New world, old tools

www.airlinecargomanagement.com
March 2014 Volume 13, Issue 1
New world,
old tools
It’s time to
turn to tech
Our results are in...
The state of freight revealed
Trendsetters
Air cargo’s in fashion
Arabian heights
Saudia gets talking
Technology: Artificial intelligence
 From film to freight
Hollywood may have a point, says Niall van de Wouw, managing director of technology
company Clive. He argues that ‘artificial agents’ are the future, even for air cargo, and are
something that will help boost carriers’ bottom lines
A
rtificial intelligence is more often
associated with Hollywood movies
than with real life. This perspective
is, however, changing. John McCarthy, who
coined the term, defined AI as “the science
and engineering of making intelligent
computer programs.” In other words,
technology which thinks and acts like a
human. While Apple and Google are making
large investments in this domain to make
life easier for their consumer clients, artificial
intelligence is also slowly finding its way to
the B2B environment too. Sooner than you
might think, this technology will support
cargo airlines with managing day-to-day
challenges such as service recovery,
opportunity cost and client performance.
Recovering from operational issues
such as delays and cancellations is often a
time-consuming and complex challenge.
Getting back on track is not easy, while
the chance of causing a string of knock-on
effects is high. Meanwhile, the growing
number of multi-hub networks makes
it increasingly complicated to keep an
Technology which thinks and acts like humans sounds
far-fetched, but it’s not limited to Hollywood and can
even help air cargo (photo: Traceyp3031)
50 Airline Cargo Management
overview of the available recovery options.
With every minute of inactivity after the
issue becomes known, the chances of being
able to use the fastest and/or cheapest
solution dwindles. Imagine if each shipment
was given a ‘voice’ that would speak up the
moment an issue becomes known. That is
exactly what an ‘artificial agent’ representing
an individual shipment would do – it
is software that continuously monitors
progress according to a plan, but the
concept does not simply stop there.
These artificial agents would not just
have voices, but also beliefs, desires and
objectives. Their beliefs relate to how they
see themselves in relation to other agents.
For example, it would understand the
relative importance of priority shipments
versus standby shipments. The desires
contain the goals of the agent – should the
shipment get to the final destination in the
fastest possible manner or at the lowest
possible cost? And finally, the objectives are
the actions the artificial agent is planning to
execute once it has combined the received
information with these beliefs and desires.
Thus, once an operational issue becomes
known, all affected artificial agents will start
cooperating and negotiating with each
other on a recovery plan that adheres best
to their beliefs, desires and intentions.
A recent project with artificial agents
in the rail industry proved that the agents
could efficiently provide realistic plans in
response to complex issues. The focus of
this particular project was to solve the issue
of freight trains competing for space on a
one-track railway line. Taking into account
the continuously changing dynamics,
the agents were capable of delivering an
efficient plan within a matter of minutes.
This kind of turnaround time is far quicker
than any conventional optimisation
approach could deliver.
Controlling opportunity cost is another
challenge where artificial intelligence can
help. Finding alternative freight volumes
for capacity that became idle due to
an operational or commercial issue, for
example no-show or under-tendering,
would also require swift action. The type
of artificial agent that would help in these
situations would look after the interests of
an individual flight, maximising its weight
utilisation, for example. It continuously
monitors the developments of shipments
planning to be moved on its flight and
will immediately spot incidents that could
www.airlinecargomanagement.com − March 2014
Technology: Artificial intelligence
van de Wouw: new technology is bringing opportunities
for the air freight industry (photo: Clive)
result in (more) capacity becoming idle,
such as feeder delays, under-tendering and
no-shows. Artificial agents are proactive:
they can anticipate issues and avoid
potential problems.
An excellent example of this would be
having a couple of hours more time to
respond to a client under-tendering. This
could make all the difference in successfully
attracting additional volumes from the
market and/or bringing forward shipments
that were planned for later flights.
Monitoring client performance is another
challenge that cargo airlines frequently
face. Incentive deals and upfront volume
discounts are just two examples of
agreements that require close attention
to ensure they provide the best possible
value for the airline. But monitoring actual
client performance, whether their tendering
behavior is in line with the desired behavior,
can be quite cumbersome. With tightening
budgets, there are now less ‘arms and legs’
available to conduct such analyses. This
measurement of client performance can, for
a large part, be taken over by another type of
artificial agent. This type could be considered
the custodian of the deal, as such it would
look for developments that are not in line
with the intended behavior of the client (for
example reaching a target revenue). This
constant monitoring of client performance
will help the airline to respond quickly to
unforeseen developments, getting the best
possible bang for its buck from its incentive
and discount programmes.
March 2014 - www.airlinecargomanagement.com
Regardless of how ‘intelligent’ the artificial
agents really are, they will not always be
capable of grasping the full context of
the challenge at hand. That is why the
approach of ‘adaptive autonomy’ has
been developed. This approach allows the
agents to identify when challenges can be
resolved automatically, when they should
be referred to a user with a set of options
for resolving and when challenges should
be fully handed over to the application user.
This type of adaptive autonomy is especially
relevant to the first two types of artificial
agents that have been discussed, as both
can initiate an action. But even the third
type (the custodian of client deals) will get
smarter over time. Based on the interaction
with the application user it will gradually
‘learn’ what kind of actions and notifications
trends were useful to that user and which
were not.
Integral to the success of these artificial
agents is their access to global airline data
on, among other things, booked shipments,
delivered cargo, manifested freight and
departed flights. Without this information
there is very little room to identify issues
and coordinate/negotiate with other agents
on the possible recovery plans. But where
can this information be found? Data in
the airline’s cargo booking system (or a
live copy of it), combined with messages
that are being sent between cargo airlines,
forwarders and handling companies
provides a solid basis to build from. If
artificial agents can furthermore ‘eavesdrop’
on these messages and simultaneously have
access to the plans in the booking system,
they will be able to spot issues that need
immediate action, instantaneously.
These artificial agents would, of course,
have to delve into large sets of data.
And although it would be fashionable to
label this ‘big data’ (that is “high-volume,
high-velocity and high-variety information
assets that demand cost-effective,
innovative forms of information
processing...”, according to Gartner, an
information technology research and
advisory firm), in strict terms it is not, for
two reasons: Firstly, although the volumes
and velocity of the shipment and flight data
of traditional cargo airlines is substantial,
it is at a far lower level than what some
retail companies are currently ‘digesting’.
Macy’s, a US department store, adjusts
pricing in near-real-time for 73 million
items based on demand and inventory (see
Nicole Laskowski; Ten big data case studies
in a nutshell); and secondly, the variety in
the data that the artificial agents need to
process is not of the extent that it requires
advanced technologies to interpret it
correctly. IATA standards (the message type
and content) are being used globally by
airlines, forwarders and handling companies
alike when communicating with each other.
Automatically interpreting such messages
is not that complicated. The relatively low
volume, velocity, and variety means that the
issues at hand are very solvable. This is good
news for cargo airlines as it will limit the
investment needed for artificial agents to be
able to tap in to this wealth of information.
Artificial intelligence is going to support
cargo airlines in overcoming day-to-day
challenges in the not too distant future.
Recovering from service disruptions,
controlling opportunity cost, and actively
monitoring client performance are just three
examples of where this technology will
help. Undoubtedly, more opportunities will
be discovered. The use of artificial agents
could well be the pinnacle of humanised
technology; software that is intuitive
to operate and which provides instant
gratification. They overcome challenges
even before you knew you had them. But
that, admittedly, sounds a bit like a trailer for
a Hollywood action movie.
A recent project in rail proved that artificial agents
could provide realistic plans in response to complex
issues (photo: Steve Damron)
Airline Cargo Management 51