The use of AI within organizations is at a tipping point. Whereas previously it was mainly limited to experimentation with chatbots and automation scripts, it is now becoming more fundamental. AI agents independently perform complex tasks, make decisions, and continue to develop based on context and interaction. Where is agentic AI headed? A roundtable discussion with experts from Cloudera, Pega, Salesforce, and ServiceNow, hosted by Techzine, gives the latest insights.
AI agents are ushering in a new era of automation. It is no longer a tool limited to support work. It also makes decisions and actively participates. It quickly becomes clear that companies cannot afford to wait. The risk of falling behind will soon become too great. The experiences of pioneers show that it is not about perfect implementation, but about getting started, learning, and scaling up.
The first step is to clarify where AI can add value. Organizations that approach AI as a new way of working are progressing the most. Don’t see it as a project for the IT department, but as something that benefits the business. No matter how advanced the technology is, ultimately it’s about solving real business problems.
Read our first article based on the roundtable discussion. In it, the experts discuss the basics you need to for successful agentic AI systems.
When should you choose an agent, and when should you choose something else?
An important consideration for organizations is the distinction between processes that benefit from an AI agent and processes that are better served by a more traditional approach. This distinction often depends on factors such as variability, scale, compliance, and the degree of human interaction. Ultimately, you can do a lot with AI agents, but certain types of automation are better suited to other types of technology.
So, when exactly should you opt for traditional workflows and when for AI agents? According to Nick Botter, Head of Solution Consulting at ServiceNow, that distinction is outdated. In his view, it’s not about whether you use an agent or a workflow, but about the orchestration between them. “You always need an ‘agent orchestrator’ to determine the next step. The agent has skills, tools, and workflows, and chooses the best route for the task at hand.”
In Botter’s case, you have, as it were, an intelligent layer that decides when a task should go to a workflow and when an agent should take over. This orchestration layer becomes the hub of automation, in which AI and traditional IT solutions function side by side. This is much more of a dynamic approach, where people work adaptively.
Agents can therefore be useful in any situation. Making suggestions, performing analyses, and identifying deviations as part of a larger automation process can all be done if required. Context is one of the most critical factors for success, states Peter van der Putten, Director of AI Lab at Pega. Data and feedback are also important. Agentic AI is useful for complex tasks. “Ultimately, you have to look carefully at where you have predictability and where you don’t,” Van der Putten explains. “The technology you use will then follow naturally.”
Assess situations before making a decision
Organizations that set things up correctly will soon be able to use AI agents to answer customer questions in multiple domains. As long as you consider what is most useful in each situation, says Cloudera’s Regional Vice President Benelux Rein de Jong. “Sometimes a workflow is more efficient, sometimes an agent is better. It depends on the nature of the process, the data available, and how quickly you need to be able to switch.” He emphasizes that AI is not always the holy grail but a powerful component of a broader architecture.
Other roundtable participants agree with this comment. Practical applicability is best considered in advance, so you know exactly what to do to solve a problem. Because if you implement agentic AI for the sake of implementing it, you may be ahead of the curve and not have waited for the wave of innovation. Still, the investment may be completely unnecessary for your specific business situation.
However, according to Jan Verbrugghe, Senior Director of Solution Engineering at Salesforce, companies are quick to think in terms of technology. “But it’s not about the tool itself. It’s about the problem you want to solve and how best to do that – with AI, automation, or a combination of both.”
The story continues after the box below.
Seven lessons
Organizations that want to get started with agentic AI need a few pointers. During the roundtable, Cloudera, Pega, Salesforce, and ServiceNow also shared a number of recurring lessons. We have summarized them below.
Start with a real business question: Choose a problem that really bothers employees and customers.
Work with orchestration: Make sure your agents don’t tackle every task, but choose wisely where it is useful.
Ensure good data access: Without reliable and clear permissions, agents cannot deliver value.
Actively adjust: Use monitoring and feedback mechanisms to evaluate agent behavior and results.
Ensure compliance from the start: Traceability and governance are crucial building blocks.
Train your teams: Familiarize employees with AI and ensure they recognize its benefits in their own work.
Plan for scaling up: Think about how you can successfully repeat what works—across teams, tools, and departments.
From experiment to mature deployment
Fortunately, many companies have experimented with AI in recent years in the form of small pilots or defined projects. These were often internal applications, such as automating frequently asked questions or summarizing documents. Although these initiatives have been valuable for building knowledge, they have rarely been scaled up to broader organizational applications. The absolute breakthrough into every corner of an organization is still waiting to happen.
However, as became clear during the roundtable, this is now starting to change. What is striking is that the conversation about AI has now shifted to the highest management level. Previously, AI was mainly a topic for IT managers. Now it is being discussed with CEOs and business leaders. The reason? The impact of AI agents extends far beyond IT. It affects customer service, operations, product development, and strategic positioning.
At the same time, there is a growing awareness that AI cannot be deployed haphazardly. Companies need to think carefully about where generative AI adds value and where other forms, such as predictive or rule-based AI, are more appropriate. After all, not every process requires a self-learning agent. In some cases, traditional automation is the preferred option, as mentioned earlier. Reasons for this may include simplicity, cost savings, or reliability. At the same time, you shouldn’t try to do everything at once, warns De Jong. Budget, use cases, and relevance all need to be matched.
That said, the use cases are now piling up. Van der Putten cites the banking sector as an example. There, concrete applications of agentic AI are being demonstrated. AI agents are used for internal risk analysis and compliance. This shows that the mature use of agentic AI does not necessarily have to be spectacular or public. It often involves optimizing processes that have a significant impact but are invisible in their execution.
The lesson is clear. Companies that want to scale up must move away from project thinking. AI must be embedded in the broader business strategy, with clear objectives, measurable performance, and an infrastructure that supports agility.
Benefits for employees and the organization
The Cloudera, Pega, Salesforce, and ServiceNow experts agree that the benefits significantly outweigh the obstacles. An interesting observation is that AI agents deliver cost savings and enable personnel reorientation. Botter puts forward the idea of freeing up a certain percentage of an employee’s time. What can they do with that freed-up time? Can these employees then respond to new market opportunities they would not otherwise have had time for? This reasoning does not see AI as a threat to jobs, but as an opportunity to allow employees to perform more valuable tasks.
In addition, studies on employee sentiment are also being shared at the table. A clear shift can also be observed in this area. Almost half of employees now believe that AI agents positively impact work. They believe agentic AI can help automate repetitive tasks, allowing them to focus on more interesting work. A small proportion do suspect a negative impact, while another proportion are neutral.
Verbrugghe also shares experiences from Salesforce that show how the impact works in practice. The company has invested heavily in AI agents for support work. Since its implementation, Salesforce has seen a reduction in the number of calls received by employees in the personal contact center. This allows staff to be deployed differently. Verbrugghe emphasizes that in this case, it is not so much about laying off people, but about reorienting the company’s capacity.
Ultimately, it must be said that the term ‘agent’ quickly conjures up images of a jack-of-all-trades. However, AI is not magic. AI agents must be deployed in a targeted manner, with an eye for context and limitations. Not every problem can be solved with AI agents, but taking over repetitive, knowledge-intensive, or context-sensitive tasks that cost people time and energy is a clear benefit. It is precisely by using agents intelligently that room for real added value is created. The better the data and process definitions, the more effective the agent.
The future of work with AI agents
AI agents are changing the way people work. The phase in which organizations still doubt the usefulness of AI in the workplace is over. The question is no longer whether, but how and where you use AI agents to create value. It is striking that the need for smart support does not only come from management. Employees also want to be able to work smarter and are increasingly asking for AI tools. Employers who fail to respond to this demand risk losing talent to organizations that do.
At the same time, flexibility is becoming increasingly important for organizing work. Employees want to develop, learn new things, and contribute to meaningful work. By intelligently integrating AI agents, repetitive or routine tasks can be taken over. This creates space for employees to focus on innovation, customer value, or personal development.
The use of AI agents therefore requires a different way to look at work. Not as a fixed schedule, but as adaptive collaboration between people and technology. Organizations that manage to strike this balance can better anticipate change and make work future-proof.