work better, faster, and with less friction.
In 2025, the narrative around AI Agents was highly simplified: you activate the agent, connect it to data sources, define the rules, and the system is ready to use. The promise is immediate, almost plug-and-play productivity, and this has led many organizations to experiment independently, often within individual business units. In practice, however, when moving from initial tests to daily use, non-trivial complexities quickly emerge.
AI Agents live within processes
The first challenge is organizational, because AI Agents operate within processes, not at the margins. This applies not only at the company level, but also within individual functions. Introducing an agent is not equivalent to traditional automation, where repetitive, rule-based activities are identified and replaced with rigid automations; AI Agents operate on smart workflows, capable of adapting to context and taking different paths based on what has occurred in previous steps.
As a result, it is not enough to insert an agent into an existing process and expect automatic benefits. Often, the process itself needs to be rethought, clarifying objectives, decision points, responsibilities, and performance metrics. Without this, there is a double risk:
- Local but limited improvements;
- An increasing number of disconnected agents, optimized for specific needs but unable to impact overall productivity.
The skills challenge
Beyond the plug-and-play narrative, the second challenge is technical and architectural. Unlike general-purpose chatbots, AI Agents must access corporate data, interact with applications, and comply with security, identity management, and governance rules. In some contexts, this means dealing with legacy systems, non-uniform APIs, incomplete data, or inconsistent data quality, factors that make adoption far from a one-click operation and require specific expertise.
The risk of excessive expectations
AI Agents promise autonomy, but that does not mean the absence of control. Defining boundaries, responsibilities, escalation mechanisms, and effectiveness metrics is essential to turn AI into an ally of productivity rather than a source of risk or inefficiency. It is often at this stage that the need for skilled internal professionals or the support of a dedicated partner becomes evident: someone capable of translating the agents’ potential into measurable results.
How to approach the journey: from experiments to value at enterprise scale
The most solid benefits of agentic AI are reserved for organizations that decide to tackle complexity from the outset, reasoning at an enterprise-wide level. This is clearly a more demanding approach than a spot-based one (according to PwC, only 17% of companies have already adopted AI Agents across the entire organization), but it is also the one that enables the creation of a real competitive advantage.
This opens up two distinct areas of evaluation. On one side, strategic considerations related to objectives, priorities, operating model, and organizational impact. On the other hand, technical considerations.
Strategic evaluation: thinking like a CEO
From an executive perspective, productivity should be considered a strategic lever, not the sum of local micro-efficiencies. This means starting from a few key objectives—such as reducing time-to-market, making better use of skills, or lowering cost per unit—and then translating them into AI-powered operational choices.
To achieve tangible results, it is necessary from the outset to understand which divisions and processes are involved, and then to hypothesize the role and impact of agentic AI not only on individual activities, but on entire processes, with an end-to-end vision.
For example, an AI Agent can help reduce time-to-market by dynamically coordinating activities that proceed sequentially across multiple functions, flagging bottlenecks, anticipating critical dependencies, and suggesting operational priorities. Or, in the context of operations, it can support greater scalability by adapting workflows based on volumes, redistributing workloads, and reducing repetitive manual interventions.
In all these cases, and many others, the value does not lie in automating a single task, but in the AI Agent’s ability to influence the process as a whole.
AI Agents for Productivity: Integrated or Custom-made?
At this point, implementation must be addressed. Typically, there are two possible paths.
Integrated solutions within productivity tools
The first option is to use ready-made (or semi-ready) solutions already integrated into the tools in use. Collaboration and productivity platforms are particularly well-suited, as they centralize documents, conversations, communication tools, and operational information; essentially, much of the corporate knowledge on which agents can operate.
When the platform provides predefined agents or tools to create them, time-to-value is high, and benefits to individual and team productivity can emerge quickly. However, adoption must be consistent with the strategy defined upstream.
Custom Development of AI Agents
The second path is the custom development of agentic solutions designed to operate across complex processes and interact with multiple systems and business functions. This is a complex approach, but it is also the most suitable when the goal is to directly impact business competitiveness.
Solution development starts from the architecture, which consists of several elements:
- Foundation, such as the LLM language model;
- Corporate knowledge (and the ability to access it securely);
- Operational capabilities;
- Workflow development and orchestration;
- Interaction with other enterprise systems;
- User experience.
From an architectural standpoint, the agent takes the form of an intelligent workflow that gathers information from existing systems, processes it through an AI model (general-purpose or specialized), and returns outputs in the form of information or actions, depending on the defined level of autonomy.
This is where the narrative of simplicity meets the limits of the enterprise environment. Despite the availability of low-code and no-code tools, integration with legacy systems, security and compliance requirements, and the need to ensure reliability and control make specialized skills indispensable. Moreover, one must consider the rapid evolution of technologies and the fact that AI Agents do not operate in isolation, but can interact with one another within ecosystems that grow day by day.
Our Role: from Innovation to Concrete Adoption
At Kirey, we have been working for years on advanced technologies and complex architectures, and it is precisely this experience that allows us today to address the evolution toward AI Agents with a solid and informed approach, rather than a reactive one.
The value we bring to our clients goes beyond agent development alone. Instead, we take responsibility for the entire adoption journey: from strategic support in selecting the most suitable architecture, to governance and integration with existing systems, all the way to the organizational support required when the introduction of AI profoundly changes how people work.
Contact us to understand how to design an adoption path aligned with your business objectives.
