It
sometimes
feels
as
though
the
legal
profession’s
primary
engagement
with
AI
so
far
involves
lawyers
citing
fake
cases
generated
by
ChatGPT
and
getting
hauled
before
judges
to
explain
themselves.
And
with
global
legal
hallucination
incidents
closing
in
on
1000
cases,
that’s
an
understandable
reaction.
Everyone
gets
a
good
laugh
and
the
always
vocal
Luddite
contingent
of
the
bar
feels
vindicated.
But
focusing
on
AI’s
limitations
as
a
consistently
reliable
brief
writer
is
like
rejecting
a
grill
because
you
don’t
like
hot
dogs.
Which
isn’t
a
knock
on
what
AI
can
bring
as
a
research
and
drafting
tool
—
or
hot
dogs
for
that
matter
—
but
the
technology
can
do
other
things!
Some
of
which
carry
obvious
value
for
lawyers
and
aren’t
impacted
by
hallucinations
at
all.
No
one
wants
to
turn
over
a
“live
ammunition”
deposition
to
a
neophyte
lawyer.
But
how
does
a
neophyte
learn
to
take
a
deposition
without
experience.
Intensive
practical
simulations
with
copious
levels
of
senior
attorney
feedback
provide
the
best
—
and,
historically,
only
—
training
model.
Maybe
the
firm
puts
it
together
on-site.
Maybe
they
send
attorneys
to
an
off-site
camp
like
NITA.
Meanwhile,
law
students
should
consider
opportunities
like
the
ML
Advocacy
Academy
for
these
experiences.
That
said,
litigation
skills
aren’t
perfected
in
a
week
and
constantly
iterating
training
sessions
with
hired
actors
and
partners
devoting
the
energy
and
(lost
billable)
time
to
mentoring
and
feedback
presents
a
logistical
challenge
for
firms.
One
that
AI
can
now
address.
AltaClaro
has
taken
its
experience
building
training
tools
for
lawyers,
and
the
captioning
and
transcription
talents
of
Verbit.ai
and
launched
DepoSim,
harnessing
AI
to
construct
a
deposition
simulator.
Artificial
intelligence
takes
on
the
roles
of
deponent,
opposing
counsel,
and
court
reporter
in
a
vetted
simulation.
Armed
with
a
closed
universe
of
documentary
evidence,
users
can
ask
questions,
mark
exhibits,
and
navigate
objections
while
the
AI
composes
a
transcript
for
posterity.
When
the
exercise
ends,
the
system
generates
a
detailed,
rubric-based
feedback
report
scoring
the
user
across
multiple
deposition
skills.
If
the
user
wants
to
run
through
the
exercise
again
to
polish
up
their
mistakes,
there’s
no
scheduling
headaches
bringing
court
reporters
and
actors
and
other
lawyers
into
the
office
—
they
can
just
fire
up
the
system
again.
Users
can
even
tweak
the
system
to
alter
the
deponent’s
evasiveness
and
the
defending
attorney’s
obstreperousness
to
practice
different
skills
or
just
to
keep
the
challenge
fresh.
AltaClaro
and
Verbit
gave
me
a
hands-on
demonstration
last
week.
I
haven’t
dealt
with
a
witness
in
years
at
this
point,
but
sitting
down
with
DepoSim
felt
like
a
blast
from
the
past.
With
the
advances
in
technology
over
the
past
couple
years,
the
AI
witness
responds
quickly
and
realistically
and
the
opposing
counsel
threw
in
suitably
annoying
objections.
Anyone
committed
to
their
kneejerk
rejection
of
technology
will
spot
the
cracks
that
separate
the
exercise
from
a
flesh
and
blood
simulation,
but
none
of
these
undermine
its
prowess
at
testing
necessary
skills.
If
anything,
the
way
the
robot
can
stick
to
the
prep
and
never
get
tired
only
increases
the
challenge.
That
can
only
be
a
good
thing.
Sadly,
the
technology
fails
to
capture
that
Texas
deposition
energy.
Maybe
the
next
update.
And
don’t
overlook
the
value
of
a
computer
simulation
for
experienced
attorneys.
A
Biglaw
litigator
can
go
a
fair
amount
of
time
between
taking
depositions.
Despite
a
little
rust
setting
in,
a
senior
partner
isn’t
likely
to
schedule
on-site
training
for
themselves
or
book
a
trip
to
an
organization’s
boot
camp.
This
tool
gives
them
a
low-impact
option
to
brush
up
on
their
skills
before
the
real
thing.
From
my
experience,
I
got
dinged
on
professionalism
because
I
couldn’t
stop
snarking
at
the
AI
witness:
Joe:
Is
this
your
signature
on
the
document?
Witness
AI:
I
will
have
to
look
through
the
document
to
find
my
signature.
Joe:
My
guess
is
it’s
on
the
last
page,
genius.
Honestly,
fair
criticism
from
the
scoring
rubric.
Though
in
a
real
deposition,
my
approach
would’ve
totally
rattled
the
witness.
Or
gotten
me
sanctioned.
AltaClaro
and
Verbit
ran
a
beta
Early
Adopter
Program
with
six
firms
—
Orrick,
K&L
Gates,
McDermott,
Littler,
Taft,
and
Brownstein
Hyatt
—
that
generated
over
160
hours
of
testing.
The
numbers
from
that
pilot
are
pretty
telling:
97
percent
of
participants
strongly
agreed
the
tool
is
valuable
for
litigation
training
and
94
percent
said
they’d
use
it
again.
One
participating
partner
called
the
experience
“frighteningly
realistic,”
adding
that
the
feedback
was
“better
feedback
than
I’ve
received
from
attorneys.
Both
more
comprehensive
and
more
specific.”
Another
self-described
“confirmed
AI
hater”
among
the
associates
admitted
they
could
see
themselves
using
it
long-term.
At
launch,
the
product
only
provides
a
straightforward
civil
deposition,
but
at
my
demonstration,
they
explained
that
they
understand
the
potential
to
expand
the
available
scenarios.
As
a
technology,
there’s
no
reason
this
couldn’t
be
adapted
for
unique
situations
like
a
30(b)(6)
or
an
expert
witness.
With
a
little
tweaking,
it
could
provide
trial
examination
or
even
oral
argument
practice.
This
is
what
the
confirmed
AI
haters
miss.
For
every
task
that
AI
risks
disastrously
messing
up
without
focused
human
oversight,
there
are
tasks
that
AI
can
tackle
consistently
well…
and
more
efficiently.
While
the
greater
AI
industry
—
and
its
supporting
VC
culture
—
love
talking
about
replacing
white
collar
workers,
they
don’t
need
to
build
robot
lawyers
to
provide
value.
Crafting
a
tool
that
allows
lawyers
to
easily
train
for
depositions
on
their
own
schedule
is
a
perfect
use
case.
And
no
worrying
about
hallucinations.
Joe
Patrice is
a
senior
editor
at
Above
the
Law
and
co-host
of
Thinking
Like
A
Lawyer.
Feel
free
to email
any
tips,
questions,
or
comments.
Follow
him
on Twitter or
Bluesky
if
you’re
interested
in
law,
politics,
and
a
healthy
dose
of
college
sports
news.
Joe
also
serves
as
a
Managing
Director
at
RPN
Executive
Search.