Page 9 - Logistics News July 2019
P. 9
Thought Leadership
2. Access to community (multi-party) data 5. Decision process must be continuous,
The ability to access data outside of the self-learning and self-monitoring
enterprise or, more importantly, receive Data in a multi-party, real-time network is
permission to see the data that is relevant always changing. Variability and latency are
to your trading community must be made recurring problems, and execution effi ciency
available to any type of AI, Deep Learning or varies constantly. The AI system must be
Machine Learning algorithms. Unless the AI looking at the problem continuously, not just
tool can see the forward-most demand and periodically, and should learn as it goes how
downstream supply, and all relevant constraints to best set its own policies to fi ne-tune its
and capacities in the supply chain, the results abilities. Part of the learning process is to
will be no better than that of a traditional measure the eff ectiveness ‘analytics’, then
planning system. Unfortunately, this lack of apply what it has learned.
visibility and access to real-time, community
data is the norm in over 99 percent of all 6. AI engines must be autonomous
supply chains. Needless to say, this must decision-making engines
change for an AI tool to be successful. Signifi cant value can only be achieved if
the algorithm can not only make intelligent
3. Support for network-wide objective decisions, but can also execute them.
functions Furthermore, they need to execute not just
The objective function, or primary goal, of within the enterprise, but, where appropriate,
the AI engine must be consumer service level across trading partners. This requires your AI
at lowest possible cost. This is because the system and the underlying execution system to
end consumer is the only consumer of true support multi-party execution workfl ows.
fi nished goods products. If we ignore this fact,
trading partners will not get the full value 7. AI engines must be highly scalable
that comes from optimising service levels and For the supply chain to be optimised across
cost to serve, which is obviously important as an entire networked community of consumers
increased consumer sell-through drives value to suppliers, the system must be able to
for everyone. process huge volumes of data very quickly.
A further enrichment of the decision Large community supply chains can have
algorithm should support enterprise level millions, if not hundreds of millions, of
cross-customer allocation to address product stocking locations. AI solutions must be able to
scarcity issues and individual enterprise make smart decisions, fast, and on a massive
business policies. Thus, AI solutions must scale.
support global consumer-driven objectives
even when faced with constraints within the 8. Must have a way for users to engage
supply chain. with the system
AI should not operate in a ‘black box’. The
4. Decision process must be incremental user interface (UI) must give users visibility
and consider the cost of change to decision criteria, propagation impact and
Replanning and changing execution plans enable them to understand issues that the AI
across a networked community in real time can system cannot solve. The users, regardless of
create nervousness in the community. Constant type, must to be able to monitor and provide
change without weighing the cost of the additional input to override AI decisions when
change creates more costs than savings and necessary. However, the AI system must drive
reduces the ability to eff ectively execute. An AI the system itself and only engage the user on
tool must consider trade-off s in terms of cost an exception basis, or allow the user to add
of change against incremental benefi ts when new information the AI may not know at the
making decisions. request of the user. •
July 2019 | Logistics News 7