1
00:00:06,520 --> 00:00:08,660
Thank you, brother. Thank you, Eric.

2
00:00:08,960 --> 00:00:12,080
What it takes to trust AI with real work.

3
00:00:16,160 --> 00:00:18,420
Thank you, Eric. Thank you, George.

4
00:00:18,820 --> 00:00:22,280
Thank you, AIBC. And
thank you to our team.

5
00:00:22,460 --> 00:00:28,320
We're very honored to win the best
enterprise AI platform for 2026 at the

6
00:00:28,320 --> 00:00:31,660
award show the other night.
This has been a fantastic week.

7
00:00:31,800 --> 00:00:36,500
And I'm really honored to be here and
to tell you a bit of our vision for the

8
00:00:36,500 --> 00:00:40,560
future and how it will impact your
business today and how you can get

9
00:00:40,560 --> 00:00:44,140
involved. So if you're having a chat
right behind these seats, consider

10
00:00:44,140 --> 00:00:46,520
sitting down and dropping
in with us for 10 minutes.

11
00:00:47,860 --> 00:00:51,080
I'm Will O 'Brien, founder
and CEO of Provenonce.

12
00:00:51,260 --> 00:00:54,380
My background is in building
trust infrastructure.

13
00:00:54,860 --> 00:00:59,500
Trust infrastructure means systems
that help people protect their assets,

14
00:00:59,800 --> 00:01:03,040
safely transact, and retain control.

15
00:01:03,780 --> 00:01:07,840
Today, I want to talk about what it
takes to trust AI with real work.

16
00:01:07,840 --> 00:01:11,860
The opportunity is much bigger
than a better answer in a chat.

17
00:01:12,460 --> 00:01:16,720
We're giving these systems the ability
to act, and we need to be confident

18
00:01:16,720 --> 00:01:19,060
about what we're trusting them to do.

19
00:01:20,500 --> 00:01:24,560
I'm first going to take you back to the
early days of crypto and a few lessons

20
00:01:24,560 --> 00:01:27,580
that we learned building
trust in a new industry.

21
00:01:28,240 --> 00:01:31,580
Then I want to take a look at some of
the business challenges, opportunities

22
00:01:31,580 --> 00:01:37,440
today, and looking ahead to the future
in which AI agents work and transact on

23
00:01:37,440 --> 00:01:38,490
our behalf.

24
00:01:39,640 --> 00:01:41,400
Let's go back to 2013.

25
00:01:41,880 --> 00:01:46,500
In 2013, I co -founded BitGo
and served as its early CEO.

26
00:01:46,840 --> 00:01:50,560
At the time, there was only single -key
wallets, which meant that if someone

27
00:01:50,560 --> 00:01:54,000
gets access to your private
key, someone can get your coins.

28
00:01:54,400 --> 00:01:59,500
We innovated on a new technology called
multi -signature, but technology was

29
00:01:59,500 --> 00:02:02,560
not sufficient. Cryptography
was only the beginning.

30
00:02:02,780 --> 00:02:05,720
People needed an understandable interface.

31
00:02:06,920 --> 00:02:11,100
Businesses needed process and human
decisions needed to be involved.

32
00:02:11,240 --> 00:02:13,600
Things like treasury
controls, spending limits,

33
00:02:14,540 --> 00:02:20,140
audit trails. In January of this year,
BitGo made its public market debut on

34
00:02:20,140 --> 00:02:24,560
the New York Stock Exchange, reaching
a market cap of nearly $2 .6 billion.

35
00:02:25,740 --> 00:02:30,740
We can carry these lessons into this
frontier of artificial intelligence and

36
00:02:30,740 --> 00:02:35,140
emerging superintelligence. The
capability to act creates a need for

37
00:02:35,140 --> 00:02:37,880
infrastructure of trust
around that action.

38
00:02:38,680 --> 00:02:44,040
Again, going back to 2013,
the price of Bitcoin was $138.

39
00:02:44,100 --> 00:02:48,980
The Financial Times and all the media
was having this frenzy about the price

40
00:02:48,980 --> 00:02:50,920
going up and the price going down.

41
00:02:52,400 --> 00:02:54,240
I looked at it a different way.

42
00:02:54,400 --> 00:02:59,040
In 2014, February, I wrote an
article that said, with Mt.

43
00:02:59,040 --> 00:03:02,700
Gox effectively dead, Bitcoin officially
enters its third stage of evolution.

44
00:03:02,960 --> 00:03:07,780
I looked back and noticed that since
2009, Bitcoin had gone through mining,

45
00:03:08,720 --> 00:03:13,900
evolving liquidity exchange, and
then in 2014, the beginnings of

46
00:03:13,900 --> 00:03:16,500
institutionalization, financial services.

47
00:03:16,740 --> 00:03:21,040
But I also looked ahead and I said the
fourth stage of evolution is something

48
00:03:21,040 --> 00:03:22,960
I call machine to machine commerce.

49
00:03:23,240 --> 00:03:29,020
In 2014, I said, what happens when
software participates in economic

50
00:03:29,020 --> 00:03:34,720
activity on our behalf? So here we are
12 years later, that horizon is taking

51
00:03:34,720 --> 00:03:40,520
shape. We call that software, we
call them agents, systems that

52
00:03:40,520 --> 00:03:43,100
work, systems that
coordinate and transact.

53
00:03:43,220 --> 00:03:46,000
And the question of trust
has come along with us.

54
00:03:47,840 --> 00:03:50,660
The same gap is visible in business today.

55
00:03:50,820 --> 00:03:56,600
In McKinsey's 2025 survey, it was observed
that 88 % of respondents regularly

56
00:03:56,600 --> 00:04:01,180
use AI, but only 6 % met this
definition of an AI high performer.

57
00:04:01,420 --> 00:04:07,340
Where significant value with at least
5 % of EBIT contribution attributable

58
00:04:07,340 --> 00:04:10,320
to AI. So that's not a 94 % failure.

59
00:04:10,540 --> 00:04:14,860
That's actually a reminder that using
AI and changing a business are different

60
00:04:14,860 --> 00:04:18,840
achievements. Look Look at the outcomes
we're seeing in generative coding, in

61
00:04:18,840 --> 00:04:21,540
marketing, in sales, right?

62
00:04:21,540 --> 00:04:23,920
The generative use of AI is extraordinary.

63
00:04:23,960 --> 00:04:28,400
The tools are incredible. But business
operations depend on commitments that

64
00:04:28,400 --> 00:04:32,420
people have made. They depend on
the institution behind the work.

65
00:04:33,580 --> 00:04:35,780
What is the institution behind the work?

66
00:04:36,540 --> 00:04:41,820
It's the years of compounded
understanding and learning and legal and

67
00:04:41,820 --> 00:04:44,200
agreements. It's more than regulation.

68
00:04:44,280 --> 00:04:48,460
It's more than rules. At the end of
the day, an employee might just make a

69
00:04:48,460 --> 00:04:50,660
click, send an agreement, update.

70
00:04:51,040 --> 00:04:54,780
But behind that is everything that
the company has developed and learned,

71
00:04:54,820 --> 00:04:57,240
sometimes mistakes they
don't want to make again.

72
00:04:58,320 --> 00:05:02,440
So how does the agent know about
the institution behind the click?

73
00:05:02,960 --> 00:05:05,480
How does the agent know what matters?

74
00:05:05,840 --> 00:05:09,040
We need to carry this
context into agentic work.

75
00:05:11,680 --> 00:05:16,260
Zappos, great company built by a
great entrepreneur named Tony Hsieh.

76
00:05:16,460 --> 00:05:21,260
And they made a promise to consumers
called delivering happiness.

77
00:05:21,600 --> 00:05:24,280
They had a policy of
free shipping both ways.

78
00:05:24,480 --> 00:05:29,880
And that culture and that promise
led to a $1 .1 billion acquisition by

79
00:05:29,880 --> 00:05:30,930
Amazon.

80
00:05:31,380 --> 00:05:35,060
Now, I want you to imagine another
company that's trying to figure out what

81
00:05:35,060 --> 00:05:37,260
their promise is. What is their culture?

82
00:05:39,380 --> 00:05:43,540
2026, why don't we just ask the agent,
what should our customer service policy

83
00:05:43,540 --> 00:05:46,000
be? So the agent recommends a benefit.

84
00:05:46,120 --> 00:05:50,140
The customer experience and customer
expectations get entrenched.

85
00:05:50,320 --> 00:05:52,540
But guess what? The
company's losing money.

86
00:05:52,640 --> 00:05:56,600
The company's making mistakes. So
you don't know, unless you bring that

87
00:05:56,600 --> 00:06:00,420
institution behind, whether the agents
are actually advising you in the right

88
00:06:00,420 --> 00:06:06,260
direction. And this is what leads
to an expectation that we're

89
00:06:06,260 --> 00:06:08,660
creating about the new handoff.

90
00:06:09,120 --> 00:06:11,900
And a category we're
calling governed autonomy.

91
00:06:12,480 --> 00:06:17,440
So in the industrial economy, the
handoffs were physical, bills of lading,

92
00:06:17,540 --> 00:06:23,500
physical goods. In the digital economy
and software economy, you have APIs and

93
00:06:23,500 --> 00:06:28,180
tickets. In the agent -to -agent economy,
which is an essay here I published

94
00:06:28,180 --> 00:06:32,040
this year in July of 2026,
the handoff is evidentiary.

95
00:06:32,140 --> 00:06:34,740
It's not enough for the
agent to say, I did the work.

96
00:06:35,040 --> 00:06:37,900
It's not enough for ChatGPT
to just write your email.

97
00:06:37,900 --> 00:06:39,960
You need to know here's the evidence.

98
00:06:40,440 --> 00:06:41,980
Here's who authorized it.

99
00:06:42,260 --> 00:06:44,340
Here's how to replay that path again.

100
00:06:44,500 --> 00:06:48,820
So when the next agent or system or
human picks it up, there's that trail.

101
00:06:49,400 --> 00:06:53,400
Just like we needed a BitGo, we needed
an audit trail, human decisions and

102
00:06:53,400 --> 00:06:56,400
business processes. We need
that with the agentic economy.

103
00:06:56,680 --> 00:07:00,660
And if you build that, and if you
incorporate that into your business, you

104
00:07:00,660 --> 00:07:03,840
will outpace your competition
and you will win the market.

105
00:07:06,400 --> 00:07:07,940
So how do you get there?

106
00:07:14,320 --> 00:07:19,280
Well first you have to decode your
business for an agent so to you your

107
00:07:19,280 --> 00:07:24,600
business makes sense but to an agent
your business is a labyrinth and here we

108
00:07:24,600 --> 00:07:29,580
are in Rome i studied Latin for seven
years i studied classics at Harvard so

109
00:07:29,580 --> 00:07:34,460
you know i like the mythology i like the
story of Theseus who went to kill the

110
00:07:34,460 --> 00:07:38,160
Minotaur, who went into the labyrinth,
and the only way he was able to get out

111
00:07:38,160 --> 00:07:42,400
was when Ariadne provided him that
thread to come out and find his way out.

112
00:07:42,500 --> 00:07:44,720
So we provide that service at Provenonce.

113
00:07:44,720 --> 00:07:48,640
The first step of getting into AI is
decoding the DNA of your business, the

114
00:07:48,640 --> 00:07:53,780
institution. We provide this as a free
service on Provenonce on our website,

115
00:07:54,540 --> 00:07:58,640
provenonce.ai. The second
step is to run work.

116
00:07:59,100 --> 00:08:00,780
We talk about governed autonomy.

117
00:08:00,780 --> 00:08:05,380
Once you have a definition of the paths
and of the orchestrations you want to

118
00:08:05,380 --> 00:08:09,020
run in your business, you need to run
it in a place and in a system that

119
00:08:09,020 --> 00:08:12,880
generates evidence, that tells the
agents this is where you're allowed to

120
00:08:12,880 --> 00:08:17,140
work. Don't expand beyond
those roles and expectations.

121
00:08:17,460 --> 00:08:21,600
And the third step is independent
assurance of the work.

122
00:08:21,620 --> 00:08:25,760
And this is a project we're working on
in Bittensor for decentralized AI with

123
00:08:25,760 --> 00:08:26,810
subnet 87.

124
00:08:27,960 --> 00:08:29,760
We know about systems of record.

125
00:08:29,760 --> 00:08:33,180
We know about CRMs, HubSpot, Salesforce.

126
00:08:33,560 --> 00:08:35,720
That's where your business
facts are recorded.

127
00:08:35,820 --> 00:08:39,740
And we know about systems of
intelligence like Claude and ChatGPT.

128
00:08:40,000 --> 00:08:43,620
That's where the interpretation
and proposals can take place.

129
00:08:43,660 --> 00:08:47,380
But what's missing in the market and
what Provenonce is providing is the

130
00:08:47,380 --> 00:08:51,720
systems of operations.
Purpose, action, and evidence.

131
00:08:51,840 --> 00:08:57,140
How do people's agents and models work
together with existing systems into one

132
00:08:57,140 --> 00:08:58,340
defined operation?

133
00:08:59,320 --> 00:09:02,760
Now, whether you're building some of
that yourself, whether you're working

134
00:09:02,760 --> 00:09:07,260
with other providers, we welcome
collaboration, but we're defining what

135
00:09:07,260 --> 00:09:12,200
this future means. We're defining the
standards for how this future is going

136
00:09:12,200 --> 00:09:14,440
to move forward, just
like we did back at BitGo.

137
00:09:16,180 --> 00:09:21,700
The future is going to cost a lot if
you're looking to run it on centralized

138
00:09:21,700 --> 00:09:27,340
software. If you're looking to assure
and make sure AI is not checking its own

139
00:09:27,340 --> 00:09:30,640
homework and not making mistakes,
you're going to run out of tokens.

140
00:09:31,120 --> 00:09:34,080
And that's the opportunity that
we're working on with Bittensor.

141
00:09:34,900 --> 00:09:39,020
We're creating an open market for
assurance where miners in this

142
00:09:39,020 --> 00:09:44,480
decentralized AI ecosystem can compete
to create methods to make sure that AI

143
00:09:44,480 --> 00:09:48,560
is stewarding your business in the
right way, that you're becoming a Zappos

144
00:09:48,560 --> 00:09:50,220
and not the opposite.

145
00:09:51,540 --> 00:09:56,900
The business needs to decide what to
do next but the assurance can come from

146
00:09:56,900 --> 00:10:02,300
independent work in a
decentralized ecosystem so to recap

147
00:10:02,300 --> 00:10:08,220
governed autonomy is the solution it
is the path forward agents will do the

148
00:10:08,220 --> 00:10:12,700
work but trust will make the market if
you need to get into this business come

149
00:10:12,700 --> 00:10:17,080
talk to our team at the booth over here
if you want to run your orchestrations

150
00:10:17,080 --> 00:10:20,620
in business we would love to support
you and if you're building something

151
00:10:20,620 --> 00:10:24,940
interesting in the market let's
collaborate the future is abundance and we

152
00:10:24,940 --> 00:10:30,120
welcome everyone to help us develop the
trust infrastructure for this market

153
00:10:30,120 --> 00:10:31,170
thank you