Page 34 - gs261002
P. 34
Insights and Expertise
AI won't fix a analysis can suffer. And if the underlying workflow is
inefficient, AI may simply make that inefficient workflow
broken payments faster.
The data problem behind the AI problem
back office Payment organizations generate enormous amounts of
data. Every transaction can produce information related to
authorization, reconciliation, adjustments, fees, settlement,
disputes and more. The challenge isn't necessarily a lack of
data. It is often where that data lives and how easily it can
be connected. In a fragmented environment, transaction
information may reside in multiple applications, databases
and operational systems. Different systems may have
different data structures, identifiers and processing
schedules. An employee investigating a transaction
may have to move between several applications just to
understand what happened.
Now consider asking AI to perform that same investigation.
AI doesn't eliminate the underlying fragmentation. It has
to navigate it. A modern AI strategy therefore begins with
a fundamental question: Can your organization provide
AI with the right information in the right context?
By Michael Meeks You can't automate what you can't see
BHMI Visibility is one of the fundamental requirements of
effective payment operations. Organizations need to
rtificial intelligence is rapidly changing the understand what is happening across their transaction
payments industry. AI is being used to detect environment, including:
fraud, automate customer interactions, analyze
A transaction patterns, assist with investigations • Which transactions have processed?
and identify opportunities for greater operational effi- • Which transactions haven't reconciled?
ciency. As organizations look for ways to reduce costs
and manage increasingly complex payment activity, the • Which exceptions require attention?
potential of AI is difficult to ignore. But there is an impor- • Where are financial differences occurring?
tant question payment organizations should ask before • What fees were assessed?
putting AI to work in their back office: Is the underlying
operation ready for AI? AI can make a good operation • Which transactions have settled?
better. It can automate a bad process. But it can't funda- • Which transactions are associated with a dispute?
mentally fix a broken foundation.
The AI opportunity is real If answering these questions requires assembling
information from multiple disconnected systems, there is
There is enormous potential for AI in payment operations. a fundamental operational problem to address before AI
Imagine a back office where AI can help identify unusual enters the picture. AI may eventually help employees find
transaction patterns, prioritize reconciliation exceptions, the answer, but a better-designed back office can make
assist employees in researching transactions, identify the answer accessible in the first place. The goal shouldn't
potential fee discrepancies or help determine why a be to use AI to compensate for poor visibility. The goal
settlement position doesn't balance. These applications should be to create the visibility that allows AI to deliver
could save significant time and allow payment even greater value.
professionals to focus on higher-value work. But each AI needs transaction context
of them depends on something that is often overlooked:
access to reliable, connected and meaningful data. Payment transactions don't exist in isolation. A transaction
that begins with an authorization can continue through a
AI is only as useful as the information it can access and complex lifecycle involving reconciliation, adjustments,
the processes surrounding it. If the data needed to answer fees, settlement and potentially a dispute. Understanding
a question is scattered across multiple systems, AI has a one part of that lifecycle often requires understanding the
more difficult problem to solve. If transaction information others.
is incomplete or inconsistent, the quality of the resulting
34

