No AI panic on the AI tanik?!
A few years ago, I first heard about the ‘1-metre rule’ during a talk given by a Dutch managing director. You may recall that I’ve written about it here before. Essentially, this rule suggests that anyone who is within a metre of a problem is likely to have a sensible idea about how to solve it. I’m still convinced by this idea. However, in the current discussions about AI agents, I find myself wondering whether we don’t sometimes experience exactly the opposite: the further away someone is from the problem, the simpler the solution seems to them…
This brings to mind the image of the Titanic. Imagine the ladies and gentlemen on the upper decks, still lulled into a false sense of security, whilst water is already pouring in further down. Upstairs, they place their trust in modern technology. Downstairs, they’d simply like to have a pump that works.
Fortunately, the companies we advise are not sinking ocean giants. But why not ask your management team whether your company is ready for AI agents? And then ask the same question of the people who handle daily scheduling, adjust production plans and chase up missing materials over the phone. I suspect the answers will be far from consistent.
At the top, they dream of automated agents that independently trigger orders, resolve bottlenecks and optimise production schedules. At the bottom, they’re all too familiar with the unresolved conflicts of interest between departments, the stock levels that don’t add up, and the delivery times that are more like historical footnotes.
My assessment is therefore quite clear: most companies are still a long way from being ready to let AI agents make operational decisions independently. Two key tasks have often simply not been completed.
The first is the definition of clear business rules. Who receives the goods if there is not enough to go round? Which order is allocated the last available component? When a production line is overloaded, what takes priority: meeting delivery deadlines, contribution margin or minimising set-up time?
Of course, these questions are answered these days too. However, this often only happens once the bottleneck has already arisen. And sometimes it is the customer who last rang the managing director (or plays golf with him) who indirectly determines the order of production. That, too, is a form of decision-making logic. Whether one wishes to automate it, however, is something one should think carefully about…
The second challenge is data quality. After all, even the best decision-making logic is of little use if it is based on incorrect information. An experienced planner may know that a particular supplier generally delivers later than promised. They also know which stock items warrant a call to the warehouse. The AI agent would also need to be able to reliably take this knowledge into account. However, to begin with, it receives only the data that we provide. And an incorrect figure does not become correct simply because an artificial intelligence processes it. If anything, it is more likely to form the basis of a wrong decision more quickly.
Before you jump to the conclusion that I am against AI agents: quite the opposite! But anyone wishing to delegate responsibility to them in the medium term must define their business rules today, fundamentally improve data quality and clarify when a human must make the decision. This work does not diminish our enthusiasm for the new technology.
At Abels & Kemmner, we have had very positive experiences with developing binding business rules, incorporating them into rule sets, and systematically improving data quality through automated rule sets. That certainly sounds less exciting than an autonomous AI agent. But that is precisely how you get the essential groundwork done. And you are already achieving improvements today that help in day-to-day operations.
For me, this is the sensible way to gradually introduce a company to the use of AI agents. The organisation, so to speak, grows into this way of working: decision-making rules become transparent, data more reliable, and automated processes become standard practice. On this basis, AI agents can then gradually take over many decisions in a meaningful way. After all, there’s no need to put the whole ship on autopilot straight away…
My suggestion, therefore, would be to apply the ‘1-metre rule’ before embarking on your next AI project. Go below deck and ask the people who, despite all the shortcomings, are keeping your processes running today what is still standing in the way of an automated decision. The answers are likely to be less impressive than the next AI presentation. However, they’ll show you where to start and ensure that your AI project doesn’t turn into the ‘AI-tanic’.
FAQ – Frequently Asked Questions
What requirements must AI agents meet in supply chain management?
For AI agents to be able to take over operational decisions in the supply chain, companies need reliable data, clear business rules and defined responsibilities. In materials planning, production planning and inventory management, for example, it must be established how conflicting objectives are to be handled and when a human should make the decision. Abels & Kemmner supports companies in laying these foundations for the gradual automation of their planning and control processes.
Why are business rules important for AI-powered scheduling?
Business rules define the criteria according to which decisions are made. For example, they determine which order takes priority in the event of a material shortage, or how on-time delivery, contribution margin and set-up costs are weighed against one another. For AI-supported materials planning, these rules must be unambiguous and consistent. Binding sets of rules make decisions transparent and provide a basis for the automation of supply chain planning.
What role does data quality play in the use of AI agents?
AI agents can only make sound decisions if the underlying information is reliable. Incorrect stock levels or unrealistic delivery times can lead to inappropriate order recommendations and production plans. That is why improving data quality is one of the key prerequisites for AI in supply chain management. Automated rules can help to systematically identify anomalies and improve data quality in the long term.
Can AI agents take over materials planning entirely?
The extent to which materials planning can be automated depends on data quality, decision-making rules and the specific exceptional circumstances. A sensible approach is to automate clearly defined decisions in stages. Where there are unresolved conflicts of interest or insufficient information, defined limits and human judgement are required. This allows responsibility to be transferred in a controlled manner whilst the company gains experience with automated processes.
How should companies prepare for an AI project in the supply chain?
The first step should be to address the actual problems in scheduling and production planning. Staff involved in day-to-day operations can highlight which information is missing and which decisions have so far relied on experience. Companies should then align their business rules, improve data quality and define when human approval is required. This preparatory work lays the foundations for practical supply chain automation using AI agents.
How does Abels & Kemmner help with preparations for AI agents?
Abels & Kemmner combines supply chain consultancy with the systematic design of planning and scheduling processes. A&K supports companies in developing binding business rules, mapping these into rule sets and improving data quality through automated rules. This results in transparent decision-making and more reliable processes right from day-to-day operations. At the same time, the company lays the groundwork for gradually integrating AI agents into its supply chain processes.
