AI Agents or Custom AI Tools?

What a week! The world is picking up steam again after the annual holidays. Of course nothing changes suddenly when the clock strikes 12, but for many companies January is still a great moment to look ahead at new plans and projects. This fresh start is clearly palpable at our company too, because in 2025 we're really going to scale.

I sometimes have to pinch myself: by the end of January we suddenly have five people at NinA AI. And I'm also talking to a few more applicants.

Growth has been set in motion, with a brand new team eager to realize our ambition. The foundation is in place, our proposition is crystal clear, and we're increasingly moving into execution.

That doesn't mean we're stopping with lectures and workshops – those continue unabated, and we're happy to organize them here at our office (come visit sometime!).

But the focus is increasingly shifting towards fulfilling our clients' needs: becoming a real AI partner. We help companies with consulting, implementation projects and integrations in the field of AI. Two-thirds of the team now even consists of developers.

And precisely because of that, the same question keeps coming up from many (potential) clients:

"Olaf, should I buy an AI tool, or should we build and configure one ourselves?"

Because I notice this question is on many minds, I thought it would be useful to dive deeper into it this week. Additionally, I'd like to give you more insight into the 'AI Agents' that are currently in the spotlight.

AI Agents

The word of 2025 is already 'AI Agents'. Funnily enough, I wasn't so positive about that a few months ago; I said I didn't believe in AI Agents. And I'm talking about AI systems that supposedly perform tasks and make decisions entirely independently, almost like an 'AI employee'.

My current view: fully self-operating AI doesn't exist. An AI needs to know your company and processes inside out to truly add value. At the same time, all sorts of articles and research are emerging that portray AI Agents as the next logical step. In my eyes that's correct – provided we see AI as a collaboration partner for humans.

I strongly believe in the harmony between human and AI; a kind of 'hybrid colleague' that supports and advises you.

So yes, I now embrace the concept of AI Agents too, but for our clients we still call them 'AI implementations'. Because let's be honest, who wants to work with yet another software tool when you can also integrate AI into the tools you already use every day?

AI tool or custom?

This brings me directly to my first point: the great added value of a custom solution is that we can integrate AI into the software you and your team already use. As long as we can make an API call, we're usually good. And yes, that's possible with virtually all common software packages.

To be clear: by "AI Tools" I mean niche solutions here. Not the large language models like ChatGPT or CoPilot, but rather tools that perform very specific tasks.

Think for example of:

For all these applications, great AI Tools exist. And sometimes they're a perfectly fine choice, just like off-the-shelf software packages. I regularly recommend these solutions to clients.

But in conversations I also often hear two specific pain points:

And that's where a custom solution really shines. Where a software company might saddle you with an invoice of hundreds of thousands, we build (together with AI) MVPs in a few months for a fixed fee of a few thousand euros per month. That's a world of difference. The first version to test is often on the client's table within a few weeks.

Because the faster we can test and provide feedback, the faster a solution truly fits your organization.

Summary

Keep following me for some great client cases.

Until next time!

Olaf Lemmens
Founder, NinA AI Agency

P.s. Want to know more about what we do? Visit www.nina-ai.nl