What can you do with AI agents?

In the blog “What are AI agents and why are they becoming increasingly important?” we explained what makes AI agents different from the forms of AI we already know. Where traditional AI mainly supports thinking work, AI agents carry out actions independently. But what does that mean in practice? Below we share three scenarios where AI agents are already running in production today, including examples from our own experience.

Incoming contracts, expense claims or reports? An AI agent can automatically open these documents, extract the relevant data, check if they meet requirements and record the output in the right system. If everything is correct, the file is closed automatically. If something is missing or unclear, the agent automatically sends a notification to the right colleague, including context and a suggested next step.

Example from practice: A healthcare provider processes hundreds of supplier invoices daily. Thanks to an AI agent that analyses PDFs, recognises amounts, flags exceptions and posts everything to the financial system, processing time has been halved. Human intervention is only needed for exceptions.

Agents are also valuable on the customer side, especially when they work alongside staff. A support agent can interpret incoming emails, retrieve relevant information from a knowledge base or CRM, generate a draft reply, and, if needed, create a follow-up ticket in another system. All fully transparent, repeatable and trainable. 

Example from APG: Together with Future Facts Conclusion and Conclusion, Dutch pension provider APG uses generative AI to improve customer interactions. Not with a standard chatbot, but with an intelligent assistant that prepares conversations, suggests responses and enriches customer contact with data from multiple sources. The employee stays in control but is supported and empowered.

Commercial teams lose a lot of time on administrative tasks. Think of updating notes, creating follow-up tasks, or sending summaries. An AI agent can take over much of this. Based on emails or meeting notes, the agent automatically fills in the right fields in the CRM, schedules follow-up actions, and informs relevant colleagues.

Example from practice: An account team uses an AI agent that automatically generates a follow-up proposal after each client meeting, including CRM updates, status changes, and calendar items for the next contact moment. Nothing gets forgotten, and staff have more time for real conversations.

A fair question to ask would be: isn’t this just clever automation? The difference lies in the goal-oriented and adaptive nature of AI agents. They don’t follow rigid, pre-scripted steps. Instead, they work with an objective and decide for themselves which route is most effective within the boundaries you set. AI agents are agile. They can handle variations, exceptions, or missing information. Precisely when the standard flow breaks down, AI agents show their value.

Beyond that, AI agents can switch seamlessly between multiple systems. Think of an AI agent that analyses an email, looks up additional information in a knowledge base, creates a task in the ticketing system, and automatically logs everything for compliance. The combination of coordination, context-awareness, and adaptability is what makes it agentic, and fundamentally different from traditional automation.

At Conclusion, we help organisations deploy AI agents smartly and responsibly. We’ve found that successful pilots often meet three important conditions:

1. Start internally and small
   Choose a process that’s internal and not directly customer-facing. Let the agent support rather than make autonomous decisions. This limits risk, maintains control and builds trust. Think of tools for staff rather than public interfaces.
   
   
2. Choose repetitive and common work
   Focus on tasks that happen often, are well understood, and currently require a lot of manual effort. Examples include processing meeting notes, drafting reports, or logging tasks. The impact is quickly measurable (e.g., time savings), and staff immediately see the value.
   
   
3. Focus on information-driven tasks
   AI agents perform best on tasks where existing information can be used intelligently. For example, from manuals, customer data, or internal knowledge bases. They build on what’s already there and help enrich those sources.

Well-chosen use cases deliver not only working agents but also better documentation, smarter processes and sometimes even new insights into how work is organised.

AI agents aren’t just valuable in processes but also at the individual level. More and more employees use tools to summarise text, rewrite emails, or generate action points. By structuring this further, for example via a personal work assistant, AI agents can contribute to greater focus and higher effectiveness.

Still, the real value lies in integration with processes. That’s where scalability emerges. Then you can measure, improve, and repeat. And that’s when it becomes truly structural.

Reading tip: Read our blog Human-centred technology as a strategic foundation.

AI agents aren’t a distant dream. They’re already usable today, in small, manageable contexts, with direct impact. You don’t need to start big to create value.

In fact, the analysis that precedes building an AI agent often delivers value on its own. You discover inefficiencies, clarify work agreements, and improve documentation. This way, your first AI agent becomes a building block for trust, collaboration, and further exploration. 

At Conclusion, we believe AI agents help organisations work smarter. With technology that fits both the practical reality and people. And it all starts with one well-chosen, practically deployable use case.

In the next blog, we’ll take a closer look at the strategic question: How do AI agents fit within broader digital transformation and organisational goals?