1
00:00:02,360 --> 00:00:02,960
Hi guys.

2
00:00:02,960 --> 00:00:04,000
Welcome to this session.

3
00:00:04,000 --> 00:00:08,760
So in this session we wanted to talk about the generative AI business case studies, some successes

4
00:00:08,760 --> 00:00:11,240
and failures we have seen in this particular field.

5
00:00:11,800 --> 00:00:16,680
So the first one we are going to look at is the Khan Academy's Conmigo which they created.

6
00:00:16,680 --> 00:00:19,520
So this is an AI tutor which they built out.

7
00:00:19,520 --> 00:00:25,240
The Khan Academy built out after ChatGPT was launched, which helps students through critical thinking.

8
00:00:25,520 --> 00:00:29,600
It helps teachers with lesson planning and for creating curriculum.

9
00:00:29,840 --> 00:00:35,320
And how it worked is it worked because of the deep integration with the Khan Academy platform they had.

10
00:00:35,360 --> 00:00:41,600
It involved 10,000 plus students and 8000 plus tutors before releasing.

11
00:00:41,800 --> 00:00:47,560
So the key learnings from this particular integration was that the aligning of the AI capabilities with

12
00:00:47,560 --> 00:00:50,880
real user needs, and it worked out really well.

13
00:00:51,080 --> 00:00:56,960
It had a rigorous amount of testings done so that we could understand the success rates were much more

14
00:00:57,000 --> 00:00:58,040
higher with this.

15
00:00:58,960 --> 00:01:03,480
Another example can be of Coca-Cola's Create Real Magic campaign.

16
00:01:03,480 --> 00:01:11,360
So this was a campaign which they ran, where it was a bold mix of AI tools and Coca-Cola's iconic brand

17
00:01:11,360 --> 00:01:12,040
elements.

18
00:01:12,240 --> 00:01:18,550
So the users created unique artwork using logos, bottles and established guidelines.

19
00:01:19,150 --> 00:01:25,710
Why it worked majorly was because of the creative freedom people got paired with clear brand boundaries.

20
00:01:25,870 --> 00:01:31,470
It scaled globally across using ChatGPT and Dall-E across 40 plus markets.

21
00:01:31,750 --> 00:01:38,270
Now the key learnings from this campaign was the generative AI thrives with freedom and structure specifically,

22
00:01:38,270 --> 00:01:43,710
and the strategic guardrails maximizes impact while protecting the brand.

23
00:01:44,110 --> 00:01:50,070
Now, if you look at another use case which was of Air Canada's chatbot incident, which happened.

24
00:01:50,270 --> 00:01:56,390
So basically what happened was Air Canada's chatbot gave some incorrect information about bereavement

25
00:01:56,390 --> 00:01:57,110
fares.

26
00:01:57,550 --> 00:02:03,670
Now, because of this misinformation, it led to financial losses and legal trouble for a customer.

27
00:02:04,230 --> 00:02:11,390
The fallout of that was that Air Canada deflected responsibility and calling the chatbot or third party,

28
00:02:11,670 --> 00:02:13,070
a separate legal entity.

29
00:02:13,270 --> 00:02:20,270
And because of which a civil resolution tribunal ruled against the company, setting a legal precedent.

30
00:02:20,630 --> 00:02:25,750
Now, the key learnings from this incident was that the there was insufficient testing and lack of human

31
00:02:26,070 --> 00:02:29,710
oversight, which was seen in this particular case.

32
00:02:29,950 --> 00:02:34,110
And there was poor crisis management, which made things worse.

33
00:02:34,940 --> 00:02:41,100
Lastly, what we want to look at another scenario which is Google's Gemini image generation feature.

34
00:02:41,340 --> 00:02:45,140
So the goal of this particular feature, it was launched.

35
00:02:45,180 --> 00:02:52,060
The Gemini tool was launched by Google in February 2024, wherein the AI model was to reflect the company's

36
00:02:52,060 --> 00:02:55,740
values around diversity and inclusion.

37
00:02:55,780 --> 00:03:03,940
Now, what went wrong was it produced historically inaccurate images, which sparked a lot of backlash.

38
00:03:04,180 --> 00:03:05,420
It criticized.

39
00:03:05,420 --> 00:03:11,020
It was criticized for potentially trying to reshape the whole societal narratives.

40
00:03:11,180 --> 00:03:17,780
And the key learnings which they had from this was that there was no not enough robust data, and human

41
00:03:17,780 --> 00:03:22,420
oversight was, uh, was used and which was critical for accuracy.

42
00:03:22,700 --> 00:03:28,580
Uh, the balance includes inclusivity with historical context through rigorous training.

43
00:03:28,620 --> 00:03:30,180
Testing was not done.

44
00:03:30,220 --> 00:03:34,780
The positive and responsible, responsible decision to remove the feature.

45
00:03:34,780 --> 00:03:41,220
So what the brand did was they went ahead and they took the right decision to remove the feature eventually.

46
00:03:41,540 --> 00:03:48,300
So these are some case studies scenarios in real life which you can see how AI integrated with different

47
00:03:48,300 --> 00:03:51,020
brands and what impact it brought into the world.