From thinking to doing: how AI-agents finally can do work for you

Imagine this: you arrive at the office on Monday morning and while you pour your first cup of coffee, an AI assistant has answered simple emails, updated the schedule for the week as well as handled a number of routine tasks.

Sounds like science fiction? Until recently, that was indeed the case. But we are on the eve of a valuable shift in the AI world: from thinking to doing.

AI-agents: from prediction to execution

The current generation of AI systems, such as ChatGPT, is incredibly good at generating text and images and therefore answering all kinds of questions. Yet these AI systems also have two fundamental limitations. The first is that they can tell you what an apple tastes like and generate a beautiful image of it, but they don't understand what an apple is. I don't see that limitation being solved for the time being.

The second fundamental limitation is that it is models that make the predictions: the next pixel, or the next word. But actual execution, these models cannot do.

Think of it this way: it's like having a great consultant who gives excellent advice, but doesn't actually understand anything about the subject matter. So you're on your own when it comes to executing based on that good advice. But change is on the way: the inability to execute is now changing radically with the rise of AI agents.

What are AI-agents? 

AI agents are systems that not only think, but also do: they bridge the gap between prediction and execution. An agent can perform tasks, communicate with other systems and make decisions without constant human intervention.

And the beauty? You don't have to be a programmer to take advantage of this.

AI automation in action: smart applications for entrepreneurs

What might that executive AI work look like? Here are a few examples:

• Automated appointment scheduling

An AI agent can manage your calendar, check open time slots and schedule appointments. Imagine this: a customer sends a WhatsApp message for an appointment and an AI agent schedules things without you having to do anything.

• Inventory management

An AI agent can monitor your inventory levels and automatically place orders with suppliers.

• Social media poster

An AI agent can automatically post an article you've written to your social media channels.

And in case you think “that's not possible yet”, let me tell you that in two hours I built a basic AI agent that:

I could ask to make a calendar appointment and invite the colleagues I listed, while it was also able to find a free moment for that for the specified time in all our calendars.

Are you thinking “but I can't program at all”? That's really not a problem. There are more and more tools available that allow you to build AI agents without writing code. And if you want to build such a system, there are two possible solutions.

1. Give you AI superpowers with Model Context Protocol

In the first option, you grab an AI tool, like Claude, that you then link to systems that can perform work. For example, you might give Claude access to the files on your computer so the tool can manage them. Or you give Claude access to Notion so the tool can retrieve articles or write new articles there. Another option is to give Claude access to YouTube so the tool can create summaries of videos.

You could even go a step further and give Claude access to a browser or search engine API that would allow the tool to look things up for you and - shocker - even order things for you. The latter is something an AI tool refuses to do so far, but technically it's possible.

Claude uses Model Context Protocol (MCP) servers for this purpose, and these allow AI systems to execute code better and thus talk to a browser or a communication tool such as Slack. These MCP servers are standardized and thus work not only with Claude, but also with other AI systems that support MCP.

Want to know more about Claude and MCP? Then check out this video.

2. Use a work flow tool which supports AI

There are several workflow platforms with AI support. Two examples are Relevance AI and N8N. Make and Zapier are alternatives whose AI capabilities I have not tested. Whereas Relevance is just a little simpler, N8N has many more features because the tool has more integrations. In addition, N8N allows you to reuse solutions - templates - built by others.

In doing so, you can run N8N on your own systems, which solves a very important privacy and legal issue, because, among other things, you are not allowed to just use customer data in third-party systems like ChatGPT or N8N, which I will write about in detail later.

Read more about building agents in N8N or go one step further and build your personal assistant in N8N.

When to use what? 

Use ClaudeAI primarily for your own use and personal productivity, and choose N8N if you want to automate work processes. I suspect there are also MCP servers for N8N by now, which will allow you to use N8N again from Claude.

💡 Tip: Start small. Pick one specific, repetitive process in your business that takes a lot of time but doesn't require a lot of creativity. That is the perfect candidate for your first AI agent.

Implementation without code: AI automation for everyone

An important consideration when using AI agents is privacy. For an agent to be effective, it quickly needs access to a lot of data, including sensitive company or customer data. There are restrictions on that within the GDPR, and rightly so.

A simple example is e-mail. If you give an agent access to your e-mail to search e-mail addresses among contacts for appointments, they can, in theory, also read the content of e-mails.

Want to deploy AI for privacy-sensitive corporate data? Then follow these practical steps to do so securely.

• Use platforms that support local or private deployments

• Start with non-sensitive data sources
• Be transparent to customers about how their data is being used
• Provide clear boundaries on what the AI agent can and cannot do

It might also be a good idea to take a peek at Voys' AI ethics and usage policy and be inspired by it. After all, it's really important that you handle your customer data properly. I recently wrote an extensive article about this for the launch of our new Call Transcription feature.

AI solutions for small and medium businesses

As a small business, you have one tremendous advantage over large companies: agility. You can experiment and implement faster than large and unwieldy organizations.

If you start using AI agents now, you can:

• Free up time for strategic work while automating routine tasks
• Improve the customer experience with faster responses and 24/7 availability
• Save precious hours that you can invest in growth
• Position yourself as an innovative player in your market

To help you do that, I like to make it concrete: what can you do today to get started?

From AI concept to functioning automation in 5 steps

Now that you know why AI agents can be so valuable to your business, you want to get started with them as soon as possible. These five steps will help you do just that.

1. Identify the “low-hanging fruit.”
   What tasks take a lot of time but require little creativity or human insight? In particular, look for administrative tasks that you either find very tedious or processes that take up too much of your time.
2. Explore available tools
   Start with platforms such as N8N or Claude and experiment with MCPs.
3. Start a small experiment
   Start with a non-critical process and document the results and then learn from them.
4. Involve your team
   Make sure your colleagues understand that AI agents are there to take over repetitive work and certainly not to replace people.
5. Tweak and improve
   Monitor results and adjust as needed. Especially if the beginning is a lot of learning and tweaking to get to the desired results. You especially need to learn how an AI “thinks” and “acts,” and once you get the hang of that you will soon start getting much better outcomes and building solutions.

So above all, start with small but meaningful steps. Rome wasn't built in a day, and neither will your AI agent ecosystem.

AI automation: the future of productive entrepreneurship

We have arrived at a fascinating point where AI is no longer just a thinker, but is becoming a doer. AI agents offer a revolutionary way to automate business processes and increase personal productivity. For small businesses, the technology opens doors previously accessible only to large organizations with extensive IT departments.

I am convinced that entrepreneurs who now invest in understanding and implementing AI workflow automation will build a huge advantage in their market segment. Not because they use more technology, but because they can spend more time on what really matters: creativity, customer relationships, and strategic growth.

So, where do you start? What automation process in your business would you like to implement tomorrow if you could?

By the way, did you know that with Voys, we recently launched Call transcriptions based on privacy-friendly AI? A super valuable feature for organizations that really want to deepen customer relationships.