.png?width=400&height=62&name=Logotype%20(1).png)
How planning algorithms inch closer to the optimal solution
Professor Hein Fleuren has supported Bottomline in the development of its BX algorithm. We asked him to give us a look behind the scenes.
How AIR is transforming my work as fuel delivery planner
"AI will not replace you at your job but will change the way you work," I read somewhere recently. It’s an interesting statement from my perspective as a fuel delivery planner. To what extent can my work be automated? And is that something to look forward to or not?
I have now worked for two years as a fuel delivery planner at Bottomline, planning deliveries to hundreds of fuel stations, for several customers throughout Europe. My background was in logistics, although planning was new to me. I liked Bottomline as a company, its atmosphere, and the possibility of working from home a lot of the time. But what I have enjoyed most so far is the work itself. Planning, at its core, is about solving puzzles. And each time all the pieces fall into place – in other words, when all transports have been neatly planned in an efficient way and all fuel stations receive their drops on time – you get the same sense of satisfaction that you get from solving a challenging problem.
The question is, though, how long I will be fitting together that puzzle myself. As planning software gets smarter, easier to use, and may even go fully autonomous, it will make planning more efficient. But as it leaves fewer problems for me to solve, will I still enjoy it? And will I even have a job as a planner, a few years from now? These questions came up some time ago when I heard that we would transition to a new application for inventory routing, BX, which was being developed by our own software department at Bottomline.
Intuitive and streamlined fuel delivery planning
News like that will always be met with certain reservations from the people who have to use the software. As a planner, you have an established way of working. The current system may not be perfect, but once you’ve been through the learning curve, you know how it works and how to work around any quirks and problems it may have. Of course, every new software will promise it’s going to make life easier for you. However, I’ve often found it introduces new problems as well. Little things perhaps, that developers may have overlooked or underestimated, but which can be a real nuisance in your day-to-day work.
In this case, however, I had the advantage of being invited to work with the developers, along with some of my colleagues. Which meant that our perspective, insights, and practical concerns as end-users took center stage right from the outset. And as the new platform took shape, of course, I was still using the old program, so I was in a good position to see whether it really would be an improvement.
And let me tell you: it is. For a start, because the new interface is much more visual and intuitive. The new design focuses on the functionality you need 99% of the time and has replaced huge tables and menus full of options you rarely use with a clean, streamlined interface. It has incorporated maps that display routes and the real-time location of trucks. I can adjust a planning by simply dragging color-coded bars across a neat chart that visualizes all routes. To me, it feels as if I’ve moved from Windows 98 to the latest Apple operating system, in one massive step.
The next step: fully autonomous inventory routing
There are other improvements. For example, we can reconcile all completed trips in a few clicks, without having to switch to another program. Yet one of the most significant changes for my everyday work has taken place under the hood, so to speak. The new platform enables real-time planning. It automatically receives live updates of all the data I need. Current stock levels at fuel stations and at depots, the truck’s progress along a given route, how much fuel it is carrying in each compartment, you name it. We’re even working on incorporating real-time traffic data.That means the system makes it easier than ever to adjust the planning to unforeseen developments. As a planner, you always have to deal with some level of uncertainty when making a plan. Yes, we can estimate how stocks will develop, but we can never be sure. With real-time planning, however, that uncertainty is cut to a minimum. The system flags up stations where stocks are running out faster than expected, and the visual interface allows me to change the planning in a few clicks, rerouting trucks if necessary.
In short: a more efficient way of planning is simply not possible. Or is it? Because the next development is waiting in the wings, and it’s the most exciting one yet: autonomous planning. We are currently already testing BX’s ability to use built-in AI to create plans by itself. And the first results are very promising. I find that the system is very good at generating the most efficient planning in the blink of an eye. Within the parameters I’ve set beforehand. After all, each client has their own priorities. And as live updates come in, the planning will adjust itself and seize every opportunity to fine-tune the planning for even more efficiency.
Embracing change: from planner to issue manager
However, that takes me back to the question: what will my work look like once planning goes fully autonomous?So far, I’ve seen nothing to indicate humans will be removed from the equation. Plans may be generated and fine-tuned by algorithms. But as a planner, I’m the one setting the parameters, keeping an eye on the planning, and solving any issues that pop up. And issues will come along. After all, a plan remains a plan. In the real world, all kinds of unforeseen complications may arise. From traffic jams to a broken-down truck, sudden surges in demand, or truck drivers who have to call in sick.Actually, solving such issues is the most interesting part of my job. That’s when I speak to clients, to drivers, and that’s when I’m challenged to come up with creative solutions. And as planning goes autonomous, I expect to spend more of my time on issue management, not less. That’s because with the support of autonomous planning software, I can handle more trucks, more clients. Currently, we expect each planner to be able to handle up to three times as many trucks. In other words: I expect more variation in my day-to-day work.
Oh, and one final thought: as long as we don’t go autonomous, manual planning means that as planners we need to be available for weekend and night shifts, in case some rescheduling is needed. And leaving those routine rescheduling tasks outside business hours to AI is something I’m very much looking forward to!
