WEBVTT

00:00:00.160 --> 00:00:01.720
Anne, thank you so much for
being here.

00:00:01.920 --> 00:00:04.840
What are some of the biggest
risks organizations face when it

00:00:04.840 --> 00:00:05.840
comes to using GenAI?

00:00:06.880 --> 00:00:07.720
So GenAI

00:00:07.720 --> 00:00:11.960
systems have to be proficient,
so they have to bring the

00:00:11.960 --> 00:00:13.960
intended value to the user.

00:00:14.200 --> 00:00:16.360
They have to be safe
and equitable.

00:00:16.680 --> 00:00:17.840
They have to be secure.

00:00:18.000 --> 00:00:22.000
And, of course, they have to be
compliant with regulation, for

00:00:22.000 --> 00:00:25.200
example, the regulation provided
by the EU AI Act.

00:00:25.720 --> 00:00:30.520
And systems that don't, aren't
tested across all of these four

00:00:30.520 --> 00:00:34.920
dimensions bring risks to the
organizations that they are

00:00:34.920 --> 00:00:36.360
actually employing.

00:00:36.760 --> 00:00:39.560
So to expand on that a little
bit, how can these companies

00:00:39.560 --> 00:00:42.560
really make sure their systems
are both innovative and reliable?

00:00:42.880 --> 00:00:47.280
So they should really test and
evaluate them across these

00:00:47.280 --> 00:00:52.000
dimensions that I mentioned:
proficiency, safe and equitable,

00:00:52.000 --> 00:00:52.840
and secure.

00:00:53.240 --> 00:00:56.880
And we, as BCG X, have built an
open-source

00:00:56.880 --> 00:01:01.640
solution--ARTKIT--which really
helps in in doing that and

00:01:01.640 --> 00:01:06.160
automating that to some extent
by, you know, combining

00:01:06.160 --> 00:01:11.600
automation together with human
test-and- evaluation capabilities.

00:01:11.920 --> 00:01:16.840
So what can companies do to
build and also maintain customer trust?

00:01:17.160 --> 00:01:22.600
For customer trust, I think that
transparency is extremely important.

00:01:22.920 --> 00:01:27.880
So customers will need to know:
Do I interact with an AI, or do

00:01:27.880 --> 00:01:29.720
I interact with a human?

00:01:29.720 --> 00:01:33.440
And that's, by the way, also
something that is necessary from

00:01:33.440 --> 00:01:36.480
a perspective of the EU AI Act,
for example, yeah?

00:01:36.520 --> 00:01:39.240
So transparency is key.

00:01:39.440 --> 00:01:43.200
But, then of course, at the same
time, it is exactly reiterating

00:01:43.200 --> 00:01:45.120
the points: the systems
have to be helpful

00:01:45.480 --> 00:01:48.960
for the customer, they have to
be free of hate speech, they

00:01:48.960 --> 00:01:51.800
have to be free of
discrimination, and they also

00:01:51.800 --> 00:01:54.800
have to be safeguarding personal
data, for example.

00:01:55.280 --> 00:01:56.320
Thank you so much.

00:01:57.040 --> 00:01:57.520
Thank you.