Artificial intelligence, or machine learning,
is not as far off in the future as you might imagine.
The cloud is making it easier for insurers to experiment
with new ways of operating.
By Robert Regis Hyle
The headline for any story on ma- chine intelligence (or artificial intel- ligence, if you prefer) is that it’s the next platform for insurers, according
to Mike Fitzgerald, senior analyst with Celent
and co-author of the Celent report (along
with Craig Beattie), “Machine Intelligence in
Insurance: Designing the Aware Machine.”
“Not so long from now we won’t be talking
about whether your system runs on the cloud,
or if it’s Java or database-agnostic,” says Fitzger-
ald. “We are going to be talking about the use
of artificial intelligence (AI). The subtext of that
is what deep-learning algorithms are.”
Fitzgerald realizes this might be more of
an interesting topic for techies and academ-
ics, but that means nothing unless insurance
professionals learn what AI can and can’t do
and how it works.
“The only way to do that is by doing,” he
says. “The emphasis needs to be to step away
and ask, ‘What would you do with a platform
that could essentially program itself over
time?’ That’s why we need to call out for use
Fitzgerald explains that the service indus-
try has the biggest jump on everyone in terms
of using AI. Companies with a huge number
of call-center reps, in particular, are gaining
value as is the healthcare industry through the
use of AI in making a diagnosis.
“I’m sure other things are going on, but
not a lot of people are talking about what they
are experimenting with because so much of
this is proprietary,” says Fitzgerald.
Insurers are working with AI in real-world
situations, though. Fitzgerald points to USAA
as one example. The insurer is using IBM’s
Watson to work with returning veterans.
“They’ve trained the Watson platform to
take phone calls from veterans in situations
where they are returning from service to civilian life and the insurer is providing them with
information they need to consider,” he says.
“USAA trained Watson to understand all
the different programs for these veterans,”
he adds. “They fed Watson data on different
profiles other vets have chosen in the past.
They trained Watson to take these calls so if
you are 30-years-old and been in communica-
tions technology in the Marines and want to
move to a certain location, the platform will
hear the things just as if there is an advisor on
Unlike writing business rules, AI systems
can read and understand text and also read
and understand natural language typed into
the system, according to Fitzgerald. The
insurer can take a PDF or a binder on medical
plans for veterans returning to civilian life
and load it in the system. It has to go through
a human intermediary testing session so the
testers can call into the system, but the system
keep up with changes to the programs and
field the calls on their own.
populating the System
The natural language processing algorithms
are worked on and actually take words and
letters and figure out the relationship of one