Vertec meets AI: What suddenly becomes possible

Analyzing Vertec data for two years. Processing accounts payable. Prepare reminders. Prepare measures. What do these tasks have in common?

With the help of AI and Vertec, StackWorks is approaching them in a very different way than they were a short time ago.

We spoke to our customer Ishan Don, Founder & Managing Director of StackWorks AG, about how his team is using the Vertec MCP Server – and how initial experiments have become solid business processes.

What is behind the Vertec AI app?

The Vertec AI app allows Vertec to connect to AI clients such as Claude or ChatGPT. This allows an AI client to query information, use Vertec functions and features and perform actions, depending on the user rights in Vertec. Technically, the Vertec AI app is based on the Model Context Protocol (MCP), an open standard for communication between AI clients and other systems. The Vertec MCP Server establishes the connection to Vertec.

Simply put, the Vertec AI app gives AI a structured access to Vertec. Instead of thinking about where to find information or what process it takes to find it, you can describe what you want to achieve.

It all started with a simple question

When StackWorks started connecting AI to Vertec in the early summer of 2026, the Vertec AI app did not yet exist. So the team first built their own, read-only wrapper around the Vertec REST API. With the release of the official solution, StackWorks switched to it.

One of the first attempts was very simple: “Search through all the services of the last two years and summarize in which business areas we provide services.”

The result: 511 days in which services had been entered services – processed and summarized by business areas. The evaluation was near to what StackWorks expected from its own figures. It was nonetheless cross-checked against the Vertec evaluation. Anyone who flips an AI result into a decision without checking has misunderstood the tool. For Ishan Don, this was clear: The combination of AI and Vertec can do much more than simply retrieve information from the system.

An analysis turns into a weekly run

The first analyses resulted in a fixed weekly run, internally called “Vertec Hygiene”.

Behind this is a skill: a manual that gives the AI client a defined task and a fixed set of rules. For example, a search is made for:

• Services that would be chargeable and remain unpaid.
• Quotes without decision.
• Phases that run out of budget.
• Provisional invoices that no one looks at anymore.
• Anomalies that easily disappear in day-to-day business.

The run only reads and does not change anything in Vertec. Each finding is quantified and given a responsible person and a deadline. At the end there is a report that you can process. Not a block of text that you can click away.

Four skills, one month’s end

Vertec Hygiene is now one of four runs that cover essential parts of the business month at StackWorks. Billing Clearance checks services against agreed terms. Expense Margin Control reconciles billed outlays with procurement. The Account Payable run enters invoices from freelancers and assigns them to projects. And Vertec Hygiene finds what is left behind.

Each skill has its own set of rules, each run ends with a report that a human approves. The result no longer depends on who runs the run.

From considerable hours to minutes

The change is particularly evident in accounts payable. Stackwork processes a large quantity of these. In the past, individual import solutions were built for this purpose. With Claude and the Vertec AI app, Ishan Don has further developed and expanded these processes himself.

What used to require significant hours and support from Vertec, StackWorks can now do on its own in a short period of time. This shows an important effect of AI: it not only speeds up processes. It also changes who can do it. With the help of an AI client, customers can self-develop solutions and automate tasks that used to require specialized technical expertise.

And then comes the reminder

The dunning procedure at StackWorks also now follows a written procedure: From what amount is escalated?

• Who receives the reminder and who belongs to the CC?
• What is the deadline?
• When will a block be announced?

The AI client can apply these rules to the data in Vertec using the Vertec AI app, identifying the relevant cases, checking them against received payments, and preparing the letters.

Nothing is sent automatically. Every reminder is ready as a draft. The last click remains with the human. This is precisely the key point of automation with AI: Not every process needs to be fully automated. Often, the greatest benefit comes from eliminating repetitive work steps.

Measures that no one else compiles by hand

StackWorks also retrieves measures directly from Vertec via the Vertec AI app and has them built into a dashboard: revenue versus monthly target, utilization, open fees and open items.

What is decisive is not the graph, but the definition behind it. For each measure, it is determined which Vertec field it comes from and which rule applies. This shifts the discussion in the management meeting. It no longer revolves around whether the figure is correct, but what to do with it.

A report that sends itself

Every Monday and Wednesday, the StackWorks team’s Google Chat automatically displays an overview of the work that has not yet been billed for the current month.

The report runs as a scheduled task in Vertec. A Python script reads the corresponding Vertec view and posts the result to the chat via a webhook. The script was written by the AI Client based on the Vertec Knowledge Base. StackWorks tested, adapted and implemented it. The effect is less technical than cultural: Open Fees are no longer an issue that someone has to address. It is posted twice a week in the chat, visible to everyone.

The lever is not the prompt

People who try the Vertec AI app for the first time ask questions. This is impressive, but often remains inconsequential, because the next time the question is asked differently and yields a different result.

The difference arises where a good question becomes a written procedure: Which rule applies? Which value is plausible? Which trap is known? Who decides in the end? At StackWorks, this knowledge is in skills: versioned, the same for the entire team and regardless of who starts the run.

The skills are the operating manual. The Vertec AI app creates the access. The process is the yield. “The MCP server gives us the access. The yield only arises when we write down how we work,” says Ishan Don.

This is especially important for a team of less than ten people: “We are not so far because we are big. We are so far because we are small and have written down how we work.”

Vertec as a “System of Record”

Of course, knowledge about a customer is not just in ERP. Emails, documents, and other applications also contain important information. AI clients can combine such sources and create, for example, an overview of a customer relationship.

a centralized and reliable data base becomes all the more important. Vertec remains the “system of record” for the binding business data. AI can analyze this data, establish connections and prepare processes. The Vertec AI app combines the structured data and processes in Vertec with the capabilities of modern AI clients.

The AI works with the data. Vertec remains the instance in which it is managed.

Where manual work is still needed today

Boundaries are also part of this. For example, large evaluations have to be split into sections, since a query returns a maximum of 100 objects. Derive fields cannot be filtered directly. And StackWorks executes bulk changes script-driven and only after approval.

None of these limits speaks against the use of the Vertec AI app. They speak in favor of checking results and designing processes consciously.

The human remains in the loop

For Ishan Don, this is crucial: There is still a need for a human in the loop. No run sent, posted or deleted without approval. The AI client prepares, the human decides. This control remains central, especially in the case of business-critical processes. AI can make suggestions, analyze data, and prepare processes. Responsibility for the result remains with the human.

And this is just the beginning

The examples from StackWorks show that AI in ERP is no longer a theoretical vision of the future. Vertec Hygiene runs weekly. The report on the open fee is published twice a week. The charging is checked monthly. Dunning, measures and other parts of commercial operation are supported with skills and the Vertec AI app.

It’s not about replacing Vertec with AI. Vertec remains the reliable database and the system of record. AI applies where it can play its strengths – in analyzing, linking, structuring, and preparing. Human is in control. What started with an initial analysis at StackWorks shows, above all, what is possible if ERP and AI are not thought of separately:

Questions become rules. Rules become skills. And individual experiments become fixed processes.

About StackWorks

StackWorks AG supports around 200 Swiss Sme in the areas of Security & Compliance, Modern Work and Automation & AI. The company is a Google Premier Partner and Microsoft Partner. Ishan Don is Founder & Managing Director of StackWorks.