![PITNEY BOWES PACKAGE PARCEL BLOOMBERG.jpg PITNEY BOWES PACKAGE PARCEL BLOOMBERG.jpg](https://www.supplychainbrain.com/ext/resources/2024/08/10/PITNEY-BOWES-PACKAGE-PARCEL-BLOOMBERG.jpg?height=100&t=1723435795&width=150)
The act of divining the future is nothing more than a consideration of possible scenarios, and their impact on a business. For most of human history, the success of that exercise has rested on the appointed seer’s years of experience, coupled with some indefinable grasp of probability.
In recent decades, machines have stepped up to aid humans in their efforts. Software programs have gradually matured to the point where they can finally be labeled as “artificial” intelligence, aping the thought processors of their creators in the form of “expert systems.” But they haven’t been able to wipe out the plague of uncertainty.
They never will. There are simply too many variables, and too much chaos, ever to accurately predict the future of a single business, let alone an entire economy. Still, the coming of A.I. has helped humans to get a better grasp of the possibilities, as they sift through the mass of information called “big data.”
Take the relatively simple task of determining the true cost of a piece of apparel or other manufactured retail item. Of course, it’s not really simple at all. In addition to obvious elements such as the price of materials and labor, the correct decision must take into account such factors as weather, tariffs and border taxes, political and economic instability in source countries, and the impact of travel restrictions and visa charges on global talent.
Rolling all of those uncertainties into a viable prediction would seem to require a combination of science and art. But Sue Welch, chief executive officer of retail-management software vendor Bamboo Rose, believes machine learning has progressed to the point where human involvement in the process can be minimized.
A.I. can help businesses to address all of the components that go into costing, from regulatory trends to packaging. A single product can have 100 different costs based on size, color and targeted markets. It’s essential that manufacturers be able to analyze cost at extremely low levels of granularity, Welch notes.
There’s always the danger of being inundated with too many possible scenarios, most of which aren’t relevant. So a workable automated system will have the capability to weed out those parameters that don’t make sense — for example, a material that is prohibited in key markets. (Suppliers with poor quality or delivery records will also be ejected from the mix.) In the end, a proper filtering of scenarios might only present planners with the top 10 options for any given combination of product characteristics.
Adding to the complexity of pricing determinations is the ever-changing nature of the many factors that feed them. Interest rates, for example, can have a big impact on a supplier’s ability to deliver in a cost-effective manner. When they suddenly begin to rise — as they have in the first quarter of this year — the old calculations go out the window. It’s vital that any automated pricing system be able to adjust immediately.
Then there’s the question of how long to run a given scenario before a decision on costing is made. The system isn’t limited as to time, says Welch, but needs to be able to account for anomalies such as the difference between a random 12-month period and a calendar year. Regulations and customs duties often change with the new year, so that information has to be incorporated into a long-range scenario.
Creating a long horizon is important because the buyer need to have an accurate sense of a supplier’s total capacity for producing over time. But the system also has to be constantly adjusting its conclusions as the supplier nears its delivery date.
The output of an automated costing system is never going to achieve a perfect match with reality. Welch says the system must be constantly comparing estimated with actual landed cost, and making adjustments accordingly.
So perfection isn’t achievable — but can the system get better? Welch believes it can. She sees new developments in 3-D modeling and augmented reality as leading to even more accurate determinations of optimal costing. The technology allows groups dispersed throughout the world to view the same items in terms of elements such as materials, machining and workmanship. “It’s going to have a big impact,” she says.
The goal is to further remove humans from the often-numbing process of product costing, freeing them up to focus on design and development. Still, there will always be a degree to which neither human nor machine can select the ideal option from among so many choices. As Welch puts it: “Life happens.”