WEBVTT

1
00:00:00.000 --> 00:00:00.750


2
00:00:00.750 --> 00:00:03.300
SAM RANSBOTHAM:
Real-time data collection

3
00:00:03.300 --> 00:00:06.250
means organizations can make
many more informed choices

4
00:00:06.250 --> 00:00:07.790
based on metrics.

5
00:00:07.790 --> 00:00:10.090
But when do they
still need humans?

6
00:00:10.090 --> 00:00:12.190
Find out on today's episode

7
00:00:12.190 --> 00:00:15.140
AMEEN KAZEROUNI: I'm Ameen
Kazerouni from Orangetheory

8
00:00:15.140 --> 00:00:17.840
Fitness, and you're listening
to Me, Myself, and AI.

9
00:00:17.840 --> 00:00:21.560
SAM RANSBOTHAM: Welcome
to Me, Myself, and AI,

10
00:00:21.560 --> 00:00:24.570
a podcast on artificial
intelligence in business.

11
00:00:24.570 --> 00:00:28.330
Each episode, we introduce you
to someone innovating with AI.

12
00:00:28.330 --> 00:00:32.630
I'm Sam Ransbotham, professor
of analytics at Boston College.

13
00:00:32.630 --> 00:00:36.180
I'm also the AI and business
strategy guest editor

14
00:00:36.180 --> 00:00:37.790
at MIT Sloan Management Review.

15
00:00:37.790 --> 00:00:40.090
SHERVIN KHODABANDEH: And
I'm Shervin Khodabandeh,

16
00:00:40.090 --> 00:00:44.060
senior partner with BCG, and
I colead BCG's AI practice

17
00:00:44.060 --> 00:00:45.050
in North America.

18
00:00:45.050 --> 00:00:49.550
Together, MIT SMR and BCG have
been researching and publishing

19
00:00:49.550 --> 00:00:52.490
on AI for six years,
interviewing hundreds

20
00:00:52.490 --> 00:00:54.610
of practitioners and
surveying thousands

21
00:00:54.610 --> 00:00:58.090
of companies on what it takes
to build and to deploy and scale

22
00:00:58.090 --> 00:01:00.560
AI capabilities and
really transform

23
00:01:00.560 --> 00:01:01.890
the way organizations operate.

24
00:01:01.890 --> 00:01:04.349
SAM RANSBOTHAM:
Today, Shervin and I

25
00:01:04.349 --> 00:01:07.290
are excited to be joined by
Ameen Kazerouni, chief data

26
00:01:07.290 --> 00:01:09.622
and analytics officer
for Orangetheory Fitness.

27
00:01:09.622 --> 00:01:10.830
Ameen, thanks for joining us.

28
00:01:10.830 --> 00:01:11.210
Welcome.

29
00:01:11.210 --> 00:01:12.240
AMEEN KAZEROUNI: Thank you, Sam.

30
00:01:12.240 --> 00:01:13.190
It's great to be here.

31
00:01:13.190 --> 00:01:14.500
Excited for the conversation.

32
00:01:14.500 --> 00:01:16.333
SAM RANSBOTHAM: Currently,
you lead the data

33
00:01:16.333 --> 00:01:18.580
and analytics function
at Orangetheory Fitness.

34
00:01:18.580 --> 00:01:20.690
Maybe tell us a little bit
about the organization.

35
00:01:20.690 --> 00:01:22.060
AMEEN KAZEROUNI: Absolutely.

36
00:01:22.060 --> 00:01:26.110
Orangetheory Fitness is a
heart-rate-based, total-body

37
00:01:26.110 --> 00:01:27.100
group workout.

38
00:01:27.100 --> 00:01:31.500
It combines science, great
coaching, technology,

39
00:01:31.500 --> 00:01:35.210
and it's designed
to provide what

40
00:01:35.210 --> 00:01:37.660
we like to think of as
a more vibrant life.

41
00:01:37.660 --> 00:01:42.180
The workout's developed to
motivate each individual member

42
00:01:42.180 --> 00:01:44.945
to achieve their
desired results.

43
00:01:44.945 --> 00:01:46.320
And, you know, if
you're starting

44
00:01:46.320 --> 00:01:50.150
on your wellness journey or
you're a seasoned fitness

45
00:01:50.150 --> 00:01:55.280
enthusiast, each OTF
workout creates a community

46
00:01:55.280 --> 00:01:59.780
of shared experience but also
uses heart-rate-based training

47
00:01:59.780 --> 00:02:03.090
to allow you to experience
the workout in a way

48
00:02:03.090 --> 00:02:06.260
that's most comfortable for
you, and that's honestly

49
00:02:06.260 --> 00:02:07.310
where my role comes in.

50
00:02:07.310 --> 00:02:11.410
There's a tremendous amount of
second-by-second telemetry data

51
00:02:11.410 --> 00:02:13.810
from the fitness
equipment, from the heart

52
00:02:13.810 --> 00:02:16.060
rate monitors, that
allow us to create

53
00:02:16.060 --> 00:02:19.550
the most curated, personalized
kind of boutique fitness

54
00:02:19.550 --> 00:02:20.737
experience in the world.

55
00:02:20.737 --> 00:02:21.570
That's Orangetheory.

56
00:02:21.570 --> 00:02:23.640
SAM RANSBOTHAM: All right,
what do you do with all this?

57
00:02:23.640 --> 00:02:25.223
You've collected all
this data, you've

58
00:02:25.223 --> 00:02:27.900
got it, this telemetry, you've
got heart rate information,

59
00:02:27.900 --> 00:02:29.880
I assume, since you're
heart-rate-based.

60
00:02:29.880 --> 00:02:31.137
How does the process work?

61
00:02:31.137 --> 00:02:32.220
Take us through the steps.

62
00:02:32.220 --> 00:02:32.720


63
00:02:32.720 --> 00:02:36.190
AMEEN KAZEROUNI: Most
Orangetheory members

64
00:02:36.190 --> 00:02:39.480
in the studio will be wearing
what we call an OTbeat heart

65
00:02:39.480 --> 00:02:41.580
rate monitor, which
is a proprietary piece

66
00:02:41.580 --> 00:02:42.940
of wearable technology.

67
00:02:42.940 --> 00:02:45.010
There's two purposes to that.

68
00:02:45.010 --> 00:02:50.390
One is, it gives you a real-live
feedback loop as to how you're

69
00:02:50.390 --> 00:02:53.530
performing, what intensity
level you're outputting

70
00:02:53.530 --> 00:02:56.330
in the studio, but it
also allows the coach

71
00:02:56.330 --> 00:02:59.200
to see the intensity level that
you're outputting in the studio

72
00:02:59.200 --> 00:03:02.870
and help effectively provide
a kind of personalized fitness

73
00:03:02.870 --> 00:03:06.800
training experience in a
studio, in a group setting.

74
00:03:06.800 --> 00:03:12.060
My role is focused on kind of
unlocking that telemetry data

75
00:03:12.060 --> 00:03:15.520
and helping personalize
the experience even more.

76
00:03:15.520 --> 00:03:18.900
A really cool example is
that we recently launched

77
00:03:18.900 --> 00:03:22.520
a personalized max
heart rate algorithm,

78
00:03:22.520 --> 00:03:28.580
and members now experience a
much more curated experience

79
00:03:28.580 --> 00:03:33.090
in the studio, and that's
allowing us to use proprietary

80
00:03:33.090 --> 00:03:35.810
algorithms to determine
what the max heart rate --

81
00:03:35.810 --> 00:03:39.120
which is a physiological term
for the maximum output that

82
00:03:39.120 --> 00:03:42.380
your heart can beat at --
is for an individual member.

83
00:03:42.380 --> 00:03:44.830
And percentages of
that max heart rate

84
00:03:44.830 --> 00:03:47.480
tell you which heart rate
zone you're training in.

85
00:03:47.480 --> 00:03:50.450
So anaerobic training
versus aerobic training

86
00:03:50.450 --> 00:03:52.100
have different
physiological impacts.

87
00:03:52.100 --> 00:03:54.580
Time spent in different
intensity zones

88
00:03:54.580 --> 00:03:59.040
have been proven to have varying
effects on longevity and health

89
00:03:59.040 --> 00:03:59.540
in general.

90
00:03:59.540 --> 00:04:02.560
And by being able to
personalize that for a member,

91
00:04:02.560 --> 00:04:04.790
we're able to make
this experience even

92
00:04:04.790 --> 00:04:09.890
more curated in the studio,
while most places that leverage

93
00:04:09.890 --> 00:04:14.580
max heart rate will rely
on a generic age-based kind

94
00:04:14.580 --> 00:04:15.460
of equation.

95
00:04:15.460 --> 00:04:19.970
And as you can imagine, every
30-year-old or 40-year-old

96
00:04:19.970 --> 00:04:21.180
doesn't have the same heart.

97
00:04:21.180 --> 00:04:23.430
So things like
that are an example

98
00:04:23.430 --> 00:04:26.150
of how we curate the
experience for our members

99
00:04:26.150 --> 00:04:27.245
using this data.

100
00:04:27.245 --> 00:04:29.620
SHERVIN KHODABANDEH: That is
a super-cool example, right?

101
00:04:29.620 --> 00:04:32.510
Not the average for
your age and gender.

102
00:04:32.510 --> 00:04:35.220
And then that goes by
bands of 10 anyway, right?

103
00:04:35.220 --> 00:04:35.543


104
00:04:35.543 --> 00:04:36.460
AMEEN KAZEROUNI: Yeah.

105
00:04:36.460 --> 00:04:41.480
SHERVIN KHODABANDEH: As if
all 50-year-old males exactly

106
00:04:41.480 --> 00:04:42.630
have the same ability.

107
00:04:42.630 --> 00:04:44.630
AMEEN KAZEROUNI: Yeah.

108
00:04:44.630 --> 00:04:46.810
It makes the
experience safer, it

109
00:04:46.810 --> 00:04:49.120
makes you more
aware of what you're

110
00:04:49.120 --> 00:04:51.530
doing, what your
capability is, and you

111
00:04:51.530 --> 00:04:53.910
see that cardiorespiratory
fitness climb

112
00:04:53.910 --> 00:04:55.720
over your time with the program.

113
00:04:55.720 --> 00:04:58.710
SHERVIN KHODABANDEH: And I like
what's sort of inherent in what

114
00:04:58.710 --> 00:05:02.980
you're saying -- the
rapid feedback, right?

115
00:05:02.980 --> 00:05:06.560
Within a few seconds, you
get feedback, but also,

116
00:05:06.560 --> 00:05:10.430
in a broader, symbolic sense,
what you've been proposing

117
00:05:10.430 --> 00:05:13.880
is more and more
experimentation in general

118
00:05:13.880 --> 00:05:15.640
as you build AI algorithms.

119
00:05:15.640 --> 00:05:18.860
So it's not just the algo,
but it's also experimentation

120
00:05:18.860 --> 00:05:20.810
because you get feedback.

121
00:05:20.810 --> 00:05:24.270
The philosophy of how
to build AI models,

122
00:05:24.270 --> 00:05:27.400
which relies on
feedback, you're also

123
00:05:27.400 --> 00:05:31.587
translating that into a real use
case based on rapid feedback.

124
00:05:31.587 --> 00:05:32.670
There's some poetry there.

125
00:05:32.670 --> 00:05:34.770
AMEEN KAZEROUNI: I didn't
think of it that way,

126
00:05:34.770 --> 00:05:36.420
but absolutely
there's poetry there.

127
00:05:36.420 --> 00:05:39.540
I think that one of the beauties
of artificial intelligence

128
00:05:39.540 --> 00:05:41.770
algorithms is that they're
so reliant on things

129
00:05:41.770 --> 00:05:44.580
they've seen before and constant
feedback loops to get smarter.

130
00:05:44.580 --> 00:05:48.040
And that's exactly what the
Orangetheory experience is

131
00:05:48.040 --> 00:05:50.630
indexed on is, as the
members are getting stronger,

132
00:05:50.630 --> 00:05:54.120
we're getting smarter and making
sure that we move with them.

133
00:05:54.120 --> 00:05:57.460
SHERVIN KHODABANDEH: Tell us
more a bit about your data

134
00:05:57.460 --> 00:06:00.570
science development
philosophy and how

135
00:06:00.570 --> 00:06:06.000
you balance experimentation
with more methodical

136
00:06:06.000 --> 00:06:09.070
"get the algorithm to be
perfect before you launch it."

137
00:06:09.070 --> 00:06:10.930
Just give us a sense
of the trade-offs

138
00:06:10.930 --> 00:06:13.100
and how you and your
teams think about that.

139
00:06:13.100 --> 00:06:14.850
AMEEN KAZEROUNI: I
love that question,

140
00:06:14.850 --> 00:06:16.260
and it's almost
like I'll take it

141
00:06:16.260 --> 00:06:21.030
even a step outside of
developing algorithms.

142
00:06:21.030 --> 00:06:23.570
I'm a firm believer
that companies

143
00:06:23.570 --> 00:06:29.000
have started indexing so heavily
on collecting as much data

144
00:06:29.000 --> 00:06:32.100
as is humanly possible;
as much as their compute

145
00:06:32.100 --> 00:06:33.980
and their storage
costs, and their boards

146
00:06:33.980 --> 00:06:36.460
and their investors,
will allow them to,

147
00:06:36.460 --> 00:06:37.990
companies collect data.

148
00:06:37.990 --> 00:06:40.630
And I think that what happens
is, once you have the data,

149
00:06:40.630 --> 00:06:43.070
the expectation is, "Let's
jump to machine learning;

150
00:06:43.070 --> 00:06:44.360
let's jump to AI."

151
00:06:44.360 --> 00:06:47.370
And I would argue that those ...

152
00:06:47.370 --> 00:06:49.370
How many debates have you
been in on "What's AI,

153
00:06:49.370 --> 00:06:52.110
and what's machine learning,
and what do the words mean?"

154
00:06:52.110 --> 00:06:57.140
My philosophy is to go and find
those mundane, repetitive tasks

155
00:06:57.140 --> 00:07:00.000
and automate them first with
your data, where possible.

156
00:07:00.000 --> 00:07:03.720
Go and find intuitive,
gut-based decisions

157
00:07:03.720 --> 00:07:06.750
that your stakeholders
and verticals

158
00:07:06.750 --> 00:07:09.800
are uncomfortable making
based off of intuition

159
00:07:09.800 --> 00:07:11.460
and would love to
make off of data,

160
00:07:11.460 --> 00:07:16.040
and make that data democratized,
clean, and available.

161
00:07:16.040 --> 00:07:20.030
And after that, you start the
whole machine learning journey.

162
00:07:20.030 --> 00:07:24.210
And I think when it comes
to machine learning and AI,

163
00:07:24.210 --> 00:07:26.690
and developing an
algorithm in particular,

164
00:07:26.690 --> 00:07:29.370
it really depends on the
context and the domain in which

165
00:07:29.370 --> 00:07:31.950
you're working -- whether you're
focusing on that precision,

166
00:07:31.950 --> 00:07:33.830
whether you're focusing
on that recall --

167
00:07:33.830 --> 00:07:36.860
and it really depends
on what the implications

168
00:07:36.860 --> 00:07:38.310
of the prediction are.

169
00:07:38.310 --> 00:07:41.340
But generally
speaking, I always err

170
00:07:41.340 --> 00:07:44.270
on the side of, if it's
safe enough, experiment

171
00:07:44.270 --> 00:07:48.570
and learn rather than relying
on training and validation sets

172
00:07:48.570 --> 00:07:49.890
to chase perfection.

173
00:07:49.890 --> 00:07:52.435
It's kind of my
rule-of-thumb philosophy.

174
00:07:52.435 --> 00:07:54.560
SAM RANSBOTHAM: There's
something also interesting,

175
00:07:54.560 --> 00:07:55.977
and I know you've
shifted slightly

176
00:07:55.977 --> 00:07:58.540
into thinking about how
your organization uses data,

177
00:07:58.540 --> 00:08:01.050
but maybe going back
to how you're using it

178
00:08:01.050 --> 00:08:04.890
within the studios, how do you
incorporate data from outside?

179
00:08:04.890 --> 00:08:08.090
You know, let's say that I've
been training and working

180
00:08:08.090 --> 00:08:12.000
and improving within
the studio, but little

181
00:08:12.000 --> 00:08:13.980
do you know that
I've hurt my foot

182
00:08:13.980 --> 00:08:16.270
or I've pulled a leg muscle.

183
00:08:16.270 --> 00:08:18.540
How does that sort of
outside information come in?

184
00:08:18.540 --> 00:08:20.590
AMEEN KAZEROUNI: I like
that question a lot,

185
00:08:20.590 --> 00:08:23.230
Sam, because I've got
two answers there.

186
00:08:23.230 --> 00:08:28.070
One is, the coach is the
hero at Orangetheory.

187
00:08:28.070 --> 00:08:29.950
And the way we
think about it is,

188
00:08:29.950 --> 00:08:32.559
there's 20 to 30
people in a class,

189
00:08:32.559 --> 00:08:35.650
but you don't want to think
of it as group fitness as much

190
00:08:35.650 --> 00:08:37.870
as you want to think
of it as -- probably,

191
00:08:37.870 --> 00:08:40.460
at least I think of it as --
subsidized personal training.

192
00:08:40.460 --> 00:08:43.390
So you've got a pretty
good personal relationship

193
00:08:43.390 --> 00:08:46.090
with that coach, and
there's no algorithm

194
00:08:46.090 --> 00:08:48.360
that I'm going to build
that's going to be better

195
00:08:48.360 --> 00:08:50.370
than you telling
your coach, "Hey, I

196
00:08:50.370 --> 00:08:51.618
hurt my foot last night.

197
00:08:51.618 --> 00:08:52.910
What do you think I should do?"

198
00:08:52.910 --> 00:08:56.140
And we've got alternative,
low-impact cardio equipment

199
00:08:56.140 --> 00:08:59.620
in the studio so you
don't get on a treadmill

200
00:08:59.620 --> 00:09:01.560
and start trying to
run 8 miles an hour.

201
00:09:01.560 --> 00:09:04.480
In fact, one of the
things the coach says

202
00:09:04.480 --> 00:09:06.120
before every class
starts is, "If you

203
00:09:06.120 --> 00:09:08.120
got any orthopedic issues,
please let me know,"

204
00:09:08.120 --> 00:09:09.860
and so on and so forth.

205
00:09:09.860 --> 00:09:13.100
So I think that
one thing that I've

206
00:09:13.100 --> 00:09:15.150
learned about data over
the course of my career

207
00:09:15.150 --> 00:09:17.300
is that data is
valuable, and it's

208
00:09:17.300 --> 00:09:18.860
really good to make
decisions with,

209
00:09:18.860 --> 00:09:23.010
but if there's a human in the
loop that can scalably provide

210
00:09:23.010 --> 00:09:26.720
the answer instead,
it's likely going

211
00:09:26.720 --> 00:09:29.390
be hard to beat that
expert with an algorithm.

212
00:09:29.390 --> 00:09:32.010
So don't get in the
way of the expert.

213
00:09:32.010 --> 00:09:34.070
The second part
of that answer is,

214
00:09:34.070 --> 00:09:35.850
let's say you had a
bunch of caffeine,

215
00:09:35.850 --> 00:09:38.820
or you ran a marathon
the previous day,

216
00:09:38.820 --> 00:09:43.760
or you didn't sleep very well;
all those external factors

217
00:09:43.760 --> 00:09:46.140
affect your cardiac output
when you're in the studio.

218
00:09:46.140 --> 00:09:49.400
So that real-time feedback
loop allows you and the coach

219
00:09:49.400 --> 00:09:50.600
to modulate yourself.

220
00:09:50.600 --> 00:09:53.590
One thing that we encourage
is, if you're not feeling it,

221
00:09:53.590 --> 00:09:55.220
take a "green day."

222
00:09:55.220 --> 00:09:55.830
...

223
00:09:55.830 --> 00:09:58.910
The zones are broken
up into colors,

224
00:09:58.910 --> 00:10:01.750
orange and red being
the anaerobic zones --

225
00:10:01.750 --> 00:10:04.810
zones 4 and 5 -- green
being your zone 3,

226
00:10:04.810 --> 00:10:06.460
blue being your zone 2.

227
00:10:06.460 --> 00:10:09.770
So we recommend, don't shoot
for those 12 to 20 minutes

228
00:10:09.770 --> 00:10:11.300
in the orange and red zone.

229
00:10:11.300 --> 00:10:13.170
Look at your heart
rate zones, listen

230
00:10:13.170 --> 00:10:15.800
to that real-time feedback
loop, and take a green day

231
00:10:15.800 --> 00:10:16.600
if you need to.

232
00:10:16.600 --> 00:10:19.230
So it's a bit of a
cop-out answer there,

233
00:10:19.230 --> 00:10:21.570
but we've got the coach,
we've got the member,

234
00:10:21.570 --> 00:10:23.250
and we've got a
real-time feedback loop.

235
00:10:23.250 --> 00:10:26.810
And when there's an intuitive
answer like that right in front

236
00:10:26.810 --> 00:10:30.750
of you, I don't think you should
interfere with an algorithm,

237
00:10:30.750 --> 00:10:32.668
is kind of the
thought process there.

238
00:10:32.668 --> 00:10:35.210
SAM RANSBOTHAM: I don't think
that's a cop-out answer at all,

239
00:10:35.210 --> 00:10:38.670
because historically,
you had a OK,

240
00:10:38.670 --> 00:10:40.140
you could have a
personal trainer,

241
00:10:40.140 --> 00:10:41.598
or you could be in
a group setting.

242
00:10:41.598 --> 00:10:44.060
I think I personally identify
[with] this because you could

243
00:10:44.060 --> 00:10:46.100
either teach someone
one-on-one as a tutor

244
00:10:46.100 --> 00:10:48.240
or you could teach someone
in a giant classroom.

245
00:10:48.240 --> 00:10:50.710
And I'm just excited
about applications

246
00:10:50.710 --> 00:10:55.230
like this that are letting us
pull together the best of both

247
00:10:55.230 --> 00:10:56.160
of those worlds.

248
00:10:56.160 --> 00:10:58.770
I'm hungry for it in
education, but I can see that

249
00:10:58.770 --> 00:11:00.520
you're making a lot
of progress on that --

250
00:11:00.520 --> 00:11:02.850
and Shervin and I have
talked with Peloton as well.

251
00:11:02.850 --> 00:11:04.800
They're starting
to think about how

252
00:11:04.800 --> 00:11:09.873
you get to an individual-level
experience at scale,

253
00:11:09.873 --> 00:11:12.040
and that seems like what
you're really trying to do.

254
00:11:12.040 --> 00:11:12.880
AMEEN KAZEROUNI: Absolutely.

255
00:11:12.880 --> 00:11:14.460
I couldn't have put
it better myself.

256
00:11:14.460 --> 00:11:15.550
I think that's exactly it.

257
00:11:15.550 --> 00:11:17.175
SHERVIN KHODABANDEH:
I like that a lot.

258
00:11:17.175 --> 00:11:19.150
This is all about
rapid feedback,

259
00:11:19.150 --> 00:11:21.830
which is a cornerstone
of building expertise

260
00:11:21.830 --> 00:11:25.580
in any system, whether it's
chess, or whether it's running,

261
00:11:25.580 --> 00:11:27.100
or whether it's
machine learning.

262
00:11:27.100 --> 00:11:29.240
The more you combine
experimentation

263
00:11:29.240 --> 00:11:32.530
and rapid feedback and
the human in the loop

264
00:11:32.530 --> 00:11:36.230
and apply that across
different industries, the more

265
00:11:36.230 --> 00:11:37.780
opportunity and
values you're going

266
00:11:37.780 --> 00:11:41.580
to unleash for personalized
everything, not just fitness

267
00:11:41.580 --> 00:11:46.200
or shopping for pants but also
education and everything else.

268
00:11:46.200 --> 00:11:47.803
You're on to something here.

269
00:11:47.803 --> 00:11:48.970
AMEEN KAZEROUNI: Absolutely.

270
00:11:48.970 --> 00:11:50.730
In my prior life,
I was in retail.

271
00:11:50.730 --> 00:11:53.550
And when you think about
the role of machine

272
00:11:53.550 --> 00:11:55.750
learning and deep
learning in retail,

273
00:11:55.750 --> 00:11:58.120
it's that: You're
trying to re-create

274
00:11:58.120 --> 00:12:02.410
the in-person shopping
experience on a website.

275
00:12:02.410 --> 00:12:04.240
Like, how do we curate this?

276
00:12:04.240 --> 00:12:06.050
How do we create a
personal shopper?

277
00:12:06.050 --> 00:12:07.980
How do we create that
boutique experience?

278
00:12:07.980 --> 00:12:09.900
How do we predict
the size correctly?

279
00:12:09.900 --> 00:12:11.400
How do we use AR
[augmented reality]

280
00:12:11.400 --> 00:12:13.820
so you can see what a shoe
looks like on your foot?

281
00:12:13.820 --> 00:12:18.070
We're always trying to close
that gap and arrive back

282
00:12:18.070 --> 00:12:21.030
at what the real thing would
be like -- personalized,

283
00:12:21.030 --> 00:12:21.960
but at scale.

284
00:12:21.960 --> 00:12:24.760
And I think that's the
secret sauce there:

285
00:12:24.760 --> 00:12:28.150
a rapid feedback loop,
algorithmic support

286
00:12:28.150 --> 00:12:31.190
with large volumes of
data, but then also

287
00:12:31.190 --> 00:12:34.643
not trying to circumnavigate
around the expert,

288
00:12:34.643 --> 00:12:35.560
the human in the loop.

289
00:12:35.560 --> 00:12:37.190
SHERVIN KHODABANDEH: Exactly.

290
00:12:37.190 --> 00:12:38.880
And I think that's
really critical, Sam,

291
00:12:38.880 --> 00:12:41.455
because I remember when we did
the first few years of the [AI

292
00:12:41.455 --> 00:12:43.122
report](
https://sloanreview.mit.edu/big

293
00:12:43.122 --> 00:12:45.164
ideas/artificial intelligence
business strategy),

294
00:12:45.164 --> 00:12:46.462
like back 2018-2019.

295
00:12:46.462 --> 00:12:47.330


296
00:12:47.330 --> 00:12:48.540
SAM RANSBOTHAM: So long ago.

297
00:12:48.540 --> 00:12:48.678


298
00:12:48.678 --> 00:12:50.970
SHERVIN KHODABANDEH: There
were still a lot of folks --

299
00:12:50.970 --> 00:12:52.440
I mean, that's only
three years ago,

300
00:12:52.440 --> 00:12:53.560
but a lot of folks still
today are thinking of AI

301
00:12:53.560 --> 00:12:56.920
as that which replaces human,
and that which must automate.

302
00:12:56.920 --> 00:13:01.610
And the more we think that
narrowly, the more outcomes

303
00:13:01.610 --> 00:13:04.770
we're leaving on the table,
the more successful workforces

304
00:13:04.770 --> 00:13:07.100
and humans we're
disenfranchising,

305
00:13:07.100 --> 00:13:10.900
and also the more
opportunities we're leaving

306
00:13:10.900 --> 00:13:13.060
unaddressed because we
think, "Well, there's

307
00:13:13.060 --> 00:13:15.420
no way I can fully
replace a human here,

308
00:13:15.420 --> 00:13:17.025
so I'm not going to do it."

309
00:13:17.025 --> 00:13:19.400
And there are so many of these
things where it's not just

310
00:13:19.400 --> 00:13:22.700
AI versus human but
it's AI and human.

311
00:13:22.700 --> 00:13:23.145


312
00:13:23.145 --> 00:13:24.020
AMEEN KAZEROUNI: Yep.

313
00:13:24.020 --> 00:13:28.790
You know, I think that there
is a conflation of companies

314
00:13:28.790 --> 00:13:30.650
where the product is AI.

315
00:13:30.650 --> 00:13:33.240
Like, when you think of
AI, you think of Tesla,

316
00:13:33.240 --> 00:13:35.630
and there's different
industries all of them play in,

317
00:13:35.630 --> 00:13:38.030
but there's a very core
part of the product that's

318
00:13:38.030 --> 00:13:40.758
a standalone piece of
intellectual property

319
00:13:40.758 --> 00:13:42.800
that's heavily rooted in
artificial intelligence,

320
00:13:42.800 --> 00:13:46.290
and that's not how a
majority of the world

321
00:13:46.290 --> 00:13:48.000
is going to use
artificial intelligence.

322
00:13:48.000 --> 00:13:51.785
And I think that's one of the
kind of key differences in what

323
00:13:51.785 --> 00:13:53.410
you were just talking
about, Shervin --

324
00:13:53.410 --> 00:13:57.600
is that people try and use AI
the way those companies use AI

325
00:13:57.600 --> 00:13:59.540
and see them as
the gold standard,

326
00:13:59.540 --> 00:14:01.730
but our product is not AI.

327
00:14:01.730 --> 00:14:07.440
Our product is a curated,
science-backed, coach-inspired

328
00:14:07.440 --> 00:14:11.090
fitness experience that's
just merely augmented

329
00:14:11.090 --> 00:14:12.750
in parts by AI.

330
00:14:12.750 --> 00:14:14.150
SAM RANSBOTHAM:
This is something

331
00:14:14.150 --> 00:14:15.550
that Shervin and I are
pretty excited about.

332
00:14:15.550 --> 00:14:17.050
I don't want to
foreshadow too much,

333
00:14:17.050 --> 00:14:21.150
but we're thinking
about these mini uses

334
00:14:21.150 --> 00:14:24.327
of AI beyond just this
kind of poster use of AI

335
00:14:24.327 --> 00:14:25.410
that you're talking about.

336
00:14:25.410 --> 00:14:28.430
I mean, yes, we
all are attracted

337
00:14:28.430 --> 00:14:31.610
to the Boston Dynamics
robots that look very cool,

338
00:14:31.610 --> 00:14:35.510
but there's a lot going
on that isn't that level,

339
00:14:35.510 --> 00:14:37.753
and there's so much
value in those.

340
00:14:37.753 --> 00:14:39.920
And I think you're starting
to capture some of that.

341
00:14:39.920 --> 00:14:43.030
SHERVIN KHODABANDEH: I
wanted to segue from this.

342
00:14:43.030 --> 00:14:47.200
We talked quite a lot about
the importance of human

343
00:14:47.200 --> 00:14:52.160
in the loop, experimentation,
being hypothesis-driven,

344
00:14:52.160 --> 00:14:53.380
all these things you said.

345
00:14:53.380 --> 00:14:58.850
Maybe tell us a bit around the
operating model and the ways

346
00:14:58.850 --> 00:15:04.150
of working in a company that
is not an AI product company

347
00:15:04.150 --> 00:15:08.140
but is a company like yours
with a strong mission.

348
00:15:08.140 --> 00:15:13.440
What does it take to take a
use case and bring it to life?

349
00:15:13.440 --> 00:15:15.150
AMEEN KAZEROUNI: I
think it honestly

350
00:15:15.150 --> 00:15:18.850
comes down to three things.

351
00:15:18.850 --> 00:15:21.530
It's having the data;
you can't really

352
00:15:21.530 --> 00:15:23.930
get around not having the data.

353
00:15:23.930 --> 00:15:25.290
Investment.

354
00:15:25.290 --> 00:15:27.540
I think a big mistake
companies make

355
00:15:27.540 --> 00:15:30.830
is not investing
in data engineers

356
00:15:30.830 --> 00:15:33.360
early, thinking that
you can just sprinkle

357
00:15:33.360 --> 00:15:36.510
AI like some kind of magic
powder on raw data sets

358
00:15:36.510 --> 00:15:38.280
and it's going to
produce something.

359
00:15:38.280 --> 00:15:42.717
I think data engineers are a
critical commodity that you

360
00:15:42.717 --> 00:15:45.050
want to invest in early so
your data is at a point where

361
00:15:45.050 --> 00:15:46.150
you can actually use it.

362
00:15:46.150 --> 00:15:48.190
So that investment
is really important.

363
00:15:48.190 --> 00:15:50.280
Wherever it's coming
from, there needs

364
00:15:50.280 --> 00:15:54.850
to be a serious decision made
to invest in your data practice

365
00:15:54.850 --> 00:15:56.900
if you're going really
try and build one.

366
00:15:56.900 --> 00:15:59.260
And then finally, it's buy-in.

367
00:15:59.260 --> 00:16:02.850
When the product is not AI,
you're convincing domain

368
00:16:02.850 --> 00:16:06.070
experts -- in our case,
fitness experts --

369
00:16:06.070 --> 00:16:07.840
that have been doing
this for a long time,

370
00:16:07.840 --> 00:16:10.710
that are formally educated
in these fields of study,

371
00:16:10.710 --> 00:16:14.770
that are always going to know
more about the product than you

372
00:16:14.770 --> 00:16:19.380
do, that an algorithm is going
to help them and make their job

373
00:16:19.380 --> 00:16:20.367
easier.

374
00:16:20.367 --> 00:16:21.950
And I think that
that relationship can

375
00:16:21.950 --> 00:16:25.370
be a beautiful partnership
or it can be an extremely

376
00:16:25.370 --> 00:16:26.660
antagonistic one.

377
00:16:26.660 --> 00:16:28.600
One of the things that
I've kind of strived

378
00:16:28.600 --> 00:16:30.840
toward in my role
at Orangetheory

379
00:16:30.840 --> 00:16:34.998
is to have a strong partnership
with our template design

380
00:16:34.998 --> 00:16:37.540
team, our workout design team,
because at the end of the day,

381
00:16:37.540 --> 00:16:40.570
they are the protectors and
designers of that product,

382
00:16:40.570 --> 00:16:43.020
and we're, again, just a
tool that's supporting them.

383
00:16:43.020 --> 00:16:46.750
Their buy-in is very critical,
because their understanding

384
00:16:46.750 --> 00:16:50.970
of the algorithms is what then
makes it to coaches in learning

385
00:16:50.970 --> 00:16:52.650
and development material.

386
00:16:52.650 --> 00:16:56.060
At the end of the day, we've
got thousands of coaches

387
00:16:56.060 --> 00:17:00.280
across 24 countries explaining
that max heart rate algorithm

388
00:17:00.280 --> 00:17:01.320
that I mentioned.

389
00:17:01.320 --> 00:17:04.050
And they're not explaining it
like a piece of mathematics;

390
00:17:04.050 --> 00:17:07.130
they're explaining it like
a piece of exercise fitness,

391
00:17:07.130 --> 00:17:08.869
like a piece of
exercise physiology.

392
00:17:08.869 --> 00:17:11.040
And that requires that buy-in.

393
00:17:11.040 --> 00:17:13.140
Like, the AI, the data
team, and the fitness team

394
00:17:13.140 --> 00:17:15.829
need to be in lockstep;
otherwise, it's

395
00:17:15.829 --> 00:17:16.552
destined to fail.

396
00:17:16.552 --> 00:17:18.760
SHERVIN KHODABANDEH: And
what does that mean in terms

397
00:17:18.760 --> 00:17:22.050
of the talent and the team --
the technical team that that

398
00:17:22.050 --> 00:17:24.160
you oversee and you
hire and recruit?

399
00:17:24.160 --> 00:17:27.119
AMEEN KAZEROUNI: That
they're difficult to find,

400
00:17:27.119 --> 00:17:28.440
is what it means.

401
00:17:28.440 --> 00:17:31.250
I think there's
already a scarcity

402
00:17:31.250 --> 00:17:32.820
of talent in this space.

403
00:17:32.820 --> 00:17:37.490
I think in a mission-driven,
purpose-driven company

404
00:17:37.490 --> 00:17:41.360
like Orangetheory, you'd
think it's harder to find,

405
00:17:41.360 --> 00:17:44.320
but it's actually
easier in the sense

406
00:17:44.320 --> 00:17:47.450
that if you find someone that's
aligned with the mission,

407
00:17:47.450 --> 00:17:51.380
it's almost exciting to them
that there's an opportunity

408
00:17:51.380 --> 00:17:54.030
to apply that skill set
on what they considered

409
00:17:54.030 --> 00:17:57.500
an outside-the-job passion.

410
00:17:57.500 --> 00:18:01.450
But we've also focused
on our data organization

411
00:18:01.450 --> 00:18:03.000
being a separate entity.

412
00:18:03.000 --> 00:18:06.430
So we've got my role, chief
data and analytics officer,

413
00:18:06.430 --> 00:18:09.720
running a data organization
reporting in to our CEO.

414
00:18:09.720 --> 00:18:12.660
And we've got our chief
digital and technology officer

415
00:18:12.660 --> 00:18:15.510
running a separate digital
and technology organization.

416
00:18:15.510 --> 00:18:17.510
And what's really
powerful about that

417
00:18:17.510 --> 00:18:20.330
is that we're able to
riff off of each other

418
00:18:20.330 --> 00:18:23.060
and have one team
provide building

419
00:18:23.060 --> 00:18:26.010
blocks to the other
team and vice versa.

420
00:18:26.010 --> 00:18:31.130
And what you'd imagine
creates an interesting working

421
00:18:31.130 --> 00:18:34.180
experience actually
drives a lot of velocity

422
00:18:34.180 --> 00:18:36.720
and drives a really
cool partnership that's

423
00:18:36.720 --> 00:18:38.650
very exciting to be
a part of as well.

424
00:18:38.650 --> 00:18:42.050
So I think it's all about
that: partnerships and buy-ins

425
00:18:42.050 --> 00:18:43.760
and collaboration across teams.

426
00:18:43.760 --> 00:18:44.140


427
00:18:44.140 --> 00:18:45.640
SHERVIN KHODABANDEH:
Very well said.

428
00:18:45.640 --> 00:18:47.670
Ameen, we have a
special segment here

429
00:18:47.670 --> 00:18:53.220
where we ask you five
rapid-fire-style questions.

430
00:18:53.220 --> 00:18:55.880
Just give us the first thing
that comes to your mind.

431
00:18:55.880 --> 00:18:58.650
The key thing is
intuitive, whatever

432
00:18:58.650 --> 00:19:01.300
comes your mind, and
short and sweet answers.

433
00:19:01.300 --> 00:19:02.132
So, ready for that?

434
00:19:02.132 --> 00:19:03.340
AMEEN KAZEROUNI: Let's do it.

435
00:19:03.340 --> 00:19:03.970
SHERVIN KHODABANDEH: All right.

436
00:19:03.970 --> 00:19:05.800
What's your proudest AI moment?

437
00:19:05.800 --> 00:19:10.440
AMEEN KAZEROUNI: When we solved
it using a linear regression.

438
00:19:10.440 --> 00:19:12.030
SHERVIN KHODABANDEH: Love it.

439
00:19:12.030 --> 00:19:12.900
All right.

440
00:19:12.900 --> 00:19:15.150
What worries you about AI?

441
00:19:15.150 --> 00:19:19.090
AMEEN KAZEROUNI: A
lack of consultation

442
00:19:19.090 --> 00:19:19.960
with domain experts.

443
00:19:19.960 --> 00:19:20.398


444
00:19:20.398 --> 00:19:21.690
SHERVIN KHODABANDEH: Well said.

445
00:19:21.690 --> 00:19:26.160
Your favorite activity that
involves no technology?

446
00:19:26.160 --> 00:19:28.582
AMEEN KAZEROUNI: Hiking.

447
00:19:28.582 --> 00:19:29.312


448
00:19:29.312 --> 00:19:31.020
SHERVIN KHODABANDEH:
Was that a question?

449
00:19:31.020 --> 00:19:32.670
AMEEN KAZEROUNI: It's
a very simple one.

450
00:19:32.670 --> 00:19:34.878
I was going to say Orangetheory,
but there's too much

451
00:19:34.878 --> 00:19:35.720
technology in there.

452
00:19:35.720 --> 00:19:39.615
SHERVIN KHODABANDEH: The first
career you wanted -- like,

453
00:19:39.615 --> 00:19:41.240
what you wanted to
be when you grew up.

454
00:19:41.240 --> 00:19:41.896


455
00:19:41.896 --> 00:19:43.604
AMEEN KAZEROUNI:
Environmental biologist.

456
00:19:43.604 --> 00:19:44.380


457
00:19:44.380 --> 00:19:46.120
SHERVIN KHODABANDEH:
Your greatest wish

458
00:19:46.120 --> 00:19:47.680
for AI in the future?

459
00:19:47.680 --> 00:19:48.930
AMEEN KAZEROUNI: More access.

460
00:19:48.930 --> 00:19:48.965


461
00:19:48.965 --> 00:19:50.090
SHERVIN KHODABANDEH: Great.

462
00:19:50.090 --> 00:19:51.670
SAM RANSBOTHAM: Actually,
I've got to follow up:

463
00:19:51.670 --> 00:19:52.300
Access for who?

464
00:19:52.300 --> 00:19:53.008
Who needs access?

465
00:19:53.008 --> 00:19:55.880
AMEEN KAZEROUNI: I
think it would be great

466
00:19:55.880 --> 00:20:01.260
if some of the simpler portions
of AI that unlocked decisioning

467
00:20:01.260 --> 00:20:03.230
off of data that
companies have collected

468
00:20:03.230 --> 00:20:07.220
was easier to tap into without
the financial and human capital

469
00:20:07.220 --> 00:20:09.430
that you're required to
invest as an organization.

470
00:20:09.430 --> 00:20:14.770
I think the general efficiency
of the world will just go up,

471
00:20:14.770 --> 00:20:15.430
you know.

472
00:20:15.430 --> 00:20:17.733
SHERVIN KHODABANDEH: More
open-source kind of stuff.

473
00:20:17.733 --> 00:20:18.900
AMEEN KAZEROUNI: Yeah, yeah.

474
00:20:18.900 --> 00:20:22.100
SHERVIN KHODABANDEH: Ameen this
has been exceedingly insightful

475
00:20:22.100 --> 00:20:23.835
and a lot of fun.

476
00:20:23.835 --> 00:20:25.210
Thank you for
making time for us.

477
00:20:25.210 --> 00:20:26.360
SAM RANSBOTHAM: Yeah,
thanks for coming.

478
00:20:26.360 --> 00:20:27.390
AMEEN KAZEROUNI:
Thanks for having me.

479
00:20:27.390 --> 00:20:28.307
This was a lot of fun.

480
00:20:28.307 --> 00:20:29.300
I truly enjoyed it.

481
00:20:29.300 --> 00:20:29.800


482
00:20:29.800 --> 00:20:31.750
SAM RANSBOTHAM: On
our next episode,

483
00:20:31.750 --> 00:20:34.480
we'll speak with Khatereh
Khodavirdi, senior director

484
00:20:34.480 --> 00:20:36.480
of data science and
analytics at PayPal,

485
00:20:36.480 --> 00:20:39.120
where she oversees data
teams at Venmo and Honey.

486
00:20:39.120 --> 00:20:39.820
Please join us.

487
00:20:39.820 --> 00:20:39.892


488
00:20:39.892 --> 00:20:41.350
ALLISON RYDER:
Thanks for listening

489
00:20:41.350 --> 00:20:42.850
to Me, Myself, and AI.

490
00:20:42.850 --> 00:20:45.310
We believe, like you,
that the conversation

491
00:20:45.310 --> 00:20:47.530
about AI implementation
doesn't start and stop

492
00:20:47.530 --> 00:20:48.647
with this podcast.

493
00:20:48.647 --> 00:20:50.480
That's why we've created
a group on LinkedIn

494
00:20:50.480 --> 00:20:52.320
specifically for
listeners like you.

495
00:20:52.320 --> 00:20:55.070
It's called AI for Leaders,
and if you join us,

496
00:20:55.070 --> 00:20:57.090
you can chat with show
creators and hosts,

497
00:20:57.090 --> 00:20:59.800
ask your own questions,
share your insights,

498
00:20:59.800 --> 00:21:02.440
and gain access to
valuable resources about AI

499
00:21:02.440 --> 00:21:05.280
implementation from
MIT SMR and BCG.

500
00:21:05.280 --> 00:21:10.410
You can access it by visiting
mitsmr.com/AIforLeaders.

501
00:21:10.410 --> 00:21:13.130
We'll put that link
in the show notes,

502
00:21:13.130 --> 00:21:15.560
and we hope to see you there.

503
00:21:15.560 --> 00:21:21.000