Warning 1: this article is not (too) much about artificial intelligence, and not at all about generative artificial intelligence, or LLM. If you’re a fan of shiny stuff that contributes to depleting the planet’s resources by wasting as much energy as possible, you’d better move on!
Warning 2: if you think that supply chain management is an art or a craft that requires human intervention for every decision – in ERP or Excel – you’d better move on too, as your convictions may be shaken.
Warning 3: if you think that an APS system can take into account all constraints, all hazards, and in real time decide what needs to be done automatically – sorry, many have believed in it, tried it, and fallen flat on their faces – don’t bother.
Are you still here? It probably means you’re a pragmatist, and believe that many routine decisions can be largely automated while maintaining human control over exceptions. Let’s explore this together!
We have identified several areas where automation – or semi-automation – is relevant:
– Stock sizing
– Order Approval
– Scheduling
– Sales forecasts
Stock sizing
Once you’ve decided where in the value chain to keep stock, the sizing of replenishment loops can be largely automated. Beware, however, that the safety stock formulas that have long plagued this field give recommendations that are too high, requiring human correction. What’s more, these formulas only cover part of the subject, and ignore, for example, order frequencies and minimum orders.
In Intuiflow, we have for years been implementing a wizard that enables mass updating of replenishment loops, based on lead times, variability, and batch sizes. This logic has now been complemented by the patented “autopilot” wizard, which uses artificial intelligence to enable finer tuning.
This logic, coupled with dynamic buffers and intelligent adjustments, enables almost complete automation of stock and replenishment loop sizing – with planners focusing on the few exceptions and subsequent model monitoring.
Order Approval
As soon as the replenishment loops are continuously adjusted, it is possible to automatically approve the order generation process, possibly by applying business rules.
Our customers have achieved automation rates of up to 100% in distribution and procurement, and up to 70% in production planning. It remains for planners to really focus on the situations where they bring value.
By identifying the key capacity constraints, grouping rules, priority rules, and safety margins to be incorporated into the plan, we can automatically generate relevant scheduling and realistic promise dates. But be careful! There’s no such thing as a perfect plan. You can draw up a realistic plan, but execution will undoubtedly differ.
This allows the scheduler to intervene only in specific cases requiring additional business intelligence or business trade-offs.
Sales forecasts
Our approach to the sales forecasting process is to let the algorithms select the best forecasting tactics. Forecasters only act to enrich the model according to forecasted events. We integrate the measurement of the added value of these enrichments (FVA) to help limit these manual adjustments to situations in which they bring an improvement to the signal.
Dare to automate!
Automation in supply chain management is not only possible, it’s necessary. On the one hand, it enables productivity gains, but above all, it allows us to focus our attention on the points that truly require human arbitration.
However, there are some prerequisites:
– Disciplined management of technical and transactional data
– Modeling the management model and defining business rules – this is often diffuse, tribal knowledge, which needs to be transcribed into a formal, perennial model.
