The Right AI Agents for Greenhouses - What we found

Over the past year or two, we have looked at a lot of AI solutions for our Agriware customers. Inside Business Central. Outside Business Central. Copilots, Power Platform, Agentic add-ons, Third-party platforms, AI frameworks. We tested. We researched. We prototyped.  

We joined sessions, events, and hackathons. We failed often, questioned ourselves, and sometimes wondered if we were too small to take this on. At one point, we even asked: should we just stop?

But we kept going. Because we still believe AI will become disruptive in our industry, but only when it fits the real operational work of growers.

That said, I can tell that it's not easy investing in something powerful that simply isn't yet good enough for our customers while keeping up with the latest news and technology.

Our Turning Point: OpenClaw Agentic Execution

Late 2025, something big changed: OpenClaw was released.

One person built it and directly had global impact. And while traveling, he built it using coding agents, remotely controlled via WhatsApp.

Let that sink in for a moment.

If one person can do that (and yes, he's a brilliant mind), then surely, we should be able to take some of those concepts and apply them to Agriware, right? This changed our thinking completely from both "can we develop this ourselves" and "can AI work for our customers" perspectives. If one person could build a globally impactful tool using agents, then a specialized team like Agriware could absolutely build an execution layer specifically for greenhouses.

What would it take for AI to be useful in a greenhouse?

Somewhere in the last 2 years, most of us saw the slick demos and heard a lot of promising hyped buzzwords... But working in horticulture means you just need to get your daily work done. Period.

Because the operational reality at our customers looks like this:

• Tight margins, with labor and energy often representing more than 70% of direct costs
• Production planning with tough constraints on space, labor, and utilities.
• Logistics that must be right the first time on picking, packing, loading, and shipping.
• Complex finance flows around costing, claim-credits, or late payments.
• A constant stream of changes to orders, planning, and master data.
• Unreliable connectivity: Some customers simply can't move to the cloud due to rural locations, peak seasons, and unreliable internet.

In the last 15+ years, all these typical business challenges already live inside Agriware on Business Central. But until now, there were no AI tools that could handle it in a way that was reliable, cost-efficient, and privacy-friendly (all at the same time).

So instead of asking "which third-party AI tool should we plug in?", we asked our customers: what would actually make a difference for your daily processes? Which tasks take the most time or are prone to mistakes, typos, and rely too much on human availability?

What Makes Agriware AI Agents Different?

This quarter, the first Agriware customers will start working with what we call Agriware Agents. Little helpers who assist the end-user in any role in the Agriware system.

So, what makes the Agriware Agents different and why does it matter for growers:

1. It’s fast. If you've felt that most AI tools are slow, you're not alone. Agriware Agents are built for execution speed. And you control how fast they act, with or without reasoning.
   ‍
2. Cost-effective by design. Burning tokens is a new business model for AI companies. With Agriware Agents, we don't wasted AI tokens on processes that do not need AI. We only use AI where it adds value, and let it execute traditional Agriware functions just like a user would. Lightweight, predictable, and efficient.
   ‍
3. Full data control. You decide exactly which data is shared, down to the field level. Giving you full control of your own data.
   Your data stays yours. With the AI models from Agriware Intelligence, your data runs on dedicated resources, exists only during the session, and is never used to train AI models.
   ‍
4. Works on-premise and cloud. No dependency on Microsoft Copilot. No forced move to the cloud while your are active in rural area. Agriware Assistants work with both Business Central cloud and on-premise environments.
   ‍
5. It understands horticulture. Most AI tools don't understand Agriware or the business processes behind growing operations. Ours does. And it can be fine-tuned with simple, human-readable instructions. When your company grows or your processes change, you just update the instructions yourself.
   ‍
6. Predictable outcomes. Agriware Agents are not just AI models guessing words. Our framework makes it intentionally very deterministic. If you ask it to assign a work order to a production manager, it will do that right: every single time. Same input, same outcome. Only the explanation might differ slightly (because it should still feel human).

We did not build this because AI is trendy. We built it because greenhouse operations need speed, control, privacy, predictable outcomes and horticulture-specific understanding.

This is just the start. And I'm genuinely curious to see where this goes in the coming months together with our Agriware customers.

Arno Hogervorst Director Agriware Innovation and AI

Interested in joining? We have a limited number of places available for Agriware customers, so feel free to reach out.

‍