What
did
you
do
during
the
data
wars,
kid?
That’s
more
or
less
how
someone
in
the
far-flung
future
will
talk
about
legal
technology
in
2025.
Or
we’ll
all
be
driven
extinct
by
SkyNet.
One
or
the
other.
As
generative
AI
continues
its
march
into
the
legal
workflow,
the
flurry
of
press
releases
increasingly
focus
on
data
access.
At
the
risk
of
oversimplifying
the
market,
advancements
in
the
underlying
algorithms
are
going
to
stagnate
and
providers
have
already
announced
their
investment
in
building
better
guardrails
so
the
next
logical
leap
for
the
technology
is
pairing
it
with
better
data.
And
so
that’s
the
current
scramble.
Harvey
is
an
AI
platform
designed
for
legal
professionals,
LexisNexis
is
a
vast
repository
of
data
and
years
of
categorization
and
curation…
they
fight
crime!
In
all
seriousness,
a
while
back,
Harvey
wanted
to
build
a
legal
research
solution
and
LexisNexis
worked
on
an
AI
solution.
Now
they’re
joining
forces.
Within
Harvey,
customers
can
ask
LexisNexis
Protégé™
to
receive
comprehensive,
trusted
AI
answers
grounded
in
the
LexisNexis
collection
of
U.S.
case
law
and
statutes,
validated
through
Shepard’s®
Citations.
Harvey
users
can
ask
complex
legal
questions
in
natural
language
and
receive
citation-supported
answers
from
primary
sources
of
law,
refine
their
queries
through
follow-up
questions,
and
seamlessly
continue
their
research.
Answers
are
generated
using
LexisNexis
fine-tuned
models
within
a
proprietary
infrastructure
that
anchors
responses
in
legal
content,
metadata,
and
case
law
relationships,
powered
by
Shepard’s®
Knowledge
Graph
and
Point
of
Law
Graph
technology.
Now
Harvey
users
have
the
data
and
LexisNexis
can
deliver
its
product
through
a
platform
delivering
AI
to
lawyers
across
use
cases.
Why
isn’t
this
just
an
acquisition
like
when
Thomson
Reuters
bought
Casetext?
Unclear.
The
many
tentacled
LexisNexis
corporate
octopus
has
money
in
Harvey
so
it
wouldn’t
be
crazy
to
bring
it
all
under
one
roof.
And
despite
the
alliance
talk,
it
seems
as
though
this
will
eventually
create
redundancies.
Artificial
Lawyer
asked
that
question
outright
and
was
told
by
LexisNexis
North
America,
UK,
and
Ireland
CEO
Sean
Fitzpatrick,
“Who
knows
where
this
will
go?”
Pretty
sure
the
answer
is
some
kind
of
merger.
The
deal
also
includes
a
tease
for
some
workflows
getting
the
robo-treatment:
Harvey
and
LexisNexis
will
also
develop
sophisticated
legal
workflows
built
on
the
latest
generative
AI
technology.
These
co-developed
workflows
will
initially
include:
Motion
to
Dismiss
Workflow:
Generates
high-quality
Motion
to
Dismiss
arguments
and
related
client
communications
with
legal
research
content
from
LexisNexis
Motion
for
Summary
Judgment
Workflow:
Automates
key
steps
in
drafting
a
Motion
for
Summary
Judgment
with
supporting
legal
research
content
from
LexisNexis
And
oh
yes,
motions!
Because
this
brave
new
world
includes
—
wait
for
it
—
AI-generated
“Motion
to
Dismiss”
and
“Motion
for
Summary
Judgment”
workflows.
So
the
two
most
overused
filings
in
the
history
of
American
jurisprudence
are
now
getting
the
robo-treatment.
I
remain
worried
about
of
the
automated
workflow
as
a
concept.
Crunching
a
file
into
a
usable
statement
of
facts
can
save
time
and
generative
AI
can
also
perform
the
top-level
reasoning
to
draft
research
queries
based
on
the
case
materials.
But
while
AI
can
be
a
force
multiplier
along
every
step
of
the
way,
putting
together
a
brief
seems
like
more
than
the
sum
of
its
parts.
Legal
research
is
a
morass
of
linguistic
subtlety,
precedent,
and
jurisdictional
nuance.
Even
with
Shepard’s
duct-taped
to
its
forehead,
an
LLM
is
one
semantic
landmine
away
from
confidently
pulling
the
good
case
while
overlooking
the
great
one.
As
an
unrelated-but-maybe-related
aside,
in
Bob
Ambrogi’s
LawSites
coverage
of
this
announcement,
Harvey
CEO
Winston
Weinberg
said,
“Lawyers
have
trusted
LexisNexis
for
centuries,
so
this
alliance
allows
us
to
provide
our
customers
with
data
sources
they
know
and
rely
upon
while
collaborating
with
on
AI
systems
that
make
daily
life
for
our
joint
customers
significantly
easier.”
LexisNexis
was
founded
in
1970.
When
he
says
“for
centuries,”
does
he
mean
the
20th
and
the
21st?
Because,
one,
I’m
not
cool
with
that.
And
two,
it
feels
like
the
sort
of
“technically
true
but
wrong
because
it
misses
the
nuance”
response
that
worries
me
about
the
whole
generative
AI
workflow
experience.
The
more
providers
try
to
put
the
entire
lawyering
process
on
greased
rails,
the
less
I
trust
the
humans
to
perform
the
necessary
work
to
make
sure
the
work
product
is
the
best
it
can
be.
It’s
one
thing
to
“do
legal
research”
and
evaluate
the
output
and
another
to
“do
all
the
steps
at
once”
and
evaluate
that
output.
It’s
hard
to
describe,
but
I
approached
research
differently
in
the
advocate
role
than
in
the
arbitrator/mediator
role.
When
I
would
receive
briefs
from
the
parties
and
check
up
on
them,
my
research
was
always
already
contextualized
by
their
finished
work
product.
I
had
to
affirmatively
divorce
myself
from
that
context
to
guarantee
I
got
to
the
right
answer.
Obviously
this
product
is
still
in
development
and
might
take
pains
to
prevent
people
from
doing
this,
but
the
hype
around
these
sorts
of
workflow
agents
creeps
toward
“it
will
do
all
the
steps
to
give
you
something
finished
that
you
can
just
check
up
on,”
but
when
the
output
arrives
looking
finished,
there’s
a
bit
of
a
psychological
barrier
to
ripping
it
back
down
to
the
studs
to
make
sure
it’s
right.
Good
lawyers
will
change
how
they
approach
the
job
to
make
sure
the
allure
of
an
automated
workflow
doesn’t
dull
their
professional
judgment.
I’m
just
not
sure
I
trust
lawyers
enough.