You get out what you put in

IN MY 40-year career in logistics, I’ve seen a new, transformative technological event come along roughly every half-decade. From huge, standalone computers with flickering green text on black screens, to primitive networking which then linked these computers together, to the mobile data revolution, the latest of these is undoubtedly artificial intelligence (AI) and machine learning (ML).

Adoption of and reference to AI technology – not just in logistics – has of course increased dramatically in recent years and it has developed a reputation as something of a panacea that cures all ills, regardless of sector. While it’s certainly true in logistics that AI can potentially solve many different problems, there are also considerations which must be taken into account to get the best from a proposed solution.

AI isn’t something to have simply for the sake of having it. Just because you can have something does not mean that you should. Any solution needs to have operational value that reflects what people are doing day-to-day, makes tasks quicker and more accurate and optimises fleets by reducing costs and emissions. Simply saying “We need AI” without an idea of what form this might take will not work.

You get out what you put in

Technology providers cognisant of the AI revolution will have core data science teams targeting its use within their product portfolio and customer solutions, which increases both productivity and operational performance while also reducing risk.

Principally, AI solutions have a dependence on high-quality data. When using a solution in conjunction with a Transport Management System (TMS) that has the potential to optimise route-planning and operational efficiency, feeding poor and incomplete data into the system will result in flawed route and optimisation plans. 

As much useful data as possible is therefore vital. Almost 600,000 Microlise assets have drawn around 1.2 petabytes of text-based data (with a petabyte being the equivalent of 1,024 terabytes) from usage patterns. A  typical 24-hour period might be informed by as many as 112 million events, including 12.7 million of those from tachographs.

It is this high-quality, real-time data that AI and subsequently ML uses to determine optimised routes, predict delays accurately and improve overall operational efficiency. Without it, transport managers would be faced with what might feel like a retrograde step of combining a limited solution with the capacity of their own knowledge, which has a margin for error. 

However, as anyone who has ever had to send a supplementary text message saying ‘Sorry, predictive text’ knows, technology doesn’t always get things right. There is still a need for human intuition and situational awareness which means that the robots will not take over the world just yet, but it is not a case of transport managers going back to issuing handwritten manifest lists.

Collected, analysed and optimised high-quality data can be overlaid over complex customer requirements to ensure that everyone is as happy with the end result as possible. This could range from imperative requirements like chilled and frozen freight arriving at a supermarket loading bay in the appropriate condition to making sure that the consignment has been executed in a trailer that does not bear the name of a competitor. Every last variable can be taken into account, if the AI has the opportunity to do so. 

The more data you have, the smarter your AI and the more optimised your operations will be. Different levels of value in an AI product give different levels of reward for a logistics firm, its operations and its customers. 

There are however other applications of AI beyond increased route and operational efficiency. It can analyse and correct photographs taken using mobile technology, and in-cab cameras can analyse situations and alert operators to instances in which a lone driver might need help or support. It will also intelligently monitor trailer brake performance and identify accident hotspots using historical and real-time execution data.

Unmet expectations

Although AI has genuine game-changing capabilities for logistics, there are still some things that it can’t quite manage because of distinctly human characteristics.

Systems could dramatically reduce inefficiencies that clog up the motorway network and emit unnecessary carbon dioxide if logistics firms shared real-time data. Such collaboration could prevent empty running and help companies meet government targets on net zero, but it is a concept that remains unfulfilled while logistics firms are reluctant to share their data with competitors.

Also, as intuitive as AI is fast becoming, it still cannot solve fundamental issues such as traffic congestion, poor road conditions and outdated transport networks. It is also restricted by being unable to physically repair vehicles, load and unload cargo or handle unexpected mechanical failures without human interaction.

AI is also susceptible to inheriting biases from historical data which can lead to unfair decision-making in areas such as pricing, assigning drivers to jobs and prioritising deliveries.

Watch this space

While the potential of AI may not ultimately be totally limitless, there is still much to anticipate for its usage in the logistics sector in the coming years.

Coupling AI with business intelligence (BI) technologies and platforms will enable users to ask systems questions in common languages rather than with complicated prompts that do not understand nuance. This could be as simple as asking, for instance, “How much did empty running cost me last month compared to this month?”

Jeopardy models are also on the horizon. These will report on ‘at-risk’ activities that are consuming data from telematics platforms and which then map feeds and historical behaviours to proactively advise of issues which may soon arise before they are able to gain a foothold. As an extension of this, suggestive AI engines will provide the best option to mitigate this risk in a way that is most efficient, both financially and with the minimum of environmental impact.

We’ve come a long way in the last 40 years but it’s safe to say the journey isn’t over yet.

John Taylor, strategic solutions manager, Microlise

For more information, visit www.microlise.com