1
00:00:00,120 --> 00:00:03,000
At the start of every single 
year, I think it's. 

2
00:00:03,160 --> 00:00:06,400
Important. 
To provide what we believe is 

3
00:00:06,400 --> 00:00:10,560
going to be the trends that. 
Shape the next 12 months of AI. 

4
00:00:10,920 --> 00:00:13,920
2025 was a year of huge 
improvement in what? 

5
00:00:14,000 --> 00:00:16,239
AI can actually deliver for 
people. 

6
00:00:16,800 --> 00:00:19,760
However, I think. 
Expectations are still. 

7
00:00:20,080 --> 00:00:24,160
Outstripping reality, 2026 is 
going to be another year of huge

8
00:00:24,160 --> 00:00:26,280
progression, and in today's 
episode, we're going to talk 

9
00:00:26,280 --> 00:00:29,080
about the top. 
Trends that are going to shape 

10
00:00:29,080 --> 00:00:33,000
the world of AI this year, 
there's no doop with Jack 

11
00:00:33,000 --> 00:00:34,840
Harton. 
Hope you enjoy the show. 

12
00:00:47,140 --> 00:00:49,540
So we're actually 21 days into 
January. 

13
00:00:50,340 --> 00:00:52,580
I was going to take a couple of.
Weeks off over Christmas to have

14
00:00:52,580 --> 00:00:55,660
a break. 
Then was moving house. 

15
00:00:56,020 --> 00:00:57,260
Lost a bunch. 
Of things. 

16
00:00:57,260 --> 00:00:59,460
And even today, we're actually 
recording. 

17
00:00:59,660 --> 00:01:02,900
I've still not got the, let's 
say the studio to, so to speak, 

18
00:01:02,980 --> 00:01:04,980
set up. 
The actual SD cards are 

19
00:01:04,980 --> 00:01:08,240
somewhere still living in boxes 
from the Move, so I'm going to 

20
00:01:08,240 --> 00:01:10,720
apologise me wearing a silly 
headset because actually it was 

21
00:01:10,720 --> 00:01:13,560
probably the best microphone I 
could possibly get I think. 

22
00:01:14,240 --> 00:01:15,480
Hopefully. 
We're going to use an AI 

23
00:01:15,480 --> 00:01:18,240
software to make it sound as 
good as we possibly can. 

24
00:01:29,750 --> 00:01:30,830
So now anyway. 
Moving on to. 

25
00:01:30,830 --> 00:01:32,950
Trend number. 
One which I'm going to describe 

26
00:01:32,950 --> 00:01:35,070
in. 
A couple of words which is. 

27
00:01:35,670 --> 00:01:39,310
Differentiation for companies. 
So ultimately, right now, 

28
00:01:39,750 --> 00:01:42,710
differentiation 12 months. 
Ago in the AI space used. 

29
00:01:42,710 --> 00:01:43,950
To look a. 
Lot different. 

30
00:01:43,950 --> 00:01:46,720
To what it. 
Looks like today because 

31
00:01:46,720 --> 00:01:48,280
fundamentally the way users 
interact. 

32
00:01:48,280 --> 00:01:50,080
With AI is just converging on 
the exact. 

33
00:01:50,080 --> 00:01:51,560
Same pattern everywhere. 
Really. 

34
00:01:51,960 --> 00:01:54,480
You know every company building 
an AI product is. 

35
00:01:55,400 --> 00:01:58,440
Essentially a chat bot. 
You know, you type something, 

36
00:01:58,440 --> 00:02:00,680
the AI responds. 
There's maybe interactive 

37
00:02:00,680 --> 00:02:03,880
elements that pop up and appear.
You know you've got Amazon's ad.

38
00:02:03,880 --> 00:02:06,320
Management software's the exact 
same as Google. 

39
00:02:06,320 --> 00:02:09,560
'S and you know every single. 
Interface just starts to look 

40
00:02:09,560 --> 00:02:12,080
quite similar with unique. 
Parts of branding and. 

41
00:02:12,080 --> 00:02:15,840
Colours and vibe as well. 
But mostly it's a chat interface

42
00:02:16,680 --> 00:02:18,400
and you know it makes. 
Sense there's over. 1 1/2 

43
00:02:18,400 --> 00:02:20,960
billion people. 
Using AI every single day right 

44
00:02:20,960 --> 00:02:23,280
now. 
And you know that that. 

45
00:02:23,280 --> 00:02:25,040
Expectation is transferred to 
every product. 

46
00:02:25,040 --> 00:02:27,080
So everyone, naturally. 
Is becoming a chat. 

47
00:02:27,080 --> 00:02:30,280
Bot and it's. 
Therefore, quite difficult as 

48
00:02:30,280 --> 00:02:33,080
organizations, as people to go. 
What is it that makes this 

49
00:02:33,080 --> 00:02:35,720
company unique Then? 
You know, if everyone looks the 

50
00:02:35,720 --> 00:02:38,840
same and it's an agent doing 
tasks for people, what makes 

51
00:02:38,840 --> 00:02:41,560
them, you know, unique? 
So what's happening is really 

52
00:02:41,560 --> 00:02:43,160
interesting. 
Questions being posed, which is 

53
00:02:43,520 --> 00:02:45,520
OK, therefore. 
What isn't isn't 

54
00:02:45,520 --> 00:02:46,840
differentiation. 
That's going to be a big 

55
00:02:46,840 --> 00:02:49,280
question of this year. 
My belief is. 

56
00:02:49,280 --> 00:02:53,640
It's not the interface anymore, 
it's 12 months ago strategically

57
00:02:53,640 --> 00:02:55,440
companies building. 
New chat interfaces and 

58
00:02:55,440 --> 00:02:57,800
experiences that was. 
Strategic, clever work. 

59
00:02:58,400 --> 00:03:01,040
Now it's just. 
Not again, because building 

60
00:03:01,040 --> 00:03:03,440
better chat interfaces is really
important, but doesn't 

61
00:03:03,440 --> 00:03:06,200
differentiate. 
Anybody you know, everyone has 

62
00:03:06,200 --> 00:03:09,120
to be solving that exact same 
problem the exact same way. 

63
00:03:09,920 --> 00:03:11,600
And you? 
Everyone has it, every user 

64
00:03:11,600 --> 00:03:15,280
expects it, so it becomes table 
stakes, so therefore it raises. 

65
00:03:15,280 --> 00:03:17,680
An interesting question if the. 
Chat interface or the interface 

66
00:03:17,680 --> 00:03:20,360
layer that used to be for 
companies, a place that offered 

67
00:03:20,360 --> 00:03:22,480
massive differentiation. 
You know, big companies like 

68
00:03:22,480 --> 00:03:26,280
let's say Salesforce or HubSpot 
have ecosystems and training 

69
00:03:26,280 --> 00:03:29,200
programs and certificates for 
becoming a master at my 

70
00:03:29,200 --> 00:03:32,000
software. 
They used to literally monetize 

71
00:03:32,000 --> 00:03:35,800
friction of having lots of 
things and features on a page 

72
00:03:36,160 --> 00:03:38,240
because a salesperson could 
easily point it in and say, 

73
00:03:38,240 --> 00:03:40,240
look, we do everything and it 
doesn't. 

74
00:03:40,240 --> 00:03:42,280
Matter that you don't know how 
because we've got big training 

75
00:03:42,280 --> 00:03:43,600
programs to make. 
You an expert? 

76
00:03:44,320 --> 00:03:46,160
So that used to be 
differentiation because you used

77
00:03:46,160 --> 00:03:46,840
to. 
Look. 

78
00:03:46,840 --> 00:03:49,040
Complex, whereas now everything 
is a. 

79
00:03:49,040 --> 00:03:52,040
Chat interface and an agent 
orchestrates and just takes all 

80
00:03:52,040 --> 00:03:53,960
those actions. 
So therefore it's a really 

81
00:03:53,960 --> 00:03:56,680
interesting one because. 
The question becomes, where does

82
00:03:56,680 --> 00:03:58,760
competitive advantage therefore 
come from and. 

83
00:03:58,760 --> 00:04:00,720
This is where I'm going to go to
trend #2 which? 

84
00:04:00,720 --> 00:04:02,920
Is where companies are going to 
start investing effort. 

85
00:04:03,280 --> 00:04:05,520
So trend 2. 
Broadly is is where? 

86
00:04:05,520 --> 00:04:09,480
Companies are redirecting. 
Their efforts towards what let's

87
00:04:09,480 --> 00:04:11,600
describe as their unique 
intelligence. 

88
00:04:11,720 --> 00:04:12,200
So what? 
Do I? 

89
00:04:12,200 --> 00:04:14,880
Mean by intelligence? 
So the intelligence. 

90
00:04:14,880 --> 00:04:20,360
Is about unique ways of solving 
customer problems in your 

91
00:04:20,360 --> 00:04:22,320
specific domain. 
So for example, if you're 

92
00:04:22,320 --> 00:04:25,400
building, let's say. 
For example, a company that is 

93
00:04:25,400 --> 00:04:28,720
selling sales software. 
Which is maybe they're going to 

94
00:04:28,720 --> 00:04:30,360
expose. 
Unique intelligence, which is 

95
00:04:30,360 --> 00:04:34,200
the ability to, say, predict 
which deals might close based on

96
00:04:34,200 --> 00:04:38,560
five years of deal history. 
And maybe showing patterns of 

97
00:04:38,560 --> 00:04:41,960
historic. 
Alternatives and competitors and

98
00:04:42,160 --> 00:04:45,960
behaviours from salespeople that
have led to more deals being won

99
00:04:46,280 --> 00:04:50,600
or less deals and that being 
exposed for an agent and a user.

100
00:04:50,600 --> 00:04:53,080
So that's intelligence. 
So if it's a logistics platform,

101
00:04:53,080 --> 00:04:55,760
so it helps people get packages 
from A-Z. 

102
00:04:56,720 --> 00:04:59,120
Maybe they can. 
Analyse millions of shipments. 

103
00:04:59,120 --> 00:05:00,840
Across every region in the 
world. 

104
00:05:01,160 --> 00:05:04,800
And then expose that to the 
customer through AI or if your 

105
00:05:04,800 --> 00:05:06,800
customer support platform for 
example. 

106
00:05:07,080 --> 00:05:08,440
It's understanding exactly. 
Which? 

107
00:05:08,480 --> 00:05:11,880
Issues really need human. 
Intervention based on historic 

108
00:05:11,880 --> 00:05:14,560
data. 
So that's what we mean by 

109
00:05:14,560 --> 00:05:15,920
intelligence. 
And that's where. 

110
00:05:15,920 --> 00:05:18,360
Companies, we believe. 
Over this next period is going 

111
00:05:18,360 --> 00:05:19,600
to be. 
Trickling down on. 

112
00:05:19,800 --> 00:05:21,880
Because if they expose more and 
more and more of that. 

113
00:05:22,320 --> 00:05:24,760
That is extremely. 
Unique and something no one else

114
00:05:24,800 --> 00:05:26,840
has. 
Rather than spending all that 

115
00:05:26,840 --> 00:05:29,840
time on the chat interface, it's
just about exposing that. 

116
00:05:29,840 --> 00:05:31,800
Value to. 
Customers and users. 

117
00:05:31,960 --> 00:05:34,160
So yeah, I think that's going to
be a massive trend of companies 

118
00:05:34,160 --> 00:05:35,880
trying to. 
Expose that very unique 

119
00:05:35,880 --> 00:05:37,720
intelligence, rather. 
Than think through. 

120
00:05:37,720 --> 00:05:39,960
Long user journeys and 
experiences anymore because 

121
00:05:39,960 --> 00:05:42,760
ultimately. 
The chat interface all looks. 

122
00:05:42,760 --> 00:05:45,160
The same. 
It's the outcomes an agent. 

123
00:05:45,160 --> 00:05:48,160
Can deliver for a person and how
clever those outcomes are. 

124
00:05:48,480 --> 00:05:50,440
That is what delivers value in 
the future anyway. 

125
00:05:50,520 --> 00:05:53,040
We're going to go to trend #3. 
Trend #3 is that. 

126
00:05:53,040 --> 00:05:55,840
AI conversations are about to. 
Become immensely visual. 

127
00:05:56,040 --> 00:05:58,320
So yeah, AI conversations are 
moving from like text. 

128
00:05:58,320 --> 00:06:01,160
Based to. 
Really visual, interactive. 

129
00:06:01,160 --> 00:06:02,920
Experiences. 
So you've seen this with Chatchi

130
00:06:02,920 --> 00:06:05,160
BT? 
So Open AI and Chatchi BT 

131
00:06:05,160 --> 00:06:08,960
released integrations into 
Spotify, Expedia, Instacart, and

132
00:06:08,960 --> 00:06:11,800
essentially the Spotify playlist
renders in a conversation. 

133
00:06:12,440 --> 00:06:15,360
Or you can do learning. 
Through the big learning 

134
00:06:15,360 --> 00:06:18,960
platforms straight into track 
GBT and their platform appears 

135
00:06:18,960 --> 00:06:22,000
inside the chat interface. 
So that what's becoming a trend 

136
00:06:22,000 --> 00:06:24,200
really is that the chat 
interface. 

137
00:06:24,200 --> 00:06:27,160
Wants to render beautiful user 
interface components from 

138
00:06:27,160 --> 00:06:29,480
different platforms because 
fundamentally again. 

139
00:06:29,920 --> 00:06:32,840
The more outcomes and agent. 
Can deliver, the more valuable 

140
00:06:32,840 --> 00:06:35,600
it is, and so for all companies 
they want more. 

141
00:06:35,840 --> 00:06:38,160
UI. 
Components deliver more outcomes

142
00:06:38,280 --> 00:06:40,440
in their chat. 
Because if you think about, 

143
00:06:40,440 --> 00:06:43,280
let's say. 
I want to select a playlist that

144
00:06:43,280 --> 00:06:46,360
might have taken a conversation 
that lasted 10 minutes suddenly.

145
00:06:46,600 --> 00:06:48,880
The user interface. 
Visual component appears in the 

146
00:06:48,880 --> 00:06:51,200
chart and you just select your 
playlist and press go. 

147
00:06:52,240 --> 00:06:54,760
So you want to book a flight. 
Instead of reading a description

148
00:06:54,760 --> 00:07:00,000
of flight information, you just 
have a comparison of flights in 

149
00:07:00,000 --> 00:07:02,360
a visual format that has ticks 
and crosses. 

150
00:07:02,360 --> 00:07:04,840
For example, just as you would 
get on a normal website that's 

151
00:07:04,840 --> 00:07:07,200
straight from ChatGPT or 
straight from a conversation. 

152
00:07:07,360 --> 00:07:11,120
So you can bet. £1,000,000 that.
Claude is about to do the same. 

153
00:07:11,120 --> 00:07:13,640
Microsoft, Apple, when they 
actually get their act together.

154
00:07:14,680 --> 00:07:17,600
A slightly technical diversion 
here is that the model context 

155
00:07:17,600 --> 00:07:23,360
protocols, so MCP protocol. 
So MCP is actually expanding 

156
00:07:23,680 --> 00:07:26,880
what it enables you to do. 
So actually you can provide UI 

157
00:07:26,880 --> 00:07:30,920
components inside of that very 
soon, which means that as one 

158
00:07:30,920 --> 00:07:34,800
vendor to another or an 
organization, you can then let's

159
00:07:34,800 --> 00:07:38,880
say you sell your maps software,
you can not just put your maps 

160
00:07:38,880 --> 00:07:42,400
information into an MCP, you can
put your maps user interfaces 

161
00:07:42,400 --> 00:07:45,440
into an MCP and then other 
platforms might go to surface 

162
00:07:45,440 --> 00:07:47,680
that up through a conversation. 
So yeah, that's. 

163
00:07:47,680 --> 00:07:50,360
The three trends on. 
Let's describe it as kind of the

164
00:07:50,360 --> 00:07:55,400
interface part, so where the AI 
interface is changing. 

165
00:07:55,720 --> 00:07:58,720
I want to make a change of 
direction and move on to I guess

166
00:07:59,000 --> 00:08:00,800
trends on how. 
Agents are built. 

167
00:08:12,110 --> 00:08:14,590
So trend #4. 
Of this list I think. 

168
00:08:14,590 --> 00:08:16,910
Agent reasoning is. 
Going to start to really 

169
00:08:16,910 --> 00:08:18,190
squeeze. 
And compete. 

170
00:08:18,550 --> 00:08:21,990
Rigid workflows or even clever 
workflows from the likes of 

171
00:08:21,990 --> 00:08:24,350
Zapier NAN. 
Lindy. 

172
00:08:24,600 --> 00:08:27,120
So the very big popular 
platforms for AI automation. 

173
00:08:27,320 --> 00:08:29,960
And I think really there's a. 
Big debate going on online all 

174
00:08:29,960 --> 00:08:32,159
the time. 
Which is you see a lot of 

175
00:08:32,159 --> 00:08:34,120
people. 
Sharing their great NAN workflow

176
00:08:34,120 --> 00:08:36,320
that saved ME3. 1000 hours of 
time. 

177
00:08:36,559 --> 00:08:38,760
People see that, they play with 
it and it works, It makes 

178
00:08:38,760 --> 00:08:41,240
complete sense, and for many 
tasks it makes a lot of sense to

179
00:08:41,240 --> 00:08:43,880
use that. 
But as a result, people assume 

180
00:08:43,880 --> 00:08:46,920
that's the way agents should be 
built and LED, and that's where 

181
00:08:46,920 --> 00:08:49,640
the fundamental debate comes in 
and you've got the likes of Lang

182
00:08:49,640 --> 00:08:52,640
Chen. 
CEO who really complain that a. 

183
00:08:52,720 --> 00:08:55,680
Open AI release yet another 
workflow builder. 

184
00:08:56,640 --> 00:08:58,120
You've got anthropic 
complaining. 

185
00:08:58,120 --> 00:08:59,840
About the exact same. 
Because they don't. 

186
00:08:59,840 --> 00:09:02,080
Believe workflows are the future
and so. 

187
00:09:02,080 --> 00:09:03,680
A lot of. 
Companies default to workflows 

188
00:09:03,680 --> 00:09:05,640
because they feel like they're 
able to. 

189
00:09:06,280 --> 00:09:08,960
Be in control, so I. 
Know this agent is going to do 

190
00:09:08,960 --> 00:09:10,240
this, then it's going to do 
that. 

191
00:09:11,040 --> 00:09:13,640
But often when? 
It comes to real world 

192
00:09:14,280 --> 00:09:17,480
complexity for, let's say, 
deploying workflows. 

193
00:09:18,640 --> 00:09:21,520
They're going to break because a
human person wants to go in 100 

194
00:09:21,520 --> 00:09:24,920
different directions. 
And that's where agents who can 

195
00:09:24,920 --> 00:09:26,320
reason and create a plan like 
you. 

196
00:09:26,320 --> 00:09:28,440
Might see on cloud. 
So it creates a specific plan, 

197
00:09:28,440 --> 00:09:31,560
it asks itself questions, it 
ticks itself off on a plan and 

198
00:09:31,920 --> 00:09:34,440
then uses information to shape 
that plan. 

199
00:09:34,560 --> 00:09:36,720
That's what actually a workflow 
is doing. 

200
00:09:36,720 --> 00:09:39,280
But you're trying to build a 
workflow and try and plan for 

201
00:09:39,280 --> 00:09:42,280
hundreds of scenarios, whereas 
an agent can do that and change 

202
00:09:42,280 --> 00:09:44,240
it on the fly. 
So yeah, Zapier. 

203
00:09:44,240 --> 00:09:47,240
NAN and tools like that. 
I definitely feel attractive 

204
00:09:47,240 --> 00:09:50,480
because you know, you can drag 
and drop things, you can select 

205
00:09:50,480 --> 00:09:54,400
arrows, you can define logic for
an agent must do this and then 

206
00:09:54,400 --> 00:09:56,520
that. 
You can literally see what's 

207
00:09:56,520 --> 00:09:58,720
going to happen. 
See, as I said, but life. 

208
00:09:59,160 --> 00:10:02,360
Tasks. 
And humans just don't like 

209
00:10:02,360 --> 00:10:04,000
following that sequence of 
steps, so. 

210
00:10:04,000 --> 00:10:06,960
If I wanted. 
To do a workflow, I might go and

211
00:10:06,960 --> 00:10:09,400
set a step for saying. 
Check for duplicates and if 

212
00:10:09,400 --> 00:10:11,080
found. 
An issue? 

213
00:10:11,160 --> 00:10:13,320
Refund them, then check 
cancellations. 

214
00:10:13,320 --> 00:10:16,480
And if you find an issue, retry.
Check the payments, but if 

215
00:10:16,480 --> 00:10:18,600
there's an invalid prompt for 
update. 

216
00:10:18,760 --> 00:10:19,880
But the duplicate. 
Charge was. 

217
00:10:19,880 --> 00:10:21,360
Actually an upgrade the 
cancellation. 

218
00:10:21,360 --> 00:10:23,400
Would then fail because there's 
a pending charge. 

219
00:10:23,520 --> 00:10:24,960
So everything becomes. 
Interconnected. 

220
00:10:24,960 --> 00:10:27,520
The workflow can't handle that 
level of interconnectedness and 

221
00:10:27,520 --> 00:10:29,240
chaos. 
It becomes very difficult. 

222
00:10:29,560 --> 00:10:31,280
Whereas agent reasoning can 
assess. 

223
00:10:31,280 --> 00:10:33,120
A full situation. 
Have a conversation with the 

224
00:10:33,120 --> 00:10:35,320
user. 
Identify the three issues. 

225
00:10:35,680 --> 00:10:37,720
And know that they're. 
Related because it is able to 

226
00:10:37,720 --> 00:10:40,480
analyse more information. 
And create a plan and again it 

227
00:10:40,480 --> 00:10:42,280
can. 
Change its plan and then check 

228
00:10:42,280 --> 00:10:43,960
in with the users Say is this 
good or bad? 

229
00:10:44,080 --> 00:10:46,600
See, I think a massive trend 
will be moving towards agent. 

230
00:10:46,600 --> 00:10:49,160
Reasoning because people who 
love cloud code and how powerful

231
00:10:49,160 --> 00:10:52,720
it is and how effective it is 
and it's something we're mindset

232
00:10:52,720 --> 00:10:55,320
tripling down on. 
Because I ultimately. 

233
00:10:55,800 --> 00:10:59,080
We see the value in that. 
It's where we're just releasing 

234
00:10:59,080 --> 00:11:02,400
new technology where I can look 
over millions of rows of data 

235
00:11:02,400 --> 00:11:05,520
and create a plan for how to. 
Click into very. 

236
00:11:05,520 --> 00:11:08,760
Complex questions and. 
Find the right data to help it. 

237
00:11:08,760 --> 00:11:12,760
Answer those questions. 
And give real value to users. 

238
00:11:12,800 --> 00:11:14,280
So trend 5. 
Is. 

239
00:11:14,480 --> 00:11:16,320
Again, within the theme of how 
agents. 

240
00:11:16,320 --> 00:11:18,080
Are built. 
And specifically, this theme is 

241
00:11:18,080 --> 00:11:21,320
how I think Building Agents is 
going to move to. 

242
00:11:21,320 --> 00:11:23,800
More conversational interfaces. 
And I think that's what's. 

243
00:11:23,800 --> 00:11:25,080
Going to get widespread 
adoption. 

244
00:11:25,920 --> 00:11:27,560
So I think it's going to move 
from quite technical 

245
00:11:27,560 --> 00:11:30,440
configuration or even. 
Workflows or even. 

246
00:11:30,440 --> 00:11:35,360
Code to configuration. 
Via conversations, you know 

247
00:11:35,360 --> 00:11:37,200
describe what the agent should 
do and the system should just 

248
00:11:37,200 --> 00:11:40,600
generate the configuration and 
this might expand who builds 

249
00:11:40,600 --> 00:11:42,000
agents. 
From maybe 2%. 

250
00:11:42,400 --> 00:11:46,280
Of people and knowledge workers 
to, you know, 50 percent, 60%. 

251
00:11:46,760 --> 00:11:48,040
Because right now, building an 
agent. 

252
00:11:48,040 --> 00:11:50,840
Requires quite a few technical 
skills and thinking a lot of the

253
00:11:50,840 --> 00:11:52,680
time, even if you try and 
simplify it. 

254
00:11:53,040 --> 00:11:55,480
You know, Zapier and other 
similar tools. 

255
00:11:55,840 --> 00:11:57,360
Make you think like a 
programmer. 

256
00:11:58,400 --> 00:12:01,120
You have to understand variables
and data types and condition 

257
00:12:01,120 --> 00:12:04,280
logic and error handling and 
code first. 

258
00:12:04,280 --> 00:12:06,200
Approach. 
Requires actual programming. 

259
00:12:07,000 --> 00:12:10,200
And the issue there is that. 
An agent's very rarely one and 

260
00:12:10,200 --> 00:12:11,160
done. 
You're then going to. 

261
00:12:11,160 --> 00:12:13,040
Have to maintain it and improve.
It and change. 

262
00:12:13,040 --> 00:12:15,160
It So yeah, what I think is 
going to start emerging. 

263
00:12:15,160 --> 00:12:17,680
For agent building is very 
conversational studio 

264
00:12:17,680 --> 00:12:19,560
interfaces. 
It's something we're actually 

265
00:12:19,560 --> 00:12:22,560
about to start launching as AV 
two of our agent. 

266
00:12:22,560 --> 00:12:25,480
Building Google AI. 
Studio is now moving towards 

267
00:12:25,480 --> 00:12:28,600
that where you can. 
Configure these experiences. 

268
00:12:28,600 --> 00:12:31,360
Through pretty. 
Minimal. 

269
00:12:31,360 --> 00:12:33,840
Interface, really, so soon 
you'll be able to say I just 

270
00:12:33,840 --> 00:12:36,440
want an agent that monitors 
support tickets, figures out you

271
00:12:36,440 --> 00:12:38,920
know, which ones are urgent and 
just reach them to the right 

272
00:12:38,920 --> 00:12:40,280
person. 
And if a ticket sits there for 

273
00:12:40,280 --> 00:12:43,160
more than 24 hours, escalate it.
That's how we think and in terms

274
00:12:43,160 --> 00:12:45,960
of instructions and the system 
should just know how to do that 

275
00:12:45,960 --> 00:12:49,560
and build that trend 6, I think.
Agent testing is about to become

276
00:12:49,720 --> 00:12:52,880
a major. 
Area of focus for all companies 

277
00:12:53,560 --> 00:12:54,880
because the problem. 
Often isn't AI. 

278
00:12:54,880 --> 00:12:57,480
Capability and This is why 
people gravitate, we think to 

279
00:12:57,480 --> 00:12:59,600
workflows the. 
Problem is confidence. 

280
00:12:59,960 --> 00:13:01,920
Confidence that agents are going
to make good decisions 

281
00:13:01,920 --> 00:13:04,560
repeatedly in scenarios. 
Actually kind of matter. 

282
00:13:04,880 --> 00:13:07,800
Large scale AI LED training and 
testing is going to be what 

283
00:13:07,800 --> 00:13:11,160
builds that confidence because 
traditional say software testing

284
00:13:11,160 --> 00:13:13,960
of a feature doesn't work Agent.
Testing is quite different 

285
00:13:13,960 --> 00:13:15,240
because. 
Behavior isn't, let's say, 

286
00:13:15,240 --> 00:13:17,640
deterministic. 
Give an agent a support. 

287
00:13:17,640 --> 00:13:19,720
Query. 
It might handle it in five ways 

288
00:13:20,080 --> 00:13:22,960
over 5 separate occasions, even 
though it should be consistent. 

289
00:13:23,840 --> 00:13:25,840
Multiple. 
Approaches could all be correct,

290
00:13:26,000 --> 00:13:28,440
but slightly different, which 
makes us all as humans feel 

291
00:13:28,440 --> 00:13:29,840
quite nervous. 
It's like a human. 

292
00:13:30,640 --> 00:13:33,680
And so the answer really for. 
Us is I think we're at mindset 

293
00:13:33,680 --> 00:13:36,000
and what we're going to expect 
us to see everywhere in the 

294
00:13:36,000 --> 00:13:39,120
world very soon, which is that 
often ahead of things by, you 

295
00:13:39,120 --> 00:13:42,400
know, three to seven months is 
scenario based testing at 

296
00:13:42,400 --> 00:13:45,240
massive scale using AI to 
evaluate everything. 

297
00:13:45,640 --> 00:13:47,600
So you. 
Describe the scenarios that 

298
00:13:47,600 --> 00:13:49,800
represent real scenarios or 
situations. 

299
00:13:50,280 --> 00:13:53,800
Define success criteria based on
judgement and use AI to. 

300
00:13:53,800 --> 00:13:55,440
Test itself and. 
Test AI. 

301
00:13:55,640 --> 00:13:57,800
And I think this is all. 
Going to connect to ROI as well 

302
00:13:57,800 --> 00:14:01,000
because I think people are going
to move from say vibes of AI 

303
00:14:01,000 --> 00:14:04,240
deployments to real ROI and 
impact of performance. 

304
00:14:15,640 --> 00:14:17,560
So that's how agents are. 
Billed as a theme set. 

305
00:14:18,160 --> 00:14:20,320
I think let's move on now to 
another theme which. 

306
00:14:20,320 --> 00:14:21,600
Is how work. 
Changes. 

307
00:14:21,600 --> 00:14:23,720
As a result of all of this and 
as part of this. 

308
00:14:23,720 --> 00:14:26,200
Let's go into. 
Trend 7, which is I think. 

309
00:14:26,200 --> 00:14:27,600
Knowledge work is. 
Going to start. 

310
00:14:27,600 --> 00:14:31,040
Shifting from doing. 
To now managing because all of 

311
00:14:31,040 --> 00:14:32,960
these things that we're going to
start to see is going to. 

312
00:14:32,960 --> 00:14:35,480
Simplify agent creation. 
But make it more powerful 

313
00:14:35,680 --> 00:14:38,960
because workflows and complex 
interfaces is not what's going 

314
00:14:38,960 --> 00:14:43,560
to deliver wide scale, let's 
say, adoption amongst the 

315
00:14:43,560 --> 00:14:46,280
workforce, especially in 
knowledge work. 

316
00:14:46,400 --> 00:14:49,000
So I think all of these things 
are going to start culminating 

317
00:14:49,000 --> 00:14:52,960
in knowledge workers moving from
say 1 or 2% or 8% of people 

318
00:14:52,960 --> 00:14:56,520
working really fast to let's say
50, sixty, 70%. 

319
00:14:56,680 --> 00:14:59,720
You're starting to see this now 
with Anthropic launching Cloud 

320
00:14:59,720 --> 00:15:01,120
Co work. 
Which is essentially. 

321
00:15:01,120 --> 00:15:04,560
Trying to give the experience of
Cloud Code to. 

322
00:15:04,720 --> 00:15:07,280
Non-technical people because the
number of non. 

323
00:15:07,280 --> 00:15:10,040
Technical people using cloud 
code again, trying to build and 

324
00:15:10,040 --> 00:15:12,880
provide an agent LED experience 
with a series of tools like I 

325
00:15:12,880 --> 00:15:15,160
was describing. 
No workflows and making it 

326
00:15:15,160 --> 00:15:16,480
really simple to manage for 
them. 

327
00:15:17,640 --> 00:15:19,120
So I think knowledge work is 
going to shift from. 

328
00:15:19,280 --> 00:15:23,480
Doing tasks to managing agents 
and therefore testing frameworks

329
00:15:23,480 --> 00:15:26,240
and simplicity is going to be 
really important within this 

330
00:15:26,240 --> 00:15:27,880
theme, I think. 
Trend A is. 

331
00:15:27,880 --> 00:15:29,360
Going to be. 
The year. 

332
00:15:29,360 --> 00:15:30,680
Of agent manager. 
As a result. 

333
00:15:30,760 --> 00:15:34,120
And the reason I say agent 
management, not full autonomy is

334
00:15:34,120 --> 00:15:36,000
because I don't think full 
autonomy is close. 

335
00:15:36,040 --> 00:15:37,520
I think we're quite a while away
from that. 

336
00:15:37,960 --> 00:15:39,360
So I think that's why. 
Agents are so. 

337
00:15:39,360 --> 00:15:41,120
Powerful versus, say, a 
workflow. 

338
00:15:41,520 --> 00:15:43,880
Because an agent will. 
Do a series of tasks and then 

339
00:15:43,880 --> 00:15:48,320
check in with you, ask the user 
to make that judgement call, 

340
00:15:48,320 --> 00:15:50,680
make strategic decisions. 
But the agent will start to 

341
00:15:50,680 --> 00:15:53,800
increasingly handle more and 
more of the execution because 

342
00:15:53,800 --> 00:15:56,360
it's simpler just to provide 
agents tools and agents to 

343
00:15:56,360 --> 00:15:58,000
reasoning. 
So yeah, I think This is why 

344
00:15:58,000 --> 00:15:59,440
it's going to be so. 
Powerful The a lot. 

345
00:15:59,440 --> 00:16:02,080
Of these trends are just going 
to culminate in some of these 

346
00:16:02,080 --> 00:16:04,840
things because ultimately. 
Human judgements going to be. 

347
00:16:04,840 --> 00:16:07,560
Still really important, but 
agents begin to do powerful. 

348
00:16:07,560 --> 00:16:10,400
Things with powerful reasoning. 
Is going to be widespread across

349
00:16:10,400 --> 00:16:13,440
many different products, not 
just Chatter, BT and cloud. 

350
00:16:14,000 --> 00:16:15,800
It's something that we're 
providing all of our customers. 

351
00:16:15,800 --> 00:16:19,440
For example, the ability to have
reasoning agents and the ability

352
00:16:19,440 --> 00:16:23,080
to run large scale testing on 
your agents and AI experiences. 

353
00:16:23,200 --> 00:16:26,160
Another big theme I think we're 
going to start to see as Tread 9

354
00:16:26,160 --> 00:16:29,760
is forward deployed vibe coding.
So. 

355
00:16:29,800 --> 00:16:32,160
We we shouldn't actually. 
Use the word vibe coding because

356
00:16:32,160 --> 00:16:36,320
it almost, let's say, lowers the
perceived value of using AI as a

357
00:16:36,320 --> 00:16:38,360
coding tool. 
But anyway, there's a there's a 

358
00:16:38,360 --> 00:16:39,840
new role that's emerging more 
and more. 

359
00:16:39,840 --> 00:16:41,400
Which is people in departments 
who come. 

360
00:16:41,400 --> 00:16:45,640
Back who combine domain 
expertise with AI building 

361
00:16:45,640 --> 00:16:48,320
capabilities essentially. 
So not necessarily professional 

362
00:16:48,320 --> 00:16:51,720
engineers, but domain experts 
who can build solutions to 

363
00:16:51,720 --> 00:16:54,080
problems so these. 
People know really how to use 

364
00:16:54,120 --> 00:16:56,720
AI. 
Tools and the domain experts and

365
00:16:56,720 --> 00:17:00,400
they forward deploy them to 
solve big problems and not get 

366
00:17:00,400 --> 00:17:04,040
it to let's say full production 
or ready for everyone to work on

367
00:17:04,040 --> 00:17:06,560
and or to deploy to customers or
launch to customers. 

368
00:17:07,040 --> 00:17:09,480
What we are seeing is they may 
have to solve big knotty 

369
00:17:09,480 --> 00:17:11,520
problems because maybe they're 
expert coders and they have 

370
00:17:11,520 --> 00:17:14,640
domain expertise or they're just
good at AI tools and have 

371
00:17:14,640 --> 00:17:17,200
fantastic AI and domain 
expertise. 

372
00:17:17,720 --> 00:17:20,800
And the reason domain expertise 
is important is ultimately they 

373
00:17:20,800 --> 00:17:24,560
have a lot a lot of context to 
make good judgement calls and 

374
00:17:24,560 --> 00:17:26,520
solve big problems for 
companies. 

375
00:17:26,920 --> 00:17:30,320
And that could be. 
Ideas for new customer products?

376
00:17:30,360 --> 00:17:34,200
It could be solving big problems
for teams and essentially 

377
00:17:34,200 --> 00:17:38,320
building tools and systems using
AI and then eventually handing 

378
00:17:38,320 --> 00:17:51,730
it over to the rest of the team.
So these are the trends that we 

379
00:17:51,730 --> 00:17:54,010
think will. 
Define AI in 2026. 

380
00:17:54,130 --> 00:17:58,850
The interface is experiencing 
the biggest revolution since 

381
00:17:59,400 --> 00:18:02,160
command line interfaces. 
Companies will be investing more

382
00:18:02,160 --> 00:18:03,480
and. 
More in their own data to 

383
00:18:03,480 --> 00:18:07,040
provide as unique intelligence 
to their users and customers and

384
00:18:07,040 --> 00:18:08,160
ultimately. 
You'll see more. 

385
00:18:08,160 --> 00:18:10,320
Deployments than ever throughout
2026. 

386
00:18:10,440 --> 00:18:12,120
So that's it for this week in 
this episode. 

387
00:18:12,120 --> 00:18:14,040
I hope you found it useful. 
Next week we'll have a proper. 

388
00:18:14,040 --> 00:18:17,560
Microphone, camera and setup. 
But yeah, we'll see you next 

389
00:18:17,560 --> 00:18:17,800
week.