WEBVTT
X-TIMESTAMP-MAP=LOCAL:00:00:00.000,MPEGTS:0

00:00.240 --> 00:01.840
At PostNL we aim to be the

00:01.840 --> 00:05.680
favourite deliverer within the Benelux
and from and towards the Benelux.

00:05.720 --> 00:09.000
We like to offer distinctive experiences
to our customers

00:09.000 --> 00:12.920
and Generative AI is a really new technology,
and promising technology.

00:13.320 --> 00:16.360
It will help us to make our processes
more efficient

00:16.720 --> 00:19.720
and improve our services
towards customers.

00:27.880 --> 00:30.880
The opportunity is to see if,

00:31.200 --> 00:34.200
a lot of repetitive work or maybe,

00:34.560 --> 00:39.000
huge amounts of data,
which are maybe humanly almost impossible

00:39.240 --> 00:42.120
to process, that we can use the technology

00:42.120 --> 00:45.120
to get faster and better answers.

00:45.560 --> 00:47.280
So specifically in our

00:47.280 --> 00:51.240
customer service departments,
finance, maybe marketing.

00:51.520 --> 00:53.240
There's a lot of opportunity over there.

00:53.240 --> 00:54.360
With this new technology

00:54.640 --> 00:57.640
we looked for a partner
who could help us in organising

00:57.800 --> 00:59.960
the whole new knowledge that came about.

00:59.960 --> 01:03.440
Deloitte and PostNL have been partners
for many years.

01:03.640 --> 01:06.760
We made an elaborative
plan and put it to work.

01:07.000 --> 01:11.480
It is of utmost importance
to implement

01:11.480 --> 01:15.960
AI solutions that are reliable,
trustworthy and transparent.

01:17.280 --> 01:21.360
So at PostNL we
actually did a very pragmatic approach

01:21.360 --> 01:24.400
in which we developed risk checklists or

01:24.400 --> 01:27.400
enriched
the existing checklists of PostNL.

01:27.560 --> 01:32.200
A good example of that is,
for instance, a question in the AI act

01:32.560 --> 01:37.520
that requires organisations to validate
if the end users, for instance, consumers

01:38.000 --> 01:42.000
are aware that they are interacting
with an AI system.

01:42.360 --> 01:44.360
This is a straightforward check,

01:44.360 --> 01:48.400
but needs to be embedded or enriched
within the existing risk frameworks

01:48.560 --> 01:52.240
to ensure that the organisation complies
with the regulation.

01:52.440 --> 01:56.360
The approach to this engagement
was focussed on delivering value fast.

01:56.600 --> 02:00.400
So through fast prototyping,
we were able to create proof of concepts

02:00.400 --> 02:04.840
that yielded between 60 and 80%
of resolution rate already in six weeks.

02:05.320 --> 02:08.840
This approach was done
with the developers of PostNL

02:08.840 --> 02:12.800
within PostNL, and actually scaled up
and delivered by PostNL.

02:12.920 --> 02:16.040
We have actually experimented
whether it would work

02:16.040 --> 02:19.480
in the reality by doing simulation games.

02:19.600 --> 02:23.480
So based on some of the identified GenAI
use cases,

02:23.840 --> 02:26.600
we identified different
stakeholder groups

02:26.600 --> 02:31.520
and we simulated the processes as
if it were already developed and deployed.

02:31.880 --> 02:34.800
By those simulation games,
we were able to test

02:34.800 --> 02:37.800
whether those risk elements
were covered properly

02:37.880 --> 02:40.360
and what are the gaps to address
going forward.

02:40.360 --> 02:44.360
So this is one very pragmatic example
of how we embedded it within PostNL.

02:48.040 --> 02:51.040
A lesson learned from this project is

02:51.640 --> 02:56.160
the importance of data quality
for designing reliable AI solutions.

02:56.960 --> 02:59.600
So what we have noticed
is that the concept of garbage in

02:59.600 --> 03:02.480
garbage out plays an
even more important role here.

03:02.480 --> 03:05.120
But also the adoption from end users

03:05.120 --> 03:08.120
is being driven
by the reliability of the data.

03:08.280 --> 03:11.480
And when I talk to you about data quality,
it's about completeness,

03:12.160 --> 03:13.640
correctness of the data.

03:13.640 --> 03:17.120
That really is important from the
beginning to pay attention.

03:17.440 --> 03:20.440
Key factor,
I think, is a lot of communication.

03:20.680 --> 03:24.000
So listening to people, what is your job?

03:24.000 --> 03:26.200
What do you do on a daily basis?

03:26.200 --> 03:28.160
What things do you like?

03:28.160 --> 03:29.280
What don't you like?

03:29.960 --> 03:33.040
And also
maintaining the human in the loop.

03:33.080 --> 03:34.640
So not trying to replace

03:34.640 --> 03:38.000
a human by technology, but
let the human work with the technology.

03:38.440 --> 03:41.200
AI and GenAI
is not about replacing humans.

03:41.200 --> 03:43.760
It's about supercharging productivity.

03:43.760 --> 03:48.720
Whereas sales and service agents can focus
on the things that really matter.

03:49.240 --> 03:51.800
Having meaningful conversations
with customers.

03:52.200 --> 03:54.960
A couple of tips for companies
starting with GenAI would be

03:54.960 --> 03:57.960
start small, fail fast.

03:58.080 --> 04:01.360
Talk with your colleagues
and all the teams in your company

04:01.520 --> 04:05.080
if you learn, you can increase the steps
and bring more value to the company.