MFG Labs was invited on February, 6th to a conference organized by DCBrain, on the use of AI applied to logistics, with Jean-David Elbim, Chief Innovation & Data Officer at Bolloré Transport & Logistics.
The opportunity for MFG Labs and Paul-Mehdy M’Rabet, senior data consultant, to talk about the issues, opportunities and challenges of data applied to the supply chain and logistics activities.
Here are some selected excerpts from this conference.
A change in the business model that creates new needs
We note that supply chain and logistics companies tend to have an ultra-centralized decision-making system and deployed agents, responsible for operations, are considered - whether assumed or not - as performers:
- Either the decisions are taken by a central body - far from the field - as if they were strategic decisions;
- Either the operational decisions are taken by the branches according to rules enacted “centrally”.
This operating model, too slow and often uncorrelated with realities happening on the ground, is not suitable and many companies decide to give back the decision-making ability to their agents deployed in branches.
So the decision process changes. From decisions made by senior executives, over time, at low frequency, sometimes with the help of consulting companies, we move to decisions made by operational agents, in the field, alone, every day, several times a day.
To properly perform this task, operators expect that they will be provided with simple tools adapted to their work context, quality data to rely on, and appropriate training to enter into this paradigm shift. Simply put, they want the company to empower them to become “data-driven”.
Being “data-driven” in the supply chain and logistics
For MFG Labs, being data-driven means systematically using data in the decision-making process. In other words, it is to guarantee by software solutions that all decisions are made in the light of the data present in the information system of the company.
Still, using data to make a decision is not a necessary and sufficient condition to be qualified as “data-driven”:
- Data access must be guaranteed by systems. Too often it is based on an intermediated human process (eg creation of a ticket to the BI team);
- The raw data retrieved must be processed - again automatically - to facilitate interpretation.
Put it simple, we will understand that the data presented to the decision-maker must be of quality (veracity, freshness, completeness …) so that his decision is optimal.
For supply chain and logistics companies, this is a major challenge since - in essence - they handle many complex objects at the interface between the physical world (goods, warehouses, means of transport …) and digital (documentary flows, financial flows, etc.), and must adapt to the variable quality of the data coming from their ecosystem (upstream or downstream businesses).
Operational decision-makers: key players in the path to become “data-driven”
To become data-driven, as we said, you must systematically use data in your decision-making.
If no system guarantees the use of the data for an identified decision-making, we propose the following shift and actions:
- identify useful data to enable the identified decision making;
- define in what form these data are most useful / understandable for the decision maker;
- design and implement a software solution adapted to the context of the decision-maker, which presents - at the right time - the identified data (step 1) in the defined form (step 2).
Such a change cannot be made without the involvement of operational decision-makers. At MFG Labs, we work apply service design methodologies supported by an “Experience” pole (UX / UI designers, Industrial designers, Service designers, etc.) in a co-construction mode that includes operational agents early in the design process. It should also be noted that the fact of involving future users from the design phase is a guarantee of adoption of the future solution.
A simple operational model that facilitates high value-added AI projects
Thus, having and maintaining quality data, designing and implementing software tools exploiting this data, and involving operational staff in the design can help improve decision-making processes.
In addition, we quickly notice that this allows a virtuous circle to be set in motion for the company:
- The agents who now use the data on a daily basis can quickly notice defects / deviations in quality, and have them corrected (via the implementation of a feedback loop in the software solution for example). The data asset is better monitored, and quality faults are more easily corrected, thereby increasing the total value of this asset.
- By recording and observing the choices made by the users and the data consulted to make these, the company can better understand the decision-making mechanisms. This makes it possible to identify the most influential and less influential parameters, to observe trends and seasonality and, above all, to create a set of labeled data which may be useful if we wish to move towards a greater automation of the decision making.
When this virtuous circle does work, the company acquires a better knowledge of the underlyings elements needed to make the decision, and then has a qualified and exploitable data asset.
This asset is a foundation on which to capitalize, and allows a company to go further in improving any decision-making. The company has now the opportunity to revisit the mandate given to the operator and that given to the tool. The addition of artificial intelligence services could help to determine whether the tool should be used just to present figures, to predict a phenomenon or a state, to prescribe the solution most suited to the context, to automate all or part decisions…
Note that the ambition of a company is not systematically to reach automation (i.e. the total delegation of the decision to the system). Technically it is very often difficult to set up. Often besides, beyond ethical considerations, the ROI is negligible, the cost of implementing a fully automated system being high or even technically infeasible.
It is therefore important to properly determine the mandate to be given to the system. The delegation of a part of the decision-making process must in any case enable the right decisions, “augmented” decisions, to be taken in a systematic manner.
Benefiting from the approach, delegating part of the process to a machine can allow the employee to devote the time saved to more important or complex tasks. For example, in freight forwarding, we managed to drastically reduce the time spent on transport planning so that employees could focus on managing transport exceptions. Better exception handling allows the company to reduce the number of disputes and significantly increase customer satisfaction.
Soft-information: the next obsession?
Whether the obsession with capturing a maximum of relevant data to improve systems and therefore decision-making seems necessary, we note that certain piece of information is difficult to capture, to structure and therefore to process. However, this soft information that circulates within the company is often key in decision-making.
Strikes, blocked roads, storms that nail planes to the ground, epidemics that impose the closure of certain activities in certain countries… All this information which is exchanged within the company, by email, telephone, fax, on the intranet etc. . often do not appear in any system and therefore cannot be taken into account by the artificial intelligence services to which part of the decision is delegated.
By structuring this information, a company enriches its data asset. Certain algorithms can then be revised to consider this new data and make the tools developed more efficient. Other algorithmic models, whose study has been refused because it is impossible to carry out, become credible alternatives, because of this new data, and their development could be considered.
Call to action
The road ahead for logistics companies to make their data-driven operations is not unique. Continuous and concomitant improvement of systems and data is however a necessary posture and necessarily creates long-term value.
For more information, contact us at: email@example.com