During my current supply chain planning market research, I have received briefings from several SCP companies. All are investing in artificial intelligence. Many say that they are using generative AI, a type of AI that can create new content and ideas, as part of their journey toward autonomous planning. Autonomous planning is a type of planning that takes humans out of the planning loop.

But Omer Bakkalbasi, the chief innovation officer at Solvoyo, says they are already there. “We feel we are ahead of our times. The people who work with us are those who really, truly believe in what we believed in from the start, that is, autonomous supply chains are possible. That was our vision, even starting back in 2010. Our clients have realized that dream and are benefiting from it.” They are creating new efficiencies for these customers because they are creating plans that can be executed a high percentage of the time.

Solvoyo has a metric they call the user acceptance rate. This metric measures the percentage of time the planners accept replenishment, transportation, or inventory plans as they are without any change in the timing of the delivery or the quantity to be delivered. Having a 95% user acceptance rate would mean that 95% of the planning recommendations are executed as is.

At one of their customers, A101, Solvoyo says they have reached a 99.5% planner acceptance rate. I talked to A101 in 2021, and the Turkish convenience store retailer had achieved 99% autonomous planning for all products not subject to spoilage.

At a division of one of the world’s largest consumer goods companies, 85% autonomy on manufacturing plans and 95% acceptance of proposed purchase orders has been achieved. A division of Unilever is also on an autonomous planning journey with Solvoyo.

But when he presents this to many companies, they don’t believe it. “I presented this at Home Depot, and an executive looked me right in the eye and said that’s too good to be true. It’s not possible. Nobody will give that much control of their plan to an algorithm.”

“But we’re not an algorithm,” Mr. Bakkalbasi states firmly. “We are a platform. “The platform collects data and makes sure the master data is internally consistent. If a user makes changes to the plan, they log that data. Planners that do not accept the plan presented to them need to select a reason code for why the plan was not good enough. This allows the system to learn and improves the quality of the engine’s output. It is a “continuous feedback loop.” Getting to over 95% plan acceptance rates does involve a lot of sweat and blood” from both their customers and a dedicated analyst from Solvoyo.

However, the improvement process starts with the reason codes. Once planners must select a reason code, acceptance rates improve significantly.

Further, the journey to autonomous planning does not rely on a highly accurate forecast. “I have not cared for 20 years”, Mr. Bakkalbasi states with force, what level of forecast accuracy is achieved. “Forecasting is not an actionable item.” You don’t act on a forecast; you act on what you purchase. “That’s an action. You manufacture stuff. That’s an action. You set a target inventory level. Accepting that as a parameter, that’s an action. You route a truck. That’s an action.”

The solution is creating an inventory target by SKU by channel that is automatically updated based on changes in historical sales, weather, and other external demand drivers. When forecasting accuracy improves, achieving on-time in full can be accomplished with less inventory. “That is why I don’t care about forecast accuracy.”

The journey to achieve a high degree of autonomous planning starts with becoming digital, then becoming intelligent, and finally achieving autonomy. “And you can’t skip a stage.”

“What does ‘digital” mean? Mr. Bakkalbasi asks rhetorically. “You have to have a digital platform where you get all your relevant data.” And that data has “to be internally consistent. The number one requirement for autonomous planning is master data management. If you don’t have your master data correct,” you can’t possibly succeed on this journey.

Longer terms plans are being created, the sales and operations plans, but plans will also change as circumstances change. If the amount of material a supplier can provide changes, the platform must know that immediately. For a fulfillment plan, the dimensions of the pallets and the trucks, the number of trucks available, and the capacity of a warehouse must be current and accurate.

Second, plans must be intelligent. The algorithms to be used and customer priority rules must be established. You can’t constantly swap out algorithms, change customer allocation logic, and achieve autonomous planning.

Intelligence is also related to how the solution is being used. The platform not only tracks plan acceptance, it tracks how often the different pages are used. “What does that mean? We have lots of functions, lots of analytics, lots of reports.” The system tracks who uses each of these pages and how long they use them for. What is the point of producing a report that nobody uses? “That kind of accountability is required,” Mr. Bakkalbasi explains, “before you focus on automation.”

Mr. Bakkalbasi closed by admitting that many customers see autonomous planning as risky. He admitted that many of their customers are only in the beginning stages of their autonomous planning journey. He is particularly frustrated that they have proved that autonomous planning provides strong results in one division of a global company but then not being able to get other divisions to pay attention to what has been achieved.

The post Autonomous Business Planning Is Not Only Possible, It Has Already Been Achieved appeared first on Logistics Viewpoints.