How AI Agents Are Reworking the Business Workflows
Imagine your workday that is not defined by tedious tasks but by strategic thinking and creative problem-solving. This kind of transformative technology is not just a futuristic concept; AI Agents are here now and capable of automating mundane work, anticipating needs, and even collaborating with human teams to achieve unprecedented efficiency and innovation.
What Are AI Agents?
AI Agents are AI systems that work proactively and autonomously for a given goal.
In slightly more technical terms, AI Agents operate using dynamic workflows and AI frameworks, allowing them to observe, plan, reason, iterate over tasks, use various tools and act in their environment.
In simplest terms, AI Agents are our new digital colleagues and are transforming how organisations improve their efficiency and productivity.
Adapted from ServiceNow
The Impact of the New AI Agent Workforce
In the societal context of digital work, AI Agents represent a paradigm shift and a growing business tsunami. As a recent IDC research (Aug 2024) predicts, AI will contribute $19.9 trillion to the global economy through 2030 and drive 3.5% of global GDP in 2030.
AI Agents are bringing the digital and physical worlds closer together than ever before. In his CES 2025 keynote, NVIDIA’s CEO Jensen Huang envisioned:
“There's a billion knowledge workers in the world. It is clear AI Agents are probably the next robotics industry and likely be a multi-trillion dollar opportunity.”
Regarding the present opportunities for enterprises, ServiceNow’s CEO Bill McDermott says:
“AI Agents have become key to helping businesses orchestrate processes across departments and with other businesses.” (crn.com, Oct 2024)
About the reduction or transformation of present human work, McDermott feels optimistic:
“You're going to see companies not having to cut headcount because revenues will start to grow as they think about new ways of doing things, new ways of building products. So I don't believe you're going to see these AI Agents replacing people as much as you're going to see them improving the efficiency, the productivity and the growth curves of these companies.” (investors.com, Jan 2025)
Solving Real-World Problems with Agent Teamwork
Modern workflow platforms such as ServiceNow are in a good position to streamline their customers’ workflows using AI Agents, given such platforms’ rich access to business and customer data. Such workflow platforms typically manage customer requests or internal approvals in organisations, often in the context of customer service, IT support or internal HR.
The platforms also manage more complex, often multi-department or -stage organisational workflows. Examples include identifying and addressing recurring customer service problems, monitoring IT operations for anomalies and resolving identified root causes, managing manufacturing equipment and coordinating field repairs. Or in the HR context, coordinating full recruitment processes and cross-organisational onboarding of new employees.
Example of AI Agents in Customer and IT Support
To illustrate AI agentic teamwork in an environment such as Customer and IT support, a human might only provide the AI Agent with an initial “complex problem” in need of a solution and trigger the work unless the agent takes on the work proactively. Anyway, the AI Agent would get an assignment of work along the lines of “Solve customer problem X”.
After that, the “Project Manager AI Agent” receiving the assignment would internally assemble an entire team of available supporting AI Agents to solve the problem. The cross-functional agent team would consist of the best possible mix of specialist and generalist expertise to solve the project goal, in this example the customer problem, exactly like a human-only project team would.
When the agent team starts working, a “Guideline Agent” might proceed to analyse IT manuals, intra-chats and other knowledge bases to find earlier applicable solutions to problems similar to the one being solved right now and also, update knowledge bases afterwards.
“IT Support Agent” could analyse the logs and access rights, analyse the influence of IT-related incidents or settings related to the problem that had occurred, and propose possible configuration adjustments.
A “Legal Agent” might have also been called to join the team to check that actions taken by the team are in conformance with organisational policies or contractual agreements.
The AI Agent team might also have more special “Personal Agents” like the digital twin of yourself or your customer. It is aware of the personal or customer, organisation, culture-specific preferences for interaction and communication style or ways of working or solving problems, and all this knowledge is embedded in to create a more personalised service delivery experience.
The initiator, “Project Manager Agent”, would continue to coordinate with the entire agent team about the progress and, depending on the situation:
Propose the next best team actions to take,
Come back to the human user with a request for additional important information,
“Request to hire” additional specialist agents to complete the job,
And ultimately, conclude the problem as solved.
More Complex Use Cases for AI Agents
While the above example of customer and employee support service use cases can be somewhat simple and also implemented in a more predefined (non-agentic) workflow in a more narrow business domain, it illustrates the high degree of autonomy the AI Agent project team can take when working either completely proactively or by the human initiator.
The capabilities of agentic AI systems to observe, plan, reason, and act in their environment can, however, also be expanded to include very complex and physical use cases.
For example, think of agentic AI for running an entire manufacturing factory. An AI that could be developed, optimised, tested and benchmarked (vs. competing factory management AIs) in safe isolation against a digital twin replica of the same factory before it can securely and with human control points run a real-world physical factory. This factory AI can simulate both the physical elements of the factory as well as the actual laws of physics. This kind of use case is not a mere vision but already a reality based on the case demonstrations of NVIDIAs Jensen Huang in the CES 2025 keynote.
Simplicity Is a Friend in Agentic Software Design
While pure AI Agent systems don’t rely on predefined deterministic workflows, they can come with cost and performance tradeoffs for computational complexity for time-extended reasoning. And that in itself is a topic for another blog post.
While the implementation of agentic systems comes with requirements for new ways of interacting with various LLMs, the technical design principles might not come across as transformative as the changes at the level of business or agent-driven user experience.
For the development of agentic systems, the basic software engineering principle of adding additional complexity only to the extent it is needed for delivering (business) value applies. As reminded in this text, you should start with the simplest design pattern that achieves your project goal.
Building End-to-End Process Intelligence with AI Agents
In conclusion, AI Agents are already revolutionising how businesses operate, and the rate at which this change occurs will only increase. On your journey with this transformative technology, consider how to streamline and improve your end-to-end business processes across the entire organisation. With the expertise of your business and process owners, modern AI Agents and process intelligence tools will cross the boundaries of various organisational, process, platform and technical silos.
Pekka Lehti
Pekka has over 20 years of experience in IT consulting and creating value for customers. He is a strong believer in continuous improvement, which drives him to learn and look for new ways to develop business operations in customer organisations.