WEBVTT

1
00:00:00.440 --> 00:00:04.879
<v Speaker 1>Welcome to AI assembled the top AI news for small business.

2
00:00:05.360 --> 00:00:08.119
<v Speaker 1>I'm Chris daily and each week I bring you the

3
00:00:08.160 --> 00:00:13.519
<v Speaker 1>most important AI developments that can transform your business. Let's

4
00:00:13.560 --> 00:00:15.439
<v Speaker 1>dive into this week's top stories.

5
00:00:17.079 --> 00:00:21.359
<v Speaker 2>Apple has sued open Ai and federal court in Northern California,

6
00:00:21.839 --> 00:00:25.160
<v Speaker 2>accusing the AI lab of stealing trade secrets to build

7
00:00:25.199 --> 00:00:29.640
<v Speaker 2>its own consumer hardware. This is a dramatic reversal for

8
00:00:29.679 --> 00:00:32.679
<v Speaker 2>two companies that were high profile partners just two years

9
00:00:32.679 --> 00:00:37.840
<v Speaker 2>ago when Chat GPT was baked into the iPhone's operating system.

10
00:00:38.399 --> 00:00:43.240
<v Speaker 2>Apple's complaint pulls no punches. It alleges that at every level,

11
00:00:43.600 --> 00:00:47.119
<v Speaker 2>from technical staff up to open AI's chief hardware officer,

12
00:00:47.600 --> 00:00:51.439
<v Speaker 2>the company has been quote stealing Apple's trade secrets and

13
00:00:51.560 --> 00:00:57.439
<v Speaker 2>confidential information end quote. That chief hardware officer, Tang Tan,

14
00:00:58.119 --> 00:01:01.119
<v Speaker 2>is a former Apple vice president, and he is now

15
00:01:01.159 --> 00:01:06.040
<v Speaker 2>personally named as a defendant. Apple says Tan directed job

16
00:01:06.079 --> 00:01:10.640
<v Speaker 2>candidates still working at Apple to bring quote actual parts

17
00:01:10.959 --> 00:01:14.319
<v Speaker 2>end quote from Apple to their interviews for what he

18
00:01:14.400 --> 00:01:17.840
<v Speaker 2>called show and tell sessions so his team at open

19
00:01:17.879 --> 00:01:22.840
<v Speaker 2>Ai could pull out more confidential information. Apple also alleges

20
00:01:23.000 --> 00:01:27.079
<v Speaker 2>open Ai coached departing Apple employees on how to evade

21
00:01:27.120 --> 00:01:30.959
<v Speaker 2>Apple's security processes on their way out the door, and

22
00:01:31.040 --> 00:01:35.079
<v Speaker 2>that one former employee, Chang Liu, walked off with an

23
00:01:35.079 --> 00:01:40.319
<v Speaker 2>Apple laptop. Liu is also named as a defendant, and

24
00:01:40.439 --> 00:01:44.239
<v Speaker 2>Apple claims open ai has been asking outside hardware firms

25
00:01:44.519 --> 00:01:48.480
<v Speaker 2>to use a metal finishing technique Apple invented, while misleading

26
00:01:48.519 --> 00:01:53.519
<v Speaker 2>those partners into believing they had Apple's permission. The backdrop

27
00:01:53.560 --> 00:01:58.040
<v Speaker 2>here is open AI's hardware push. Last year, open ai

28
00:01:58.200 --> 00:02:01.920
<v Speaker 2>paid six point four billion dollar for former Apple designer

29
00:02:02.000 --> 00:02:06.560
<v Speaker 2>Johnny Ives. Device startup and CEO Sam Altman has said

30
00:02:06.560 --> 00:02:10.759
<v Speaker 2>the first prototypes are finished. Apple is asking the court

31
00:02:10.840 --> 00:02:15.439
<v Speaker 2>for damages, injunctions, and an order forcing OpenAI to stop

32
00:02:15.560 --> 00:02:19.800
<v Speaker 2>using its trade secrets. Why it matters for small businesses.

33
00:02:20.840 --> 00:02:23.080
<v Speaker 2>If you build any part of your workflow on the

34
00:02:23.120 --> 00:02:26.879
<v Speaker 2>assumption that Apple and open Ai would keep playing nicely together,

35
00:02:27.439 --> 00:02:31.800
<v Speaker 2>that assumption is now shaky. The updated SERI coming this

36
00:02:31.960 --> 00:02:35.800
<v Speaker 2>fall is already switching to Google's Gemini models instead of

37
00:02:35.840 --> 00:02:41.479
<v Speaker 2>open Ai. Expect messier integrations between Apple devices and chat

38
00:02:41.520 --> 00:02:46.879
<v Speaker 2>GPT going forward, and expect OpenAI's promised hardware to arrive

39
00:02:47.039 --> 00:02:51.000
<v Speaker 2>under a legal cloud that could delay ship dates. If

40
00:02:51.000 --> 00:02:55.759
<v Speaker 2>you are an SMB running on Apple Intelligence plus Chat GPT.

41
00:02:56.560 --> 00:02:59.520
<v Speaker 2>Do not bet your Q four roadmap on that pairing,

42
00:02:59.560 --> 00:03:05.199
<v Speaker 2>staying f frictionless. Now moving on, Microsoft has just made

43
00:03:05.240 --> 00:03:07.680
<v Speaker 2>it a lot cheaper to skill up your team on

44
00:03:07.759 --> 00:03:11.759
<v Speaker 2>the single hottest topic in enterprise AI right now, which

45
00:03:11.840 --> 00:03:16.759
<v Speaker 2>is AI agents. Those are AI systems that take actions

46
00:03:16.800 --> 00:03:21.639
<v Speaker 2>on your behalf, not just answer questions. Microsoft has released

47
00:03:21.680 --> 00:03:26.319
<v Speaker 2>a completely free ten lesson course called AI Agents for

48
00:03:26.400 --> 00:03:31.400
<v Speaker 2>beginners on Microsoft Learn and an expanded companion repository on

49
00:03:31.479 --> 00:03:34.759
<v Speaker 2>GitHub with twenty one lessons that goes even deeper. The

50
00:03:34.800 --> 00:03:38.240
<v Speaker 2>course walks you from concept to code. It covers what

51
00:03:38.319 --> 00:03:44.240
<v Speaker 2>an agent actually is, the design patterns behind them, tool use, planning,

52
00:03:44.919 --> 00:03:49.319
<v Speaker 2>multi agent systems, memory, and how to actually ship one

53
00:03:49.360 --> 00:03:53.360
<v Speaker 2>to production. There is a hands on interactive version on

54
00:03:53.400 --> 00:03:57.319
<v Speaker 2>LinkedIn Learning too, running about fifty minutes if you just

55
00:03:57.360 --> 00:04:02.560
<v Speaker 2>want the fast pass, and crucially, it uses accessible real

56
00:04:02.599 --> 00:04:07.599
<v Speaker 2>world tooling, not just Microsoft's own stack. Why this is

57
00:04:07.639 --> 00:04:12.360
<v Speaker 2>a big deal for SMB owners Enterprise firms are spending

58
00:04:12.439 --> 00:04:16.160
<v Speaker 2>tens of thousands of dollars per seat sending engineers to

59
00:04:16.240 --> 00:04:21.000
<v Speaker 2>agent boot camps. This course covers the same foundational material

60
00:04:21.519 --> 00:04:25.160
<v Speaker 2>from a company that runs the underlying infrastructure at a

61
00:04:25.199 --> 00:04:29.680
<v Speaker 2>price point of zero. For a small business, that means

62
00:04:29.720 --> 00:04:34.240
<v Speaker 2>your existing developer, or even a technically curious operations lead

63
00:04:34.759 --> 00:04:38.000
<v Speaker 2>can learn enough in a weekend to prototype an agent

64
00:04:38.040 --> 00:04:43.439
<v Speaker 2>that handles quoting, appointment reminders, invoice follow ups, or customer

65
00:04:43.480 --> 00:04:47.560
<v Speaker 2>support triage. You do not need to hire a specialist.

66
00:04:48.279 --> 00:04:52.000
<v Speaker 2>You do not need to buy an enterprise agent platform.

67
00:04:52.240 --> 00:04:57.079
<v Speaker 2>You need one motivated person, ten free lessons, and about

68
00:04:57.120 --> 00:05:01.040
<v Speaker 2>a week. If AI agents have been on your maybe

69
00:05:01.079 --> 00:05:04.519
<v Speaker 2>someday list, this is the week to move them onto

70
00:05:04.560 --> 00:05:06.160
<v Speaker 2>the this quarter list.

71
00:05:07.519 --> 00:05:11.360
<v Speaker 3>Meta has yanked a brand new Instagram AI feature off

72
00:05:11.480 --> 00:05:15.319
<v Speaker 3>the platform just days after launching it after a wave

73
00:05:15.439 --> 00:05:19.879
<v Speaker 3>of public backlash. On July seven, Meta rolled out something

74
00:05:19.959 --> 00:05:24.920
<v Speaker 3>called muse Image. It was built into Meta Ai, Instagram,

75
00:05:24.959 --> 00:05:28.800
<v Speaker 3>Stories and WhatsApp, and it let users generate and edit

76
00:05:28.879 --> 00:05:33.959
<v Speaker 3>AI images, including by feeding it public Instagram photos and

77
00:05:34.120 --> 00:05:38.120
<v Speaker 3>sketching changes on top of them. By July tenth, three

78
00:05:38.199 --> 00:05:42.480
<v Speaker 3>days later, it was gone. Meta admitted in its own

79
00:05:42.519 --> 00:05:46.519
<v Speaker 3>words that it had quote missed the mark end quote

80
00:05:47.000 --> 00:05:50.720
<v Speaker 3>and said the feature was no longer available. The core

81
00:05:50.839 --> 00:05:54.959
<v Speaker 3>problem was privacy. The tool made it very easy to

82
00:05:55.000 --> 00:05:59.519
<v Speaker 3>grab images from public Instagram accounts, run them through generative

83
00:05:59.560 --> 00:06:04.759
<v Speaker 3>AI and produce modified versions of real people, potentially without

84
00:06:04.759 --> 00:06:10.199
<v Speaker 3>their knowledge. Regulators, creators, and every day users pushed back fast,

85
00:06:10.759 --> 00:06:15.319
<v Speaker 3>and Meta blinked. Why this matters for s mbs Two

86
00:06:15.439 --> 00:06:19.360
<v Speaker 3>lessons here. First, if your small business runs any part

87
00:06:19.399 --> 00:06:24.160
<v Speaker 3>of its marketing through Instagram or Meta AI tools, understand

88
00:06:24.360 --> 00:06:28.279
<v Speaker 3>that the feature set is unstable right now. Something you

89
00:06:28.319 --> 00:06:32.720
<v Speaker 3>build a campaign around this week can quietly disappear next week.

90
00:06:33.439 --> 00:06:37.560
<v Speaker 3>Do not become dependent on any single generative AI feature

91
00:06:37.920 --> 00:06:42.560
<v Speaker 3>inside a social platform. Second, the backlash tells you where

92
00:06:42.600 --> 00:06:45.680
<v Speaker 3>the line is with your own customers. If you are

93
00:06:45.720 --> 00:06:50.879
<v Speaker 3>experimenting with AI generated images of clients, employees, or user

94
00:06:50.959 --> 00:06:55.720
<v Speaker 3>submitted photos, get explicit consent and think hard about how

95
00:06:55.759 --> 00:06:58.120
<v Speaker 3>it would look if a screen shot ended up in

96
00:06:58.160 --> 00:07:02.040
<v Speaker 3>a local news story. The technology is running faster than

97
00:07:02.040 --> 00:07:05.560
<v Speaker 3>the trust, and Meta just paid the reputational bill for

98
00:07:05.639 --> 00:07:09.120
<v Speaker 3>finding that out in public. Now to the story, A

99
00:07:09.160 --> 00:07:13.720
<v Speaker 3>lot of executives are quietly panicking about this week. Enterprises

100
00:07:13.759 --> 00:07:17.279
<v Speaker 3>are getting hit with AI bills that are far far

101
00:07:17.399 --> 00:07:21.319
<v Speaker 3>bigger than they expected, and in many cases, those bills

102
00:07:21.360 --> 00:07:24.279
<v Speaker 3>are now higher than the salaries of the workers those

103
00:07:24.360 --> 00:07:27.879
<v Speaker 3>companies laid off to pay for AI in the first place.

104
00:07:28.720 --> 00:07:31.959
<v Speaker 3>A new kp MG survey of over two thousand senior

105
00:07:32.000 --> 00:07:36.680
<v Speaker 3>executives across twenty countries found that twenty nine percent of them,

106
00:07:36.879 --> 00:07:40.360
<v Speaker 3>nearly one in three, have no idea where their growing

107
00:07:40.439 --> 00:07:45.000
<v Speaker 3>AI costs are actually coming from. Another third admitted their

108
00:07:45.040 --> 00:07:49.519
<v Speaker 3>own confusion about AI economics is the main thing blocking

109
00:07:49.560 --> 00:07:54.720
<v Speaker 3>them from deploying AI effectively. In plain English, the people

110
00:07:54.800 --> 00:07:57.959
<v Speaker 3>signing the checks do not understand what they are paying for.

111
00:07:58.879 --> 00:08:04.759
<v Speaker 3>The examples are watering. Uber's chief technology officer disclosed that

112
00:08:04.920 --> 00:08:08.959
<v Speaker 3>Uber burned through its entire twenty twenty six AI coding

113
00:08:09.000 --> 00:08:13.439
<v Speaker 3>budget in four months. By March, eighty four percent of

114
00:08:13.519 --> 00:08:17.519
<v Speaker 3>Uber engineers had adopted the AI coding tool, and roughly

115
00:08:17.600 --> 00:08:22.800
<v Speaker 3>seventy percent of committed code now comes from AI. Uber's

116
00:08:22.839 --> 00:08:27.199
<v Speaker 3>own president admitted publicly that all that token usage did

117
00:08:27.240 --> 00:08:33.440
<v Speaker 3>not clearly correlate with more useful features shipped to customers. Microsoft,

118
00:08:33.600 --> 00:08:37.080
<v Speaker 3>which has invested about thirteen billion dollars in open AI,

119
00:08:37.679 --> 00:08:40.799
<v Speaker 3>told engineers in one of its major divisions to stop

120
00:08:40.919 --> 00:08:44.600
<v Speaker 3>using an AI coding assistant because the bill got out

121
00:08:44.600 --> 00:08:50.200
<v Speaker 3>of control. One unnamed company, per reporting from Axios, racked

122
00:08:50.279 --> 00:08:53.639
<v Speaker 3>up a five hundred million dollar claud bill in a

123
00:08:53.679 --> 00:08:57.279
<v Speaker 3>single month after management forgot to set a usage cap.

124
00:08:58.320 --> 00:09:01.639
<v Speaker 3>Nvidia's own VP of a fplied deep learning has said

125
00:09:01.759 --> 00:09:04.679
<v Speaker 3>on the record that the cost of compute for his

126
00:09:04.840 --> 00:09:10.080
<v Speaker 3>team now exceeds what Nvidia pays the humans using it. Meanwhile,

127
00:09:10.480 --> 00:09:13.759
<v Speaker 3>more than one hundred and fifteen thousand tech workers have

128
00:09:13.799 --> 00:09:17.000
<v Speaker 3>been laid off in twenty twenty six alone, across more

129
00:09:17.039 --> 00:09:20.440
<v Speaker 3>than one hundred and fifty companies, with the stated reason

130
00:09:20.519 --> 00:09:26.440
<v Speaker 3>being reallocation toward AI. Meta cut eight thousand rolls. Wis

131
00:09:26.559 --> 00:09:31.559
<v Speaker 3>cut a fifth of its workforce, Block halved headcount, Atlasian

132
00:09:31.679 --> 00:09:36.480
<v Speaker 3>shed sixteen hundred jobs. And here is the punchline. An

133
00:09:36.639 --> 00:09:42.039
<v Speaker 3>MIT study found AI automation is economically viable in only

134
00:09:42.039 --> 00:09:45.840
<v Speaker 3>about twenty three percent of roles. For the other seventy

135
00:09:45.879 --> 00:09:50.799
<v Speaker 3>seven percent, humans are still cheaper. Companies fired the humans

136
00:09:51.159 --> 00:09:53.240
<v Speaker 3>then found out the machines cost more.

137
00:09:53.559 --> 00:09:56.840
<v Speaker 2>So what does the enterprise AI bill shock mean for

138
00:09:56.919 --> 00:10:00.840
<v Speaker 2>you running a small or mid size business. Honestly, this

139
00:10:00.919 --> 00:10:03.159
<v Speaker 2>is one of the rare weeks where being small is

140
00:10:03.200 --> 00:10:07.639
<v Speaker 2>a genuine advantage. Big companies are stuck. They've locked in

141
00:10:07.840 --> 00:10:11.840
<v Speaker 2>enterprise contracts, laid off staff, and now they're on the

142
00:10:11.879 --> 00:10:15.960
<v Speaker 2>hook for usage based AI pricing that is climbing every quarter.

143
00:10:16.720 --> 00:10:20.759
<v Speaker 2>Analysts project that when AI pricing fully normalizes to reflect

144
00:10:20.919 --> 00:10:25.720
<v Speaker 2>real infrastructure costs, enterprise AI bills will rise another thirty

145
00:10:25.919 --> 00:10:30.600
<v Speaker 2>to fifty percent above where they are today. Anthropic already

146
00:10:30.600 --> 00:10:34.440
<v Speaker 2>moved enterprise customers from flat rate plans to usage based

147
00:10:34.480 --> 00:10:39.399
<v Speaker 2>billing tied to actual compute in April. Get Hub copilot

148
00:10:39.519 --> 00:10:43.600
<v Speaker 2>followed weeks later. The era of flat rate all you

149
00:10:43.639 --> 00:10:47.879
<v Speaker 2>can eat AI is ending. You as a small business,

150
00:10:48.080 --> 00:10:51.600
<v Speaker 2>still have time to be smart. Three things to do

151
00:10:51.679 --> 00:10:57.279
<v Speaker 2>this week. Number One, before you sign any AI tool contract,

152
00:10:57.679 --> 00:11:00.879
<v Speaker 2>ask specifically whether the pricing is flo at rate or

153
00:11:01.000 --> 00:11:05.600
<v Speaker 2>usage based, and ask for a hard monthly cap. That

154
00:11:05.679 --> 00:11:09.559
<v Speaker 2>five hundred million dollar claud bill happened because someone forgot

155
00:11:09.600 --> 00:11:14.200
<v Speaker 2>to set a cap. Do not be that someone. Number two,

156
00:11:15.039 --> 00:11:18.120
<v Speaker 2>Do not use frontier models for tasks that do not

157
00:11:18.320 --> 00:11:22.879
<v Speaker 2>need them. Roughly ninety five percent of enterprise AI usage

158
00:11:23.120 --> 00:11:26.840
<v Speaker 2>still runs on the most expensive frontier models, even for

159
00:11:26.919 --> 00:11:30.360
<v Speaker 2>work that does not require that horse power. If you

160
00:11:30.399 --> 00:11:34.000
<v Speaker 2>are using a top tier model to reformat spreadsheets or

161
00:11:34.039 --> 00:11:38.440
<v Speaker 2>answer basic FAQ emails, you are lighting money on fire.

162
00:11:39.519 --> 00:11:44.159
<v Speaker 2>Match the model to the job number three. Before you

163
00:11:44.240 --> 00:11:48.279
<v Speaker 2>replace a person with AI, do the actual math on

164
00:11:48.399 --> 00:11:54.720
<v Speaker 2>total cost of ownership over twelve months, including tokens, integration, monitoring,

165
00:11:55.320 --> 00:11:59.639
<v Speaker 2>error correction, and the human hours to supervise it. In

166
00:11:59.720 --> 00:12:03.399
<v Speaker 2>roughly three quarters of roles, the human is still the

167
00:12:03.480 --> 00:12:08.679
<v Speaker 2>cheaper option. Zooming out on the whole week Apple versus

168
00:12:08.720 --> 00:12:12.279
<v Speaker 2>Open AI tells you the AI platform wars are entering

169
00:12:12.320 --> 00:12:16.679
<v Speaker 2>a new, more legal phase and your integrations may get bumpy.

170
00:12:17.759 --> 00:12:21.159
<v Speaker 2>Microsoft's Free Agent course tells you the skills gap is

171
00:12:21.200 --> 00:12:24.919
<v Speaker 2>closing fast, and it is closing in your favor if

172
00:12:24.960 --> 00:12:29.720
<v Speaker 2>you move now. Meta's Muse Image Retreat tells you generative

173
00:12:29.759 --> 00:12:34.559
<v Speaker 2>AI features inside social platforms are volatile and consumer trust

174
00:12:34.799 --> 00:12:38.720
<v Speaker 2>is fragile. And the enterprise cost story tells you the

175
00:12:38.759 --> 00:12:42.559
<v Speaker 2>biggest companies in the world do not actually understand their

176
00:12:42.639 --> 00:12:46.759
<v Speaker 2>own AI spending, which means the playbook is still wide

177
00:12:46.799 --> 00:12:51.320
<v Speaker 2>open for a nimble SMB that picks the right models,

178
00:12:51.679 --> 00:12:55.519
<v Speaker 2>sets the right caps, trains one person on agents, and

179
00:12:55.559 --> 00:12:58.080
<v Speaker 2>does not lay off staff to chase a bill it

180
00:12:58.159 --> 00:13:03.639
<v Speaker 2>cannot forecast. This is genuinely a moment of advantage. The

181
00:13:03.720 --> 00:13:08.080
<v Speaker 2>advantage will not last forever, but it is here, right now,

182
00:13:08.559 --> 00:13:10.600
<v Speaker 2>this week. Use it.

183
00:13:11.879 --> 00:13:15.240
<v Speaker 1>That's it for this week's AI assembled the top AI

184
00:13:15.360 --> 00:13:18.960
<v Speaker 1>news for small business. If you found this valuable, share

185
00:13:18.960 --> 00:13:21.360
<v Speaker 1>it with a fellow business owner. And if you want

186
00:13:21.360 --> 00:13:24.399
<v Speaker 1>to know whether AI tools can find and recommend your business,

187
00:13:24.679 --> 00:13:30.399
<v Speaker 1>get your free AI website optimization report at freshmediaworks dot com.

188
00:13:30.440 --> 00:13:35.279
<v Speaker 1>Slash aiready until next week, keep innovating and keep it real.