1
00:00:00,440 --> 00:00:02,160
Speaker 1: Heather Duple class out.

2
00:00:02,160 --> 00:00:06,760
Speaker 2: More and more companies are using AI for recruiting staff. Spark, Kmart, Wilworths.

3
00:00:06,880 --> 00:00:08,799
Speaker 2: All of these guys in retail and hospital Apart from

4
00:00:08,840 --> 00:00:11,479
Speaker 2: Spark obviously, they all use the software by a company

5
00:00:11,520 --> 00:00:12,680
Speaker 2: called sapier Ai.

6
00:00:13,240 --> 00:00:13,440
Speaker 1: Now.

7
00:00:13,560 --> 00:00:16,599
Speaker 2: The program basically does the chat and video interviews and

8
00:00:16,640 --> 00:00:19,640
Speaker 2: then scores the candidate's written responses. The results can then

9
00:00:19,680 --> 00:00:22,720
Speaker 2: be tracked by the hiring team. Woolworths a loan, has

10
00:00:22,840 --> 00:00:25,799
Speaker 2: used the system to interview fifty nine thousand people in

11
00:00:25,800 --> 00:00:29,400
Speaker 2: the past year. Saperai founder and CEO Barb Heiman is

12
00:00:29,400 --> 00:00:32,760
Speaker 2: with us. Now, Hey, bub Hi, how does this work?

13
00:00:32,800 --> 00:00:32,879
Speaker 1: So?

14
00:00:32,960 --> 00:00:35,159
Speaker 2: Does it just ask sort of basic questions of the

15
00:00:35,159 --> 00:00:37,320
Speaker 2: first round applicants and then screen out the ones who've

16
00:00:37,320 --> 00:00:37,880
Speaker 2: got no shot?

17
00:00:39,040 --> 00:00:42,680
Speaker 1: Yeah, So what we do is basically do the interview

18
00:00:42,720 --> 00:00:44,640
Speaker 1: for you. If you think about hiring, you want to

19
00:00:44,640 --> 00:00:47,000
Speaker 1: meet someone, get to know them, figure out whether they're

20
00:00:47,040 --> 00:00:49,519
Speaker 1: the right foot of the job. The technology does it

21
00:00:49,560 --> 00:00:51,760
Speaker 1: for you. It does it in a way that is friendly.

22
00:00:52,360 --> 00:00:57,760
Speaker 1: It's not time pressured, and so you see amazing feedback

23
00:00:57,800 --> 00:01:02,400
Speaker 1: from candadate's. Candadates love it, feel stressed. It's much more inclusive,

24
00:01:02,440 --> 00:01:06,440
Speaker 1: for those with a disability, and it allows you to

25
00:01:06,440 --> 00:01:08,640
Speaker 1: get to the best candidates really fast, and often getting

26
00:01:08,640 --> 00:01:10,759
Speaker 1: to the best candidates first means that you win them.

27
00:01:11,760 --> 00:01:13,320
Speaker 2: How does it do the video interviews? I mean, I

28
00:01:13,360 --> 00:01:16,040
Speaker 2: can understand if you're doing like a text interview, you

29
00:01:16,080 --> 00:01:18,199
Speaker 2: can sort of screen out like that, But what about

30
00:01:18,200 --> 00:01:19,520
Speaker 2: a video one? What are you looking for there?

31
00:01:19,640 --> 00:01:22,240
Speaker 1: So the video there is no AI at all, and

32
00:01:22,280 --> 00:01:24,920
Speaker 1: we strongly believe that. Look, I would love video to

33
00:01:24,959 --> 00:01:28,200
Speaker 1: disappear as a tool to evaluate people because I think

34
00:01:28,200 --> 00:01:30,640
Speaker 1: it's very hard to remote bias. You know, most biases

35
00:01:30,680 --> 00:01:33,320
Speaker 1: are unconscious. The video is there is what we call

36
00:01:33,360 --> 00:01:35,839
Speaker 1: another data point, another signal, so someone has to watch

37
00:01:35,880 --> 00:01:39,200
Speaker 1: it and make a decision. You see that data the AI?

38
00:01:40,520 --> 00:01:42,880
Speaker 2: Does the AI do a video interview that's recorded and

39
00:01:42,920 --> 00:01:44,720
Speaker 2: then somebody needs to watch it? Does that how it works?

40
00:01:44,920 --> 00:01:47,280
Speaker 1: Yeah, it's not really an AI doing a video. It's

41
00:01:47,280 --> 00:01:49,200
Speaker 1: just like any video. You might use zoom to do

42
00:01:49,240 --> 00:01:51,920
Speaker 1: a video. There's no AI in the video at all.

43
00:01:51,960 --> 00:01:54,200
Speaker 1: The AI comes into the chats. So what we're doing

44
00:01:54,320 --> 00:01:57,160
Speaker 1: is taking language data. How does you respond to a

45
00:01:57,240 --> 00:01:59,240
Speaker 1: question about how you work with the team, how you

46
00:01:59,280 --> 00:02:02,280
Speaker 1: work with customers, and what do we learn from that response,

47
00:02:02,320 --> 00:02:04,520
Speaker 1: Do we learn that you're someone who's great with people,

48
00:02:05,040 --> 00:02:07,080
Speaker 1: that you really humble, that you're a fast learner, that

49
00:02:07,120 --> 00:02:09,520
Speaker 1: you're a good thinker. That's what we're doing, which is

50
00:02:09,520 --> 00:02:11,400
Speaker 1: what humans do, but we're doing it without all the

51
00:02:11,400 --> 00:02:15,200
Speaker 1: bias that we humans bring, and that's using science has

52
00:02:15,240 --> 00:02:18,880
Speaker 1: been around for decades now natural language processing, and that's

53
00:02:18,880 --> 00:02:22,120
Speaker 1: what provides for the accuracy of hiring, so that you

54
00:02:22,200 --> 00:02:24,760
Speaker 1: hire people who end up staying, not people who come

55
00:02:24,800 --> 00:02:26,359
Speaker 1: in and leave one week later.

56
00:02:27,120 --> 00:02:29,400
Speaker 2: So being a quick thinker is a positive.

57
00:02:30,840 --> 00:02:33,680
Speaker 1: Well depends on what the role requires. So each role

58
00:02:33,760 --> 00:02:36,240
Speaker 1: has its own custom requirements. If you think about your

59
00:02:36,280 --> 00:02:38,440
Speaker 1: hiring for someone in your team, or you're hiring for

60
00:02:38,480 --> 00:02:42,000
Speaker 1: someone for customer service cabin crew, it's fundamentally different what's

61
00:02:42,000 --> 00:02:44,639
Speaker 1: important for the role. So depending on what's important for

62
00:02:44,680 --> 00:02:47,280
Speaker 1: the role dictates the traits and the qualities that we're

63
00:02:47,280 --> 00:02:50,440
Speaker 1: looking for. Some jobs require really high critical thinking and

64
00:02:50,480 --> 00:02:55,400
Speaker 1: amazing common skills, others don't. And then we're identifying the questions.

65
00:02:55,400 --> 00:02:57,760
Speaker 1: We're creating the questions, which are always what we call

66
00:02:57,840 --> 00:03:00,400
Speaker 1: behavioral They ask you to draw on what you've done

67
00:03:00,440 --> 00:03:03,880
Speaker 1: in the past and in that basis, you get everyone

68
00:03:03,919 --> 00:03:05,799
Speaker 1: to share their story of who they are in a

69
00:03:05,840 --> 00:03:09,840
Speaker 1: way that is obviously a lot more comfortable than what

70
00:03:09,880 --> 00:03:13,120
Speaker 1: it often is for people facing someone across the table,

71
00:03:13,160 --> 00:03:15,360
Speaker 1: which is scary for a lot of candidates out there.

72
00:03:15,800 --> 00:03:19,680
Speaker 2: Bob, if being a quick thinker is something that's a

73
00:03:19,720 --> 00:03:23,920
Speaker 2: positive and presumably you are required to answer pretty articulately

74
00:03:24,000 --> 00:03:28,280
Speaker 2: quite fast, that would surely disadvange You're.

75
00:03:28,160 --> 00:03:30,760
Speaker 1: Not You're not. So what's really important is that the

76
00:03:30,840 --> 00:03:34,040
Speaker 1: chat is not timed. You could take an hour, you

77
00:03:34,040 --> 00:03:35,760
Speaker 1: could take twenty minutes, you could come back to it

78
00:03:35,800 --> 00:03:38,920
Speaker 1: five times. Well, we're not testing for quick thinking. What

79
00:03:38,960 --> 00:03:43,000
Speaker 1: we're testing for is attributes like critical thinking or are

80
00:03:43,040 --> 00:03:45,600
Speaker 1: you someone that's brilliant a customer service? Are you someone

81
00:03:45,680 --> 00:03:48,400
Speaker 1: that goes above and beyond? Are you someone that has

82
00:03:48,440 --> 00:03:52,000
Speaker 1: good communication skills? But speed we don't see in any

83
00:03:52,000 --> 00:03:55,560
Speaker 1: ways correlated with quality of higher So it is an

84
00:03:55,640 --> 00:03:58,160
Speaker 1: untimed interview, right.

85
00:03:58,920 --> 00:04:00,800
Speaker 2: So how do you though? I mean, one of the

86
00:04:00,800 --> 00:04:04,040
Speaker 2: things I imagine you're going to struggle to actually quantify

87
00:04:04,440 --> 00:04:09,280
Speaker 2: is charisma, personality, enthusiasm, tone, which is all the stuff

88
00:04:09,280 --> 00:04:12,440
Speaker 2: that you get in actual face to face interviews. How

89
00:04:12,440 --> 00:04:14,240
Speaker 2: can you possibly pick up whether somebody's got that.

90
00:04:15,720 --> 00:04:19,279
Speaker 1: Well, we would say that those are heuristics. They are

91
00:04:19,800 --> 00:04:22,080
Speaker 1: signs that we look for based on our gut or

92
00:04:22,160 --> 00:04:24,480
Speaker 1: we think what we think drive success in a role.

93
00:04:24,520 --> 00:04:28,039
Speaker 1: You gaves a lot of conversation about vibe hiring, but

94
00:04:28,279 --> 00:04:30,599
Speaker 1: is that actually delivering results for your business? You know,

95
00:04:30,640 --> 00:04:32,680
Speaker 1: what we're trying to solve for is who actually is

96
00:04:32,680 --> 00:04:34,880
Speaker 1: a fit for the role? Do you think to be

97
00:04:34,880 --> 00:04:37,600
Speaker 1: great and stick around? I do? I mean, what does

98
00:04:37,680 --> 00:04:41,360
Speaker 1: vibe high mean? Vibe high means basically I like you, yeah,

99
00:04:41,360 --> 00:04:42,839
Speaker 1: and I think you're a good fit for the brand.

100
00:04:42,880 --> 00:04:45,400
Speaker 1: It's what we call mirror hiring, and mirror hiring is

101
00:04:45,400 --> 00:04:46,239
Speaker 1: a recipe for buys.

102
00:04:47,720 --> 00:04:50,359
Speaker 2: But bob our guts have been have been trained for

103
00:04:51,000 --> 00:04:53,840
Speaker 2: flipp and millennia, but.

104
00:04:53,880 --> 00:04:55,960
Speaker 1: They're leading us astray. If you think about the two

105
00:04:56,040 --> 00:04:57,760
Speaker 1: most important decisions to make in our life, who do

106
00:04:57,800 --> 00:04:59,599
Speaker 1: we partner with? Where do we work? We get it

107
00:04:59,600 --> 00:05:02,360
Speaker 1: wrong with of the time, and so you know, how

108
00:05:02,360 --> 00:05:05,760
Speaker 1: do you kind of debias? The gut and the gut

109
00:05:06,080 --> 00:05:09,000
Speaker 1: and bias is a wonderful shortcut to make a fast decision,

110
00:05:09,000 --> 00:05:10,640
Speaker 1: but it doesn't necessarily make it a good decision.

111
00:05:10,720 --> 00:05:12,000
Speaker 2: Right, Well, are you going to roll out some AI

112
00:05:12,120 --> 00:05:13,279
Speaker 2: for picking husbands and wives?

113
00:05:13,320 --> 00:05:16,359
Speaker 1: Now you know what I can't. I would be a

114
00:05:16,360 --> 00:05:18,240
Speaker 1: millionaire from the number of times people have said that

115
00:05:18,279 --> 00:05:22,240
Speaker 1: to me. Really, yeah, everyone wants to use AI to

116
00:05:22,320 --> 00:05:24,400
Speaker 1: try and figure out the right partner, not the partner

117
00:05:24,440 --> 00:05:28,240
Speaker 1: that you pick because they you know, they meet your parents' expectations,

118
00:05:28,279 --> 00:05:30,040
Speaker 1: or they're a bit of a smunk. You know, how

119
00:05:30,080 --> 00:05:32,799
Speaker 1: do you actually choose someone who really aligns with your values?

120
00:05:32,880 --> 00:05:35,480
Speaker 1: Like you can do that through science. It's just that

121
00:05:35,560 --> 00:05:40,000
Speaker 1: we as humans feel really uncomfortable using science, you know,

122
00:05:40,160 --> 00:05:42,080
Speaker 1: data and anything but the gut or the gut of

123
00:05:42,120 --> 00:05:45,040
Speaker 1: our friends and family to help us on those decisions.

124
00:05:44,640 --> 00:05:49,000
Speaker 1: It's a lot of human bias really around using a

125
00:05:49,040 --> 00:05:50,560
Speaker 1: different approach for these decisions.

126
00:05:50,880 --> 00:05:53,160
Speaker 2: Interesting, Bob, thank you very much for talking us STIFYD.

127
00:05:53,160 --> 00:05:55,960
Speaker 2: They absolutely fascinating. It is Bob Heiman, founder and CEO

128
00:05:56,000 --> 00:05:57,800
Speaker 2: of sa PAAI. Do you know what I bet you

129
00:05:57,839 --> 00:05:59,520
Speaker 2: there is something out there where you can find a

130
00:05:59,520 --> 00:06:02,240
Speaker 2: partner you can I mean that's basically what Tinder is,

131
00:06:02,240 --> 00:06:03,720
Speaker 2: isn't it where you sort of like you put all

132
00:06:03,760 --> 00:06:06,320
Speaker 2: your important stuff, but it's it's not smart enough. You've

133
00:06:06,320 --> 00:06:07,680
Speaker 2: got to go for you've got to go for more

134
00:06:07,720 --> 00:06:10,279
Speaker 2: than that. You've got to go for values, like like

135
00:06:10,480 --> 00:06:13,520
Speaker 2: values to me would be how often do you wash

136
00:06:13,600 --> 00:06:18,360
Speaker 2: the sheets? Do you wipe the bench after you've made

137
00:06:18,360 --> 00:06:20,000
Speaker 2: a coffee and left the crimes? You know, like that

138
00:06:20,080 --> 00:06:21,600
Speaker 2: kind of stuff like I'm into that. I think that's

139
00:06:21,640 --> 00:06:23,680
Speaker 2: really important, and that kind of most of those things.

140
00:06:23,680 --> 00:06:25,640
Speaker 2: If the AI could just solve that stuff like no

141
00:06:25,760 --> 00:06:28,440
Speaker 2: lies please, also has the screen out whether the person

142
00:06:28,520 --> 00:06:30,440
Speaker 2: is lying just to impress you. If you could fix

143
00:06:30,480 --> 00:06:32,920
Speaker 2: all of that stuffy how many marriages would be saved

144
00:06:33,160 --> 00:06:35,320
Speaker 2: by the dishwasher washer being packed properly.

145
00:06:35,000 --> 00:06:39,000
Speaker 1: Out For more from Heather Duplessy Allen Drive, listen live

146
00:06:39,160 --> 00:06:39,839
Speaker 1: to news talks.

147
00:06:39,839 --> 00:06:43,040
Speaker 2: It'd be from four pm weekdays, or follow the podcast

148
00:06:43,120 --> 00:06:44,080
Speaker 2: on iHeartRadio