Ask-Sabine processes

The biggest AI opportunity in the German Mittelstand is not the next pilot project. It lies in the undocumented exceptions that today only Sabine knows.

After 76 AI transformation projects, one lesson stays with me above all: most companies do not have an AI problem. They have an implementation problem.

The technology works. The demo works. Employees even save time. But surprisingly little of it reaches the P&L.

Why? Because we put AI on top of old processes instead of rebuilding the processes. The AI produces the analysis in 30 seconds. Then someone copies the result into Excel. Sends an email for approval. Transfers data into the ERP. And at the first exception, the answer is: “Ask Sabine. She knows how it works.”

What an Ask-Sabine process is

Every company has them. The process is in the manual, and on paper it is clean. In reality it only runs because one person knows the exceptions. The supplier who always sends invoices without a purchase order number. The customer who insists on the old discount code. The approval the managing director only gives verbally.

None of it is documented. It lives in Sabine's head. She knows what to do, and she has done it for years. Nobody has ever asked why.

These processes are not particularly glamorous. Invoices. Contracts. Approvals. Compliance. Reporting. Exceptions. But that is where a great deal of work sits. And that is exactly where most AI projects fail. The tool handles the standard case. At the first exception, a human takes over again. Then you have a new tool, but no transformation.

Why the pilot project misses the exception

A pilot project picks the clean case. Good data, a clear workflow, a department that plays along. That is understandable, but it measures the wrong thing. The standard case was never the problem. The problem is the cases that run differently.

I see the pattern again and again: AI must not just make people faster. AI must make work disappear. As long as the exception lands with a human, no work disappears. It only gets moved. Sabine now has two jobs: her old one and the rework behind the machine.

That is why the opportunity is not in the next big pilot project. It is in the hundreds of small Ask-Sabine processes that happen every day across the company.

How you find the processes

You do not need to buy software for this. You need to listen. Three tests have proven themselves for me:

• The holiday test. Which cases stall when one particular person is away for two weeks? Every one of them is an Ask-Sabine process.
• The corridor test. Sit next to the clerical team for a day. Count how often someone gets up to ask a question. Every question marks an exception that is written down nowhere.
• The manual test. Take the process description and walk a real case through it. At every point where reality departs from the paper, you find hidden work.

The result is a list. Not a strategy, not a roadmap. A list of exceptions, sorted by frequency. That is the raw material for everything that follows.

How you rebuild them

This is where the Task Architect helps. I break the role down into its individual tasks, as concretely as possible. For each task I decide whether a human or a machine takes it on. Then I rebuild the process instead of putting a tool on top of the old workflow. In the end, the measure is the P&L, not usage of the tool.

An example from accounts payable processing. The role breaks down into eight tasks. AI can take over five of them today: capturing receipts, transferring data into the ERP, checking invoices, preparing approvals, producing reports. Three stay with the human: deciding exceptions, negotiating with suppliers, improving the process.

The decisive point sits in the first human task. Deciding exceptions does not mean Sabine keeps resolving every exception by hand. It means she formulates the rule that applies next time. Every exception she decides once becomes a rule in the system. Sabine no longer works on the case. She works on the process.

That is the difference between a tool and a transformation. The tool makes the standard case faster. The transformation makes sure the exception is no longer an exception the second time round.

What happens to Sabine

The role does not disappear. It changes: less typing, more deciding. For many employees that is a threat at first. Their knowledge of the exceptions was their security. Once it is in the system, it no longer belongs to them alone.

I tell every team we take over this on day one: your job will not disappear because of AI. It disappears if you do not use AI and the colleague next to you already does. Whoever knows the exceptions is the most valuable person in the project. Not because they keep solving them by hand. But because only they know which rules the system needs.

The biggest AI opportunity in mid-sized companies is not in the next demo. It is in the exceptions that today sit in the head of a single person. Whoever finds these exceptions, documents them and turns them into rules makes work disappear. Whoever does not buys a tool and keeps the old process. In 2026, AI has to reach the P&L. The way there runs through Sabine.