Speaker: 00:00:14

You are listening to SmarterArticles, long

form writing on technology, governance,

2

: 00:00:19

and the human cost of the things we build.

3

: 00:00:22

This week's article is AI

Integration: Transforming

4

: 00:00:25

Workplaces and Employee Futures.

5

: 00:00:28

Somewhere inside Amazon's sprawling

corporate operation, a system

6

: 00:00:32

called Clarity is watching,

not in the cinematic sense, no

7

: 00:00:36

blinking red eye, no ominous hum.

8

: 00:00:39

It watches in the way that modern

surveillance actually works quietly

9

: 00:00:44

through spreadsheets, through

dashboards, through the patient

10

: 00:00:49

accumulation of data points that

taken together constituted judgment.

11

: 00:00:54

Clarity tracks which AI tools Amazon's

developers use, how often they use

12

: 00:00:59

them, and whether they are hitting

the company's internal benchmark.

13

: 00:01:02

80% of developers engaging with AI

assisted coding at least once per week.

14

: 00:01:08

Managers can see exactly who meets

that threshold and who does not.

15

: 00:01:13

That data feeds directly into performance

reviews, promotion evaluations,

16

: 00:01:18

career trajectory conversations.

17

: 00:01:20

At Amazon, your relationship with

artificial intelligence is no longer

18

: 00:01:24

a matter of personal curiosity

or professional temperament.

19

: 00:01:28

It is a metric.

20

: 00:01:29

It is scored and it is beginning

to determine whether you move up,

21

: 00:01:34

stay put, or find yourself on a

performance improvement plan with

22

: 00:01:38

the exits quietly illuminated.

23

: 00:01:41

Amazon, it should be said

immediately, is not alone.

24

: 00:01:44

Across the corporate world from

Silicon Valley to the big four

25

: 00:01:48

consulting firms, a new orthodoxy is

taking hold with remarkable speed.

26

: 00:01:53

AI proficiency is no longer optional.

27

: 00:01:56

It is the new literacy,

the new typing speed.

28

: 00:02:00

The new must be proficient

in Microsoft Office.

29

: 00:02:05

Except that this time the surveillance

is more granular, the enforcement

30

: 00:02:10

more rapid, and the consequences for

non-compliance considerably sharper.

31

: 00:02:15

What is being built piece

by piece quarterly review by

32

: 00:02:20

quarterly review is something

that deserves to be named plainly.

33

: 00:02:25

It is a compliance regime

dressed up as innovation.

34

: 00:02:29

The companies leading this charge read

like an index of global corporate power.

35

: 00:02:34

At Amazon, the performance review

system known as Forte now integrates

36

: 00:02:38

self-reported accomplishments

with peer and supervisor feedback,

37

: 00:02:42

producing an overall value score that

shapes raises and promotions, and

38

: 00:02:46

at the darker end, the possibility

of a performance improvement plan.

39

: 00:02:51

Within the supply chain optimization

technologies team, AI adoption is

40

: 00:02:55

a required evaluation category.

41

: 00:02:58

Employees are asked how they

used AI to drive innovation,

42

: 00:03:01

improve operational efficiency.

43

: 00:03:03

Enhance customer experience.

44

: 00:03:06

Managers face even tougher scrutiny.

45

: 00:03:09

They must show concrete evidence

of boosting results through

46

: 00:03:12

AI without adding headcount.

47

: 00:03:14

The message embedded in that

requirement is not subtle.

48

: 00:03:18

The machine is not a

supplement to your team.

49

: 00:03:22

The machine is part of the

case against expanding it.

50

: 00:03:25

Meta followed.

51

: 00:03:26

From 2026, AI driven impact became

a core expectation baked into every

52

: 00:03:32

employee's performance review.

53

: 00:03:34

Regardless of role.

54

: 00:03:36

Engineers, marketers, product

managers, designers, all

55

: 00:03:42

evaluated on how effectively

they use AI to deliver results.

56

: 00:03:46

The company's biannual review platform

now reassess performance twice yearly with

57

: 00:03:52

AI driven impact woven into each cycle.

58

: 00:03:55

Meta has even gamified the transition,

rewarding employees with badges as they

59

: 00:04:00

hit milestones in AI tool experimentation.

60

: 00:04:04

Tracking progress through dashboards that

visualize adoption rates across teams.

61

: 00:04:08

Some employees have already begun

using the company's internal

62

: 00:04:12

AI assistant to draft the very

content submitted in their reviews.

63

: 00:04:16

A recursive loop that feels distinctly

like the future consuming its own premise.

64

: 00:04:21

Then again, perhaps that is the point.

65

: 00:04:25

Accenture took arguably

the most direct approach.

66

: 00:04:28

The company began collecting data

on weekly logins to its AI platforms

67

: 00:04:32

from senior staff, and sent an

internal communication to managers and

68

: 00:04:36

associate directors, making it clear.

69

: 00:04:39

Moving into leadership requires regular

adoption of artificial intelligence.

70

: 00:04:44

The CEO warned publicly last year

that the company would be exiting

71

: 00:04:48

staffers who could not be retrained.

72

: 00:04:51

KPMG followed announcing that from

: 2026

73

: 00:04:56

how well they have met AI objectives

during annual reviews with monitoring,

74

: 00:05:01

extending across the entire organization

from junior staff to senior partners.

75

: 00:05:07

Microsoft, the company that

arguably did more than any other

76

: 00:05:11

to mainstream generative ai, turned

the lens inward in June of last

77

: 00:05:15

year and told its own employees

that using AI is no longer optional.

78

: 00:05:21

The company that built copilot expects its

workforce to comply or face consequences.

79

: 00:05:26

Every other company in the world

is watching and taking notes.

80

: 00:05:30

What gives these mandates, their

particular force is the data backdrop

81

: 00:05:34

against which they're issued.

82

: 00:05:36

Gallup's most recent workforce survey

found that only 26% of American

83

: 00:05:40

workers use AI at least a few times

per week, while nearly half report

84

: 00:05:45

never using it in their role at all.

85

: 00:05:48

Daily usage sits at 12%.

86

: 00:05:51

Only 9% of employees describe themselves

as very comfortable with AI tools,

87

: 00:05:56

tools, and just a quarter said their

employer had clearly communicated how

88

: 00:06:00

AI is supposed to be used in their work.

89

: 00:06:03

The technology sector leads adoption

as one would expect, but retail

90

: 00:06:07

languishes at 33% total use.

91

: 00:06:11

Leadership skews the numbers further,

69% of leaders report using AI at

92

: 00:06:16

least several times a year compared

with 40% of individual contributors.

93

: 00:06:22

The people most likely to set adoption

mandates are also the people most

94

: 00:06:26

likely to already be using AI.

95

: 00:06:29

The perception gap that creates

colours every policy they issue.

96

: 00:06:33

McKinsey's research adds

another layer of complexity.

97

: 00:06:37

C-Suite executives consistently

underestimate actual employee AI usage

98

: 00:06:42

while simultaneously believing the

pace of adoption is not fast enough.

99

: 00:06:48

The report's most pointed finding was

that the biggest barrier to scaling AI

100

: 00:06:52

is not employees, who the data suggests,

are broadly ready and often curious.

101

: 00:06:59

But Leaders who are not providing

adequate support, training or direction.

102

: 00:07:05

The people writing the scorecards have

misdiagnosed who needs remediation, and

103

: 00:07:11

yet the scorecards keep multiplying.

104

: 00:07:14

Microsoft's research on what it

calls frontier firms - organizations

105

: 00:07:18

with comprehensive AI deployment

and genuinely mature integration,

106

: 00:07:23

does paint an optimistic picture.

107

: 00:07:25

At such companies, workers report

higher satisfaction, less fear

108

: 00:07:29

of job displacement, and greater

capacity for meaningful work.

109

: 00:07:33

71% say their company is thriving.

110

: 00:07:37

82% of leaders globally believe this

is a pivotal year to rethink strategy.

111

: 00:07:43

PWCs analysis of close to a billion job

advertisements found that productivity

112

: 00:07:48

growth in industry most exposed to

I had nearly quadrupled since: 2022

113

: 00:07:53

And that jobs requiring AI skills now

command a wage premium of 56% on average.

114

: 00:07:59

These are not trivial numbers.

115

: 00:08:01

The distinction that matters and that

the aggregate data consistently obscures

116

: 00:08:06

is the difference between organisations

that have genuinely invested in building

117

: 00:08:10

AI capability, and organisations that

have simply added an AI usage checkbox

118

: 00:08:14

to the performance review form.

119

: 00:08:16

The former involves training,

clear communication, appropriate

120

: 00:08:20

tooling, management support.

121

: 00:08:23

The latter involves

dashboards and consequences.

122

: 00:08:27

McKinsey found that 48% of

employees rank training as the most

123

: 00:08:31

important factor for AI adoption,

yet nearly half reported receiving

124

: 00:08:35

minimal or no training whatsoever.

125

: 00:08:38

Only 30% said their manager

provides support for AI usage.

126

: 00:08:43

Companies are, in other words, building

scoring systems for a competency

127

: 00:08:48

they have not adequately taught.

128

: 00:08:51

This is not a transition

program, it's an audit.

129

: 00:08:55

The timing at Amazon sharpens

the point considerably.

130

: 00:08:59

The company's intensified aI

monitoring coincided with its largest

131

: 00:09:03

workforce reduction in 30 years.

132

: 00:09:06

Approximately 30,000 corporate

positions eliminated over a few months.

133

: 00:09:11

Nearly 10% of its entire

corporate and technical headcount.

134

: 00:09:16

CEO Andy Jassy was

transparent about the logic.

135

: 00:09:20

Efficiency gains from AI would

likely cause Amazon's corporate

136

: 00:09:24

headcount to fall in the coming years.

137

: 00:09:27

Meanwhile, the company announced

capital expenditures expected

138

: 00:09:30

to reach $125 billion for 2026.

139

: 00:09:34

Much of it directed

toward AI infrastructure.

140

: 00:09:37

When AI usage becomes a performance metric

in the wake of mass redundancies, the

141

: 00:09:42

implicit message is impossible to misread.

142

: 00:09:45

Prove you can work with the

machine or you may be next.

143

: 00:09:48

Amazon's broader surveillance

infrastructure provides the context for

144

: 00:09:52

understanding what clarity actually is.

145

: 00:09:55

The company already operates a

manager dashboard aggregating

146

: 00:09:58

attendance frequency, time

spent in the office and building

147

: 00:10:01

locations over eight week periods.

148

: 00:10:04

Employees averaging less than four hours

of daily office time are labeled in the

149

: 00:10:09

systems vocabulary, low time badges.

150

: 00:10:13

Those with no building access

records are zero badges.

151

: 00:10:17

On the warehouse side.

152

: 00:10:19

A system called adapt monitors each

worker's productivity in real time

153

: 00:10:23

tracking gaps in activity, and issuing

automatic termination warnings for

154

: 00:10:28

unexplained breaks of two hours or longer.

155

: 00:10:31

Clarity then is not an isolated

experiment in measuring AI engagement.

156

: 00:10:36

It is the natural next extension

of a culture that has long believed

157

: 00:10:40

in the power of measurement and

longer still in the management

158

: 00:10:44

of human behavior through it.

159

: 00:10:46

The research on workplace surveillance

is unambiguous about what extensive

160

: 00:10:50

monitoring does to people.

161

: 00:10:52

A study published in plus one examining

AI driven productivity scoring,

162

: 00:10:57

found that pervasive tracking reduces

worker autonomy, increases stress, and

163

: 00:11:02

raises the risk of psychological harm.

164

: 00:11:05

The authors noted that surveillance

disciplines workers to conform to

165

: 00:11:08

expected behaviors which can be

measured, and that when autonomy

166

: 00:11:12

and agency are reduced, so is the

capacity for creativity more pointedly.

167

: 00:11:18

Organizations send a message simply

by the tasks they choose to monitor.

168

: 00:11:23

When you monitor whether someone

has logged into an AI platform,

169

: 00:11:26

rather than whether the work they

have produced is excellent, you have

170

: 00:11:30

revealed what you actually value.

171

: 00:11:33

The European Trade Union Confederation

has been equally direct warning that

172

: 00:11:37

AI driven automation risks causing

job displacement, de-skilling, and

173

: 00:11:42

precarious employment without appropriate

regulation and calling for collective

174

: 00:11:46

bargaining rights over AI deployment.

175

: 00:11:49

The UC Berkeley Labor Center's

Research noted a growing number

176

: 00:11:53

of unfair labor practice charges

as algorithmic management spreads.

177

: 00:11:57

And warns that the almost complete

absence of regulation creates strong

178

: 00:12:02

incentives for employers to deploy

digital surveillance tools in ways

179

: 00:12:06

that demonstrably harm workers.

180

: 00:12:08

This is the deeper philosophical

contradiction in the AI scoring movement.

181

: 00:12:13

The entire premise of workplace AI is

augmentation, freeing people to do more

182

: 00:12:18

creative, strategic, and meaningful work

by relieving them of repetitive tasks.

183

: 00:12:24

But when AI usage itself

becomes the metric.

184

: 00:12:27

The tool ceases to be a means to

an end and becomes the end itself.

185

: 00:12:32

A developer who writes elegant, efficient

code without AI assistance is rated lower

186

: 00:12:37

under these systems than a developer who

produces mediocre work while dutifully

187

: 00:12:42

clicking through an AI dashboard.

188

: 00:12:44

The measurement is not of quality,

creativity, or even productivity

189

: 00:12:49

in any meaningful sense.

190

: 00:12:51

It is a measurement of obedience.

191

: 00:12:54

The confidence dimension

is not trivial either.

192

: 00:12:57

Survey data shows that older workers,

those with the most institutional

193

: 00:13:01

knowledge and experience, are

showing significant drops in

194

: 00:13:04

AI confidence with baby boomer

confidence reportedly falling 35%.

195

: 00:13:11

If training is inadequate.

196

: 00:13:12

If tooling is imperfect, and if

AI proficiency is nonetheless the

197

: 00:13:17

criterion for advancement, then the

risks of compounding existing workplace

198

: 00:13:22

inequalities are considerable.

199

: 00:13:24

PWCs research found that in every

country analyzed more women than

200

: 00:13:29

men work in AI exposed roles,

suggesting the skills pressure will

201

: 00:13:33

fall disproportionately on them.

202

: 00:13:35

And job cuts have been most pronounced

in larger corporations affecting

203

: 00:13:39

mostly entry-level employees.

204

: 00:13:41

The workers least likely to

have prior technical advantages.

205

: 00:13:45

The workers for whom a missed

training session represents a

206

: 00:13:47

structural disadvantage rather

than a minor inconvenience.

207

: 00:13:52

The legislative response is beginning,

though it remains fragmented and

208

: 00:13:56

considerably behind the corporate reality.

209

: 00:13:59

In the United States, state level

proposals are multiplying in California,

210

: 00:14:04

Michigan, Rhode Island and New York.

211

: 00:14:07

At the federal level, bipartisan

legislation has been introduced

212

: 00:14:11

that would require certain companies

to report on personnel decisions

213

: 00:14:14

affected by AI and prohibit employers

from relying solely on automated

214

: 00:14:19

systems for employment decisions.

215

: 00:14:21

In Europe.

216

: 00:14:22

The ETUC condemned the European

Commissions Withdrawal of the AI

217

: 00:14:26

Liability Directive warning that

without clear liability rules, workers

218

: 00:14:31

affected by AI driven decisions

will struggle to seek redress.

219

: 00:14:36

Notably Accenture staff in 12 European

countries are currently exempt from

220

: 00:14:40

the policy of factoring AI usage

into promotions, as are employees

221

: 00:14:45

on certain government contracts.

222

: 00:14:47

The geographical variation makes

the underlying principle visible.

223

: 00:14:52

Your vulnerability to this system depends

in part on where you happen to work.

224

: 00:14:56

Then there is the question of

whether any of it actually works.

225

: 00:15:01

Over 80% of companies in a survey

of 6,000 executives reported no

226

: 00:15:07

measurable productivity gains from AI.

227

: 00:15:09

Despite billions in investment,

McKinsey found that only 1% of

228

: 00:15:14

leaders describe their companies

as mature in AI deployment.

229

: 00:15:19

Meaning AI is genuinely

integrated into workflows and

230

: 00:15:22

producing substantial outcomes.

231

: 00:15:25

92% of companies plan to

increase AI investment.

232

: 00:15:29

So only 1% have made it mean anything.

233

: 00:15:32

The gap between ambition and achievement

is not a detail, it is the story.

234

: 00:15:39

Amazon's own experience with Kiro, its

internal agentic coding tool illustrates

235

: 00:15:43

the problem with a certain dark precision.

236

: 00:15:46

In December last year, engineers allowed

Kiro to make changes that sparked a 13

237

: 00:15:51

hour disruption to Amazon Web services.

238

: 00:15:54

The tool had decided on its own initiative

to delete and recreate the environment.

239

: 00:16:00

It was the second AI caused

incident in a matter of months.

240

: 00:16:04

The company continues to push

developers toward Kiro and away from

241

: 00:16:08

potentially more capable external tools.

242

: 00:16:10

Reportedly because internal staff are

encouraged to rely on company developed

243

: 00:16:14

systems, particularly when AI usage

metrics influence performance reviews.

244

: 00:16:20

Some developers have described the

tool in terms that do not flatter it.

245

: 00:16:24

And yet under the Clarity system,

using it dutifully is better for

246

: 00:16:29

your career than not using it at all.

247

: 00:16:32

The metric is serving corporate

strategy, not employee effectiveness.

248

: 00:16:37

If AI tools are genuinely

useful, employees will

249

: 00:16:40

adopt them without coercion.

250

: 00:16:42

Useful tools tend to spread.

251

: 00:16:44

They spread in workplaces where people

talk to each other, where someone shows a

252

: 00:16:48

colleague a thing that saved them an hour.

253

: 00:16:51

Where the value is self-evident.

254

: 00:16:53

If tools are not yet useful enough

to drive voluntary adoption,

255

: 00:16:56

forcing employees to use them and

then grading them on frequency

256

: 00:17:00

does not make the tools better.

257

: 00:17:01

It simply produces a compliance regime,

one that rewards those with prior

258

: 00:17:06

advantages and punishes those without

one that measures the appearance of

259

: 00:17:11

engagement rather than its substance.

260

: 00:17:14

One that generates the data leaders want

to see without necessarily generating

261

: 00:17:18

the outcomes they claim to be pursuing.

262

: 00:17:21

The Gallup data sits at the end of

all this, like an unignorable fact.

263

: 00:17:26

Nearly half of American workers

have never used AI in their jobs.

264

: 00:17:30

Nearly half report receiving minimal

or no training, and the companies

265

: 00:17:36

at the top of the global economy

are tying promotions, bonuses, and

266

: 00:17:40

job security to AI adoption scores.

267

: 00:17:43

The gap between expectation

and preparation is not a

268

: 00:17:47

transitional inconvenience.

269

: 00:17:49

It is the defining feature of this moment.

270

: 00:17:52

The machines are not coming for your job,

but the scorecard tracking how dutifully

271

: 00:17:58

you collaborate with them just might.

272

: 00:18:15

You've been listening to Smarter Articles.

273

: 00:18:18

The article you just heard was first

published at smarterarticles.co.uk,

274

: 00:18:22

where you'll find our full

archive, a new article every day.

275

: 00:18:26

Thanks for listening.

276

: 00:18:27

Subscribe wherever you get your podcasts.

277

: 00:18:29

Share with someone who thinks carefully

and we'll meet here again next week.