Agriware Assistants: The hidden businesscase of AI

Agriware Assistants → From Experiment to Business Value

A few months ago, I wrote my first article about Agriware Assistants. At the time, the discussion mainly revolved around one question: what if AI not only answered questions, but actually performed work within business processes?

Over the past period, we have learned a great deal and are now applying this technology in practice with our customers. Dozens of scenarios have been developed, projects have been executed together with customers, and various business processes have been supported by our AI assistants. Agriware Assistants provide practical support for many processes that already take place today within horticultural businesses using Agriware Business Central.

In almost every conversation, I increasingly noticed that discussions about AI within horticultural businesses are no longer really about the technology itself.

It is about the value AI can add to a company, process, or role.

What business value does it actually deliver?

An assistant that answers a question is useful. An assistant that performs hundreds of checks every day, processes documents, detects exceptions, or prepares administrative tasks suddenly becomes much more interesting.

That is when real business value is created.

Especially within operational processes in horticultural businesses, this creates opportunities that were previously not economically viable. Many checks are not performed today because there is not enough time available or because the expected return does not justify the effort required. Not because they have no value.

Over the last quarter, together with our customers, we have experienced that AI changes this equation.

Where an employee may need 30 seconds to several minutes to perform a check, an AI assistant can perform the same check for a fraction of the time and cost.

A Different Way of Looking at ROI

In AI projects, we often see the same trap. Companies immediately try to identify a business case worth hundreds of thousands of euros. As a result, a smaller AI use case—saving perhaps €5,000 per year by automating a few checks or manual calculations—can quickly seem insignificant.

Customers using Agriware Assistants often experience the opposite.

In practice, the greatest value is usually created through multiple small improvements that together generate a significant impact. Moreover, these processes often would never have been automated because of the investment required.

By breaking business processes down into individual steps and measuring them against KPIs such as lead time, quality, and direct labor hours, a great deal of hidden value becomes visible.

A check that normally takes 30 seconds per inspection.

A document that requires an average of three minutes to process.

A work order that must be reviewed manually every day.

Individually, these tasks may seem minor. Multiply them by thousands of transactions per year, however, and the picture quickly changes.

We see that ROI calculations for AI Agents within business processes often consist of:

• The current processing time per process step
• Internal labor costs per minute
• AI consumption costs per transaction
• Monthly or annual transaction volume
• The value of error reduction and process improvement

With thousands of transactions per year, the business case can change quickly—and positively.

When Is AI Cheaper Than Human Labor?

To provide some indication, we have created several examples. Every organization naturally has different volumes, processes, and cost structures, but these examples illustrate how quickly direct AI costs can compare favorably to direct human labor.

Example 1: PDF to Sales Order

A common use case (and request) is processing a customer order from a PDF document.

Today, this often takes between two and four minutes per order. In some companies, it can even take up to 30 minutes because of the number of details involved.

With internal labor costs ranging between €0.30 and €0.60 per minute, this results in:

• Current cost: €0.60 to €2.40 per order
• AI consumption cost: €0.06 to €0.12 per order
• Direct savings: €0.48 to €2.34 per order

At 10,000 sales orders per year, this creates an estimated direct saving of:

€4,800 to €23,400 per year

But honestly, the greatest value is often not the direct saving itself.

The real value typically comes from the indirect benefits:

1. Fewer data entry errors
2. Shorter order lead times
3. Faster customer responses
4. Higher delivery reliability

Example 2: Sales Order Validation

Many businesses perform checks on sales order data at multiple stages. After order entry, just before sending the order confirmation, after picking the plants, and again shortly before shipment.

Typical questions include:

• Is the delivery address correct?
• Is the delivery date still feasible?
• Is the packaging correct?
• Is any required information missing?
• Have all customer remarks been processed on the order?

These checks often take between 30 seconds and one minute per order (and sometimes much longer).

At 10,000 orders per year, this results in an estimated direct saving of:

€1,400 to €5,800 per year

In addition, it prevents errors that later become far more expensive than the original check.

Think of credit notes, customer inquiries, or corrective work in logistics processes. I will avoid quantifying these indirect costs, but they are likely to be many times greater than the direct savings.

An AI Assistant Performing Thousands of Checks Every Day

Example 3: Reviewing and Scheduling Work Orders

Within production environments, we see many recurring reviews of work orders for activities such as potting, sowing, moving, propagation, and harvesting.

These reviews vary widely and often require specialist knowledge from someone who applies business rules directly within the Agriware system.

For example:

• Identifying missing information
• Detecting scheduling anomalies
• Automatically assigning work orders
• Cleaning up outdated or incorrect work orders
• Automatically moving planned but uncompleted work to the next working day

Per work order, the time involved may range from a few seconds to several minutes.

At volumes of 10,000 work orders per year, direct savings can range from:

Several hundred euros to more than €10,000 per activity.

It is equally important to note that issues become visible sooner. This helps prevent ad hoc rescheduling, urgent corrective actions, and disruptions elsewhere in the process. The value of this is also substantial.

Example 4: Time Registration Audits

Another use case that many growers immediately recognize is the review of daily and weekly labor hours.

Checking for missing hours, anomalies, or illogical registrations often takes management and administration more time than expected.

With 100 employees and weekly reviews, this can quickly result in:

€1,500 to more than €6,000 in direct annual savings

At the same time, businesses gain better management information because deviations become visible much sooner.

And let's be honest: reviewing timesheets is nobody's favorite task, yet it remains a necessary weekly routine.

Example 5: Quality Checks and Inspections

Inspection processes are also highly suitable for AI support.

An Agriware AI Assistant can for example:

• Evaluate recorded inspection values and potentially block a process
• Classify deviations and notify users in time
• Detect trends and generate alerts
• Recommend follow-up actions

The direct saving per inspection may seem very small and often appears not worth automating. In inspection processes, however, timely action and visibility are critical.

The real value lies in the indirect benefits.

Quality issues can be detected earlier before they develop into rejected products, customer complaints, or corrective actions later in the process.

If 1 out of every 100 inspections can result in a critical issue, and a company performs 20,000 inspections per year, there are 200 critical moments annually—roughly three to four per week—where immediate action is required.

Suppose a human error occurs once per month, causing an inspected batch not to be blocked in time and still be shipped, or forcing a process to be repeated. The indirect saving becomes substantial:

Twelve complaints per year cost at least the average order value—and usually much more.

Not Everything Needs to Be Fully Automated

Another misconception I frequently encounter is that AI only becomes interesting when a process is fully automated and roles are replaced.

That is not entirely true.

Much of the value comes from AI supporting employees with:

1. Monitoring
2. Validation
3. Summarization
4. Classification
5. Preparation

You remain responsible for the final decision, but you are simply supported by a tireless assistant.

This is especially true for processes involving financial impact, compliance requirements, or external communication, where human oversight remains essential.

Where Agriware Assistants Work Best

Our experience so far shows that AI is most successful when:

1. The use case is clearly defined
2. The input is recognizable and consistent
3. The desired output is clearly defined
4. Exceptions are well managed
5. Source data is of sufficient quality

AI is therefore not a replacement for a process or role.

In fact, the better a process is designed and responsibilities are clearly assigned, the greater the value an AI assistant can add.

The Next Step

What inspires me most is not the technology itself.

It is the realization that we can now support more and more processes that were previously too labor-intensive or simply not valuable enough to justify investment within horticultural businesses.

Not because employees are not doing a good job, but because there is simply not enough time to review, analyze, and follow up on everything.

AI makes it possible to continuously perform thousands of small checks, evaluations, and analyses at very low cost.

As a result, the question is no longer:

"Can AI do this?"

The more interesting question is:

"Which processes do we not perform today—or perform with errors—because they take too much time, even though they create business value?"

In my opinion, that is where many of the most interesting opportunities will be found in the years ahead.

Arno Hogervorst
AI & Innovation Director at Mprise Agriware, Senior Software Strategist for over 20 years.

‍