WEBVTT
1
00:00:09.180 --> 00:00:12.960
My name's Dan. I lead the consumer product team at Stripe, where I work on Link.
2
00:00:14.280 --> 00:00:15.113
So for this talk,
3
00:00:15.380 --> 00:00:18.840
I know this is a sort of business conference and you all here for your startups
4
00:00:18.860 --> 00:00:22.920
and businesses, but for this talk, we're all going to be just regular consumers.
5
00:00:23.460 --> 00:00:24.480
So maybe get a show of hands.
6
00:00:24.620 --> 00:00:28.820
Who in the last three months has built some kind of personal agent for
7
00:00:28.840 --> 00:00:32.140
themselves, has used Claude or Cowork to make something for themselves,
8
00:00:32.200 --> 00:00:36.980
installed OpenClaw maybe? Raise of hands. Okay. I would say most of the room.
9
00:00:37.180 --> 00:00:39.260
That's great. Okay. So
10
00:00:40.940 --> 00:00:45.080
today I'm going to talk about the growing consumer trend of personal AI.
11
00:00:45.900 --> 00:00:49.340
As you saw from a lot of the content yesterday from the keynote from Patrick and
12
00:00:49.400 --> 00:00:50.820
Sam talking about OpenAI,
13
00:00:51.220 --> 00:00:55.700
the last three months have been a real inflection point for AI and including I
14
00:00:55.740 --> 00:00:57.760
think for consumers.
15
00:00:57.760 --> 00:01:01.060
I think this rise of personal AI is going to be really impactful
16
00:01:01.100 --> 00:01:04.680
for both how consumers use AI in the future and what it means for agentic
17
00:01:04.720 --> 00:01:09.460
commerce. Okay. So today we're going to cover three things.
18
00:01:09.900 --> 00:01:12.880
I'm going to start by digging more into this rise of personal AI.
19
00:01:13.300 --> 00:01:16.500
We'll look at a bunch of data from Link and some UXR we've done to kind of
20
00:01:16.520 --> 00:01:18.040
explain the trend and what we're seeing.
21
00:01:18.980 --> 00:01:22.760
Then I'm going to talk about Link's wallet for agents that we shipped yesterday.
22
00:01:23.720 --> 00:01:27.980
Give a demo of how that works and how we're helping consumers buy with agents.
23
00:01:28.420 --> 00:01:29.253
And then finally,
24
00:01:29.840 --> 00:01:33.140
Kate Jensen from Anthropic will join us on stage and we'll talk more about what
25
00:01:33.160 --> 00:01:36.320
this means for businesses and enterprises in the room. Okay,
26
00:01:36.560 --> 00:01:40.500
let's get in.
So if you'd asked me six months ago,
27
00:01:41.280 --> 00:01:44.780
how do most consumers experience AI? I would've said one or two ways.
28
00:01:45.440 --> 00:01:49.580
The first way is chat. You go to ChatGPT, Claude,
29
00:01:49.660 --> 00:01:52.760
you ask it a question, it gives you an answer. You're building a recipe,
30
00:01:52.860 --> 00:01:55.940
you're doing some research. It's kind of a back-and-forth chat interface.
31
00:01:56.960 --> 00:01:59.980
And the second way is AI coding.
32
00:02:00.280 --> 00:02:03.940
More and more people using AI to code, certainly at Stripe,
33
00:02:04.340 --> 00:02:08.940
it's the primary way we build Stripe today. Over the last,
34
00:02:09.080 --> 00:02:12.040
I'd say three or four months, these two worlds have started to really,
35
00:02:12.460 --> 00:02:15.100
really intersect.
And they've intersected in two ways.
36
00:02:16.460 --> 00:02:19.340
The first is the rise of OpenClaw.
37
00:02:20.140 --> 00:02:24.580
Now what OpenClaw did was basically take all the power of the AI coding agents
38
00:02:24.960 --> 00:02:29.360
and bring it to a consumer experience. And it was one of the fastest growing,
39
00:02:29.480 --> 00:02:33.420
I think, consumer AI products ever. On GitHub,
40
00:02:33.760 --> 00:02:36.740
within 60 days, it's the most starred software project ever.
41
00:02:36.840 --> 00:02:39.760
It has more stars than React, more stars than the Linux operating system.
42
00:02:40.380 --> 00:02:42.300
It was a huge sort of moment.
43
00:02:42.700 --> 00:02:47.080
And the reason is it gave consumers a feel for what AI can really do.
44
00:02:47.320 --> 00:02:51.140
It can do long-running tasks. It can open up a browser. It can do things.
45
00:02:51.200 --> 00:02:53.660
It can have access to your email, your desktop.
46
00:02:53.960 --> 00:02:58.520
It can do a whole bunch more than just Q&A back and forth. Now,
47
00:02:58.660 --> 00:03:02.580
the second way that I think we've seen personal AI take off is more and more
48
00:03:02.640 --> 00:03:04.820
consumers, as I said at the start of this session,
49
00:03:05.300 --> 00:03:09.340
using Claude and Codex and things to build their own apps,
50
00:03:09.400 --> 00:03:11.680
to build their own agents. Last weekend,
51
00:03:12.860 --> 00:03:15.360
I didn't like how my family calendar was running,
52
00:03:15.420 --> 00:03:17.380
and so I had Claude in a couple of hours,
53
00:03:17.460 --> 00:03:19.460
build me a new calendar app for my family.
54
00:03:20.660 --> 00:03:24.580
I live in DC where nobody works in tech. Nobody knows what I do.
55
00:03:26.240 --> 00:03:30.060
But yet all my friends who are lawyers and teachers are in the evenings vibe
56
00:03:30.080 --> 00:03:34.000
coding, some little app, some game, whereas a year ago they wouldn't have been.
57
00:03:36.600 --> 00:03:39.280
So I think this tweet from Andrej Karpathy says it really well.
58
00:03:39.700 --> 00:03:40.540
For the first time,
59
00:03:40.660 --> 00:03:45.120
I think consumers are starting to really experience AI in the way that many of
60
00:03:45.160 --> 00:03:47.580
us who've been using it for coding have been doing.
61
00:03:48.160 --> 00:03:51.360
AI that can actually do things, not just question and answer,
62
00:03:51.720 --> 00:03:56.160
but take action on our behalf, do tasks, think for itself,
63
00:03:56.380 --> 00:04:00.880
and be a kind of personal assistant. Okay.
64
00:04:02.860 --> 00:04:05.560
It's not just individuals, I think, spinning up OpenClaw,
65
00:04:05.620 --> 00:04:09.820
though that's been a huge trend. The same idea of personal AI doing more,
66
00:04:09.920 --> 00:04:12.440
I think, is spreading across all of the ecosystem.
67
00:04:13.200 --> 00:04:16.900
So some of the big AI companies like Anthropic launch Claude Cowork.
68
00:04:17.280 --> 00:04:18.860
So the idea of bringing the power of Claude,
69
00:04:18.920 --> 00:04:23.220
but to the desktop so it can do a whole bunch more there. OpenAI, of course,
70
00:04:23.280 --> 00:04:27.240
acquired OpenClaw. Perplexity in February, launched Perplexity Computer.
71
00:04:27.720 --> 00:04:28.040
Same idea.
72
00:04:28.040 --> 00:04:32.240
How can we bring the power of these models to the desktop where it can have so
73
00:04:32.340 --> 00:04:34.400
much more access to your data?
74
00:04:36.440 --> 00:04:41.120
And we then have all of the sort of agent and app builder platforms,
75
00:04:41.320 --> 00:04:45.040
Replit, Lovable, Bolt, Vercel, have all seen huge growth in the last few months.
76
00:04:45.100 --> 00:04:47.900
Consumers just building their own little apps on these platforms.
77
00:04:48.940 --> 00:04:52.260
And startups too. I've spoken to dozens over the last few months.
78
00:04:52.400 --> 00:04:55.900
There's so many new startups building personal assistance, travel planning apps,
79
00:04:56.140 --> 00:04:59.400
chief-of-staff type operating systems for consumers.
80
00:05:00.020 --> 00:05:04.640
A third of the Forbes AI 50 are building some kind of app
81
00:05:04.740 --> 00:05:07.300
agent builders.
Now, to back this up,
82
00:05:07.960 --> 00:05:11.600
we have some great insights from Link. If you haven't heard of Link,
83
00:05:11.720 --> 00:05:13.040
it's our consumer wallet.
84
00:05:14.940 --> 00:05:17.780
If you ever buy a subscription from OpenAI or Anthropic,
85
00:05:17.840 --> 00:05:21.020
you'll see it there as a way that you can pay. It's on a bunch of merchants.
86
00:05:21.880 --> 00:05:24.600
And we have some pretty unique insights, I think, into this trend.
87
00:05:25.280 --> 00:05:29.960
So Link is on 94% of the consumer companies in the Forbes AI
88
00:05:30.100 --> 00:05:34.560
50, and we now have over 250 million consumers in the Link network.
89
00:05:34.560 --> 00:05:38.360
So we get this kind of vantage point where we can see how consumers are using
90
00:05:38.400 --> 00:05:39.260
these AI products.
91
00:05:40.600 --> 00:05:45.560
So I'm looking here at how much a consumer spends on AI in
92
00:05:45.600 --> 00:05:47.960
a given month. So this is consumers, not businesses,
93
00:05:48.080 --> 00:05:52.200
this is just individuals paying with Link for AI. So looking here at the P90,
94
00:05:52.820 --> 00:05:55.640
so this is the top 10% of Link users.
95
00:05:56.160 --> 00:05:59.200
You can see a fairly steady increase in spend over the last couple of years.
96
00:05:59.700 --> 00:06:02.760
And then starting this year, a huge inflection point,
97
00:06:03.360 --> 00:06:07.280
spending over $350 a month on AI. So this is not just one subscription,
98
00:06:07.660 --> 00:06:09.080
this is multiple subscriptions.
99
00:06:09.440 --> 00:06:13.860
This is paying for the pro plans from the big labs to get really great coding
100
00:06:13.900 --> 00:06:16.500
ability. If you look at the P75,
101
00:06:17.260 --> 00:06:20.020
so the top 25% of Link consumers, same thing,
102
00:06:20.700 --> 00:06:23.780
42% year over growth in spend on
103
00:06:23.900 --> 00:06:28.640
AI. And the P50, the median consumer,
104
00:06:28.840 --> 00:06:33.540
also still a lower amount, but getting up to the $75 a month spent on AI.
105
00:06:33.700 --> 00:06:36.780
So again, we're seeing this trend start obviously with the power users,
106
00:06:36.840 --> 00:06:39.300
but it's coming down to the median consumer too,
107
00:06:39.380 --> 00:06:42.020
where they're really starting to spend a lot of money every month on AI.
108
00:06:44.060 --> 00:06:48.680
If I just click on the builder platform, so how much they're spending on Replit,
109
00:06:48.720 --> 00:06:49.760
Lovable, Bolt, Vercel,
110
00:06:49.820 --> 00:06:52.800
places where consumers can go to build little apps and build projects,
111
00:06:53.160 --> 00:06:55.300
they're spending 5x what they were a year ago.
112
00:06:57.980 --> 00:06:58.813
So in summary,
113
00:06:58.940 --> 00:07:02.460
the trend I think we're seeing is that more and more consumers are running their
114
00:07:02.500 --> 00:07:04.600
own personal agents. They're building their own apps.
115
00:07:04.700 --> 00:07:06.800
AI is no longer just a chat interface.
116
00:07:06.860 --> 00:07:10.200
It is a thing they can actually experience and have it do tasks for them and
117
00:07:10.560 --> 00:07:13.520
sort of operate on their behalf and they're willing to spend.
118
00:07:13.520 --> 00:07:15.740
They're spending more than ever on AI.
119
00:07:16.800 --> 00:07:19.600
And so that brings us to what they want it to do.
120
00:07:20.740 --> 00:07:22.960
They're sending agents out to browse and buy,
121
00:07:23.400 --> 00:07:26.760
and now we're increasingly seeing them want to actually transact on behalf
122
00:07:28.300 --> 00:07:33.220
of the consumer.
So let's talk a bit about how do we actually get these
123
00:07:33.300 --> 00:07:36.620
AI assistants to actually transact. So first,
124
00:07:36.680 --> 00:07:40.440
let's kind of lay out the ecosystem. We have consumers, the people,
125
00:07:41.040 --> 00:07:43.780
us in this room as consumers who want our agents to be able to go buy things.
126
00:07:44.820 --> 00:07:45.800
We have the agent builders.
127
00:07:46.040 --> 00:07:48.800
So any of you in the room who are building personal assistants,
128
00:07:48.860 --> 00:07:52.980
building AI apps for consumers that you want to be able to go and buy on behalf
129
00:07:53.040 --> 00:07:55.300
of the consumer. And then we have the sellers,
130
00:07:55.740 --> 00:08:00.020
the places that the consumers and agents want to go buy from.
When we thought
131
00:08:00.040 --> 00:08:01.320
about Link, we thought,
132
00:08:01.400 --> 00:08:05.240
how can we actually address the needs of all three of these groups?
133
00:08:05.240 --> 00:08:09.380
How can we build a product that works for everybody? So what do they each want?
134
00:08:10.000 --> 00:08:12.360
Consumers want their preferred payment methods.
135
00:08:13.520 --> 00:08:16.720
I don't want to spin up a new crypto wallet and figure out what Solana is and
136
00:08:16.780 --> 00:08:19.760
gas fees in order to go fund my agent. I just have a credit card.
137
00:08:19.820 --> 00:08:22.420
I want it to be able to use my credit card and buy things for me.
138
00:08:23.720 --> 00:08:26.420
They want controls and security. As great as AI is,
139
00:08:26.680 --> 00:08:28.140
I still don't trust it with my bank account.
140
00:08:28.380 --> 00:08:31.800
I would like there to be some control in place that I'm going to be protected.
141
00:08:33.860 --> 00:08:35.140
Those of you that are building agents,
142
00:08:35.880 --> 00:08:38.800
what we've been hearing a lot of is they want to purchase across the internet,
143
00:08:38.900 --> 00:08:41.820
right? You don't just want to be able to buy from places that support certain
144
00:08:41.840 --> 00:08:42.600
protocols.
145
00:08:42.600 --> 00:08:45.000
You want it to be able to buy from all the places that consumers want to buy.
146
00:08:45.600 --> 00:08:47.440
And you want to be able to stay out of the funds flow.
147
00:08:48.360 --> 00:08:50.400
Stripe's really good at complex money movement.
148
00:08:51.340 --> 00:08:54.560
You guys don't want to be dealing with money transmission and figuring out how
149
00:08:54.560 --> 00:08:57.520
to move money around. And then there's the sellers.
150
00:08:58.280 --> 00:09:00.040
Really excited about agentic commerce,
151
00:09:00.240 --> 00:09:02.720
but don't want to keep making massive integration changes,
152
00:09:02.920 --> 00:09:06.040
figuring out how to support every new protocol. And at the same time,
153
00:09:06.320 --> 00:09:07.480
want Agentic commerce,
154
00:09:07.680 --> 00:09:11.400
but from good humans with good agents that you can understand,
155
00:09:11.720 --> 00:09:14.960
not just a sort of sprawl of bots that are hard to understand.
156
00:09:16.600 --> 00:09:20.640
So that's why we launched yesterday, Link's agent wallet.
157
00:09:21.400 --> 00:09:22.800
It's a programmable wallet.
158
00:09:23.520 --> 00:09:27.280
It gives consumers the ability to grant agents the ability to spend on their
159
00:09:27.320 --> 00:09:31.480
behalf. It can produce one-time use virtual cards. So a 16-digit card number,
160
00:09:31.640 --> 00:09:36.120
you can plug into a website checkout, or it can create Shared Payment Tokens.
161
00:09:36.440 --> 00:09:40.040
So if you want to pay with a Machine Payment Protocol, it can produce those two.
162
00:09:40.800 --> 00:09:45.240
The human today approves every transaction and it never exposes the raw card
163
00:09:45.360 --> 00:09:49.440
details. So if you're a consumer, you're protected. If you're an agent builder,
164
00:09:49.600 --> 00:09:52.920
super easy to integrate and you can enable a whole bunch of commerce.
165
00:09:55.240 --> 00:09:58.680
Okay. Let me show you how this works. I'm going to jump to a demo.
166
00:10:00.560 --> 00:10:05.400
Oh, great. Okay. So before I get to the demo, yes, we launched it yesterday,
167
00:10:05.940 --> 00:10:08.120
link.com/agents. Go check it out.
168
00:10:08.640 --> 00:10:11.800
If you are building an agent and want to integrate it with your agent so it can
169
00:10:11.880 --> 00:10:15.680
spend on behalf of consumers, come check out the GitHub, Stripe, Link CLI.
170
00:10:16.920 --> 00:10:19.280
Okay. So I'm going to do two things.
171
00:10:19.440 --> 00:10:23.320
I'm going to first give you an end-to-end demo of an app I built last weekend to
172
00:10:23.440 --> 00:10:26.160
kind of show you how Link fits in and how Link enables payments.
173
00:10:26.560 --> 00:10:29.560
And then I'm going to do a deep dive for a couple of minutes on what we call a
174
00:10:29.600 --> 00:10:30.320
spend request,
175
00:10:30.320 --> 00:10:34.880
which is kind of the core new thing that we've created here.
So for this demo,
176
00:10:35.080 --> 00:10:39.880
I'm going to spin up Claude. That's my agent here. And before I get to the demo,
177
00:10:40.320 --> 00:10:43.880
I should give you the heads up. I have three daughters,
178
00:10:44.160 --> 00:10:45.840
and they love getting physical mail.
179
00:10:46.240 --> 00:10:48.800
It's just like a really exciting thing when they open,
180
00:10:48.880 --> 00:10:53.760
the things come in the mailbox, and they see their name on it. I love dad jokes.
181
00:10:54.200 --> 00:10:57.600
I'm a dad, and I just love them. My daughters hate them, but I love them.
182
00:10:57.760 --> 00:10:58.480
So I was like,
183
00:10:58.480 --> 00:11:03.040
what if I could mail my daughters a dad joke just by pushing a button with my
184
00:11:03.160 --> 00:11:07.680
agent? So with that in mind, let's launch our dad jokes agent.
185
00:11:09.800 --> 00:11:12.960
Okay. So there's going to be three parts to this. It's going to create the joke.
186
00:11:13.640 --> 00:11:14.960
It's going to mail it,
187
00:11:15.080 --> 00:11:18.080
and it's going to use Link to help me pay for the postage.
188
00:11:18.320 --> 00:11:21.920
So the first thing it's done is loaded our "create payment credentials" skill.
189
00:11:22.200 --> 00:11:26.280
This is the Link skill that tells the agent how it can spend with Link.
190
00:11:27.000 --> 00:11:30.200
Now it's asking me what topic would you like the dad joke to be about and who
191
00:11:30.280 --> 00:11:31.113
should I send it to?
192
00:11:31.440 --> 00:11:36.440
I'm going to go with space for Iris, my middle daughter.
193
00:11:37.800 --> 00:11:42.440
Okay. So it's going to go hopefully find me a very amusing dad joke. Great.
194
00:11:42.640 --> 00:11:44.200
It's calling... I didn't notice.
195
00:11:44.320 --> 00:11:47.920
There's actually a whole public API of dad jokes if anyone wants to go integrate
196
00:11:48.200 --> 00:11:50.240
themselves.
Okay. I have three choices.
197
00:11:52.120 --> 00:11:56.080
I'm going to go with option one. How do you organize a space party? You planet.
198
00:11:56.600 --> 00:12:00.000
Okay. Option one. Yeah, that was the correct groan. Thank you.
199
00:12:01.400 --> 00:12:04.160
Okay. Now it's going to... Yep, send it. That's the right
200
00:12:04.200 --> 00:12:09.200
address.
201
00:12:09.860 --> 00:12:12.280
And it's going to make the PDF.
Now, while it's pulling that up,
202
00:12:13.000 --> 00:12:16.480
I'm just going to show you... We're using this company PostalForm.
203
00:12:16.540 --> 00:12:19.380
They're a great Stripe user, basically lets you upload a PDF and mail it.
204
00:12:20.140 --> 00:12:22.620
What they've done is integrate with our Machine Payments Protocol.
205
00:12:23.140 --> 00:12:27.880
And so basically you can, with an agent, send mail. It's a super great product.
206
00:12:28.340 --> 00:12:31.580
The founders here, maybe in the room, it's a super cool thing.
207
00:12:31.780 --> 00:12:35.720
It's just so much fun to be able to send mail just by uploading a PDF.
Okay.
208
00:12:36.000 --> 00:12:39.080
So if I go back to my agent, it's made the PDF. Let me just check it out.
209
00:12:40.860 --> 00:12:45.400
Yep. That looks like a great thing to mail. So yep, it looks good.
210
00:12:45.460 --> 00:12:46.360
Let's go ahead and mail it.
211
00:12:51.860 --> 00:12:52.060
Now,
212
00:12:52.060 --> 00:12:56.460
what it's doing here is it's kicking off the Machine Payment Protocol endpoint
213
00:12:56.480 --> 00:12:57.313
with PostalForm.
214
00:12:57.320 --> 00:13:01.480
So this is beginning the payment process using Machine Payment Protocols. Now,
215
00:13:01.540 --> 00:13:01.940
as I said,
216
00:13:01.940 --> 00:13:06.620
Link can support one-time-use cards or we can use Shared Payment Tokens for
217
00:13:07.420 --> 00:13:08.860
this authenticated, this flow here.
218
00:13:09.500 --> 00:13:12.000
Now it's calling Link and it's checking if I'm logged into Link,
219
00:13:12.160 --> 00:13:16.600
and it's realizing I'm not. And so in a second, it should prompt me to log in.
220
00:13:19.180 --> 00:13:21.860
So it's decoding the Machine Payment Protocol response,
221
00:13:22.160 --> 00:13:24.740
and it's generating me a link to log in here.
222
00:13:24.820 --> 00:13:29.640
So this is the link I need to basically grant access for my agent to
223
00:13:29.740 --> 00:13:34.660
use Link. So I'm going to pull up that URL and log in.
224
00:13:34.760 --> 00:13:37.500
Okay. So I can connect an agent. I can make agentic payments.
225
00:13:37.600 --> 00:13:41.340
I approve every spend and my card details remain protected.
226
00:13:41.680 --> 00:13:45.460
That sounds good.
I'm going to continue, confirm the passphrase.
227
00:13:46.900 --> 00:13:50.880
And yet, my dad jokes agent on my Stripe laptop wants access to my Link account.
228
00:13:50.940 --> 00:13:54.400
It's going to get my account details and it's going to get the ability to spend.
229
00:13:54.700 --> 00:13:59.660
Great. Let's do it. Done. Okay. Now that I'm logged in,
230
00:14:00.480 --> 00:14:02.400
it should be able to complete the payment.
231
00:14:04.500 --> 00:14:06.320
So it's going to create what we call a spend request.
232
00:14:06.540 --> 00:14:08.040
I'll come back to this in a minute,
233
00:14:08.120 --> 00:14:11.100
but this is the way the agent basically starts to set up the contract with the
234
00:14:11.160 --> 00:14:12.360
human to say, "Hey,
235
00:14:12.500 --> 00:14:14.980
I'd like to be able to spend this much money on this merchant.
236
00:14:15.400 --> 00:14:19.880
Can you get me a payment credential I can go use to complete that transaction?"
237
00:14:19.880 --> 00:14:21.950
So here we're seeing it create the spend request.
238
00:14:23.810 --> 00:14:28.670
And in a second, if I come to my app, let me unlock my phone.
239
00:14:30.830 --> 00:14:34.870
I should have a push notification from Link coming in a second. Great.
240
00:14:35.370 --> 00:14:38.230
So you can see there, as well as a message from my kids' school,
241
00:14:39.030 --> 00:14:43.190
a payment request for $3.40. So let me tap on that, open up the Link app,
242
00:14:44.550 --> 00:14:47.840
and I can see, okay, my agent here is looking to spend $3.40.
243
00:14:48.270 --> 00:14:50.770
I can see the merchant it's looking to buy from,
244
00:14:51.350 --> 00:14:54.010
and it's giving me some context on what it's trying to achieve with this
245
00:14:54.070 --> 00:14:57.650
purchase. I'm just going to use my United Club card, which looks good,
246
00:14:58.190 --> 00:14:59.490
and I'm going to prove it with my face ID.
247
00:15:04.830 --> 00:15:07.750
Okay. Now that's done. Back to our agent.
248
00:15:08.550 --> 00:15:10.190
It should be able to now get the spend request,
249
00:15:11.070 --> 00:15:13.410
encode the PDF and the letters on the way.
250
00:15:20.870 --> 00:15:22.850
Now, while it's sending it off, I should say again,
251
00:15:24.110 --> 00:15:27.230
I built this into my own personal agent here. If you are building an agent,
252
00:15:27.610 --> 00:15:32.090
you're building your own grocery shopping app, personal assistant,
253
00:15:32.150 --> 00:15:32.983
that kind of thing,
254
00:15:33.010 --> 00:15:36.890
you can absolutely integrate Link into that so that your consumers can spend
255
00:15:37.370 --> 00:15:41.430
with your agent. Okay. It's got the 200 back, so in a second,
256
00:15:41.490 --> 00:15:45.170
it should be complete. Great. Letter's on the way.
257
00:15:45.330 --> 00:15:46.550
I can pull up the confirmation.
258
00:15:48.290 --> 00:15:52.330
It's been sent and I can even see, it's put a little space rocket.
259
00:15:53.350 --> 00:15:56.630
Great job, Claude. Okay. Thank you. Thank you.
260
00:15:58.610 --> 00:15:59.530
Just to prove it's real,
261
00:15:59.610 --> 00:16:03.570
this is one I sent last weekend that my daughter received and she groaned
262
00:16:03.810 --> 00:16:07.370
heavily when she got it.
Okay. So let me for a sec,
263
00:16:07.490 --> 00:16:11.230
just deep dive more into what this spend request is because that's the key thing
264
00:16:11.250 --> 00:16:14.270
that we've built here. So I have a couple of examples.
265
00:16:14.990 --> 00:16:17.630
This is using the Link CLI, which is what we have on GitHub.
266
00:16:17.710 --> 00:16:20.170
This is what your agent calls to be able to spend,
267
00:16:20.570 --> 00:16:24.570
and it creates a spend request. And so it has to pass a few things.
268
00:16:24.830 --> 00:16:27.250
The first is the payment method. So from the Link wallet,
269
00:16:27.350 --> 00:16:30.070
it can select which card or bank account it wants to use to pay.
270
00:16:30.130 --> 00:16:33.730
And then it has a choice of credential types. So it can say, "Hey,
271
00:16:33.790 --> 00:16:37.690
I need a one-time used card." If it wants to go fill out a regular checkout form
272
00:16:37.730 --> 00:16:40.690
on the internet, or it can create a shared payment token.
273
00:16:41.030 --> 00:16:44.270
So if it wants to go be able to plug into an MPP server,
274
00:16:44.350 --> 00:16:46.530
it can create a shared payment token.
275
00:16:46.530 --> 00:16:49.250
We're going to be adding support for more types of credentials in the future,
276
00:16:49.350 --> 00:16:52.470
so Link can support crypto-based things,
277
00:16:52.550 --> 00:16:56.670
stablecoin-based payments as well.
The next thing it needs is a merchant name
278
00:16:56.690 --> 00:17:01.450
and URL. This is me ordering flowers for my wife for Mother's Day,
279
00:17:01.910 --> 00:17:06.220
which is in a couple of weeks. And so you pass the merchant name in the URL.
280
00:17:06.550 --> 00:17:08.070
Now this starts to build the contract.
281
00:17:08.150 --> 00:17:11.720
What we then do on Link is make sure that the credential the agent gets back
282
00:17:11.870 --> 00:17:16.470
can't be used outside of what is being specified here. You then have an amount,
283
00:17:17.230 --> 00:17:20.390
$84.98. Anything above that, we will decline.
284
00:17:20.530 --> 00:17:25.090
So the agent can't spend more than you agree. And then a context.
285
00:17:25.770 --> 00:17:29.290
We ask the agent to provide a bit of information about what it's buying and from
286
00:17:29.330 --> 00:17:31.510
where so that you as a human can approve.
287
00:17:32.690 --> 00:17:35.990
So what happens is once the spend request is made, the human confirms,
288
00:17:36.590 --> 00:17:40.970
and then the credential that gets back is scoped to exactly what they specified
289
00:17:41.030 --> 00:17:41.990
here. So this way,
290
00:17:42.050 --> 00:17:44.530
the agent can't suddenly decide to go off and spend all your money.
291
00:17:44.750 --> 00:17:48.990
It can't just go do things that you didn't allow it to do. So yeah,
292
00:17:49.230 --> 00:17:53.090
that's how we create a spend request.
Okay.
293
00:17:54.010 --> 00:17:55.690
In the last about 10, 15 minutes,
294
00:17:56.390 --> 00:18:00.170
I would love to welcome to the stage Kate Jensen Anthropic to talk more about
295
00:18:00.310 --> 00:18:02.550
how businesses are adapting here.
296
00:18:09.680 --> 00:18:11.010
Thank you so much for being here.
297
00:18:11.270 --> 00:18:12.103
Thanks for having me.
298
00:18:13.650 --> 00:18:18.630
Okay. So to get right into it, in your view from Anthropic,
299
00:18:20.130 --> 00:18:23.050
what are the best companies,
300
00:18:23.130 --> 00:18:26.510
the companies that are embracing AI in the loop with their customers or in their
301
00:18:26.530 --> 00:18:27.363
operations,
302
00:18:28.370 --> 00:18:31.150
what are the best companies doing right now with AI and their customers?
303
00:18:32.310 --> 00:18:33.650
I have a few serious answers,
304
00:18:33.710 --> 00:18:37.450
but one of them is you were just using your Stripe account on Claude Code to
305
00:18:38.030 --> 00:18:38.863
send that joke.
306
00:18:39.120 --> 00:18:43.770
And actually the best things companies are doing are
307
00:18:43.810 --> 00:18:46.330
encouraging people to just use the technology.
308
00:18:47.610 --> 00:18:50.890
It was funny to see the little disclaimer across the top as you were like,
309
00:18:50.890 --> 00:18:55.350
"I think it'll be a space joke today." But the reality is I talk to
310
00:18:55.430 --> 00:18:57.770
CTOs all the time who say,
311
00:18:57.770 --> 00:19:01.570
"I look at our power users and 99% of
312
00:19:01.610 --> 00:19:06.290
them are using this technology for stuff at work." And there's always someone
313
00:19:06.330 --> 00:19:10.770
who's writing a novel in their spare time and they're like, that's okay.
314
00:19:11.410 --> 00:19:14.630
How do I just help people get used to this technology and understand how it can
315
00:19:14.670 --> 00:19:15.630
change their day to day,
316
00:19:15.690 --> 00:19:19.170
whether it's at work or outside of work because that sort of adoption is what
317
00:19:19.250 --> 00:19:20.190
really, really matters.
318
00:19:20.530 --> 00:19:22.650
Yeah. If you think about
319
00:19:24.570 --> 00:19:28.990
this huge shift from consumers in AI and enterprises and business trying to keep
320
00:19:29.130 --> 00:19:31.850
up, what do you think people are getting wrong?
321
00:19:32.330 --> 00:19:35.270
If you were to speak to 50 of your top users,
322
00:19:35.710 --> 00:19:38.330
what is the kind of common misconception where you think, "Ah,
323
00:19:38.410 --> 00:19:40.110
they're not quite thinking about this the right way.
324
00:19:40.170 --> 00:19:43.210
I wish I could just reframe this whole thing for them." What is the
325
00:19:43.250 --> 00:19:44.083
misconception?
326
00:19:44.330 --> 00:19:46.690
Two things. One, they treat it like an IT rollout.
327
00:19:47.710 --> 00:19:51.230
This isn't necessarily just a new piece of software that folks are using.
328
00:19:51.290 --> 00:19:55.390
It should and will fundamentally change how most of their teams operate and how
329
00:19:55.430 --> 00:19:59.910
you build products. And two, they outsource it to someone who,
330
00:20:00.450 --> 00:20:01.890
and I will say this very kindly,
331
00:20:02.030 --> 00:20:04.590
many of you may be chief AI officers in this room.
332
00:20:05.110 --> 00:20:08.590
There is an important role for someone to be accountable and
333
00:20:10.040 --> 00:20:14.890
hold an organization accountable to KPIs with AI, but the CEO,
334
00:20:15.090 --> 00:20:18.070
the CTO, the entire executive team have to be fully bought in.
335
00:20:18.710 --> 00:20:22.070
So you kind of want to see leadership from the very top down embracing AI,
336
00:20:22.650 --> 00:20:24.270
driving that. Yeah, that makes a lot of sense.
337
00:20:26.550 --> 00:20:29.490
Since you're the ones building this technology day in, day out,
338
00:20:30.370 --> 00:20:34.950
and we are all both employees and individuals in the evenings,
339
00:20:35.870 --> 00:20:38.410
how are people at Anthropic using AI each day?
340
00:20:38.650 --> 00:20:43.210
And I'm curious for like the kind of like more avant-garde things that teams are
341
00:20:43.230 --> 00:20:44.063
doing with AI.
342
00:20:44.350 --> 00:20:47.630
I can tell you about the percent of code that's now written with Claude Code.
343
00:20:47.710 --> 00:20:49.750
Thank you for using Claude Code for that, by the way.
344
00:20:51.230 --> 00:20:54.370
But there are a few just standout examples throughout the organization that I
345
00:20:54.390 --> 00:20:59.350
think are pretty interesting. Our legal team, many of you feel this deeply,
346
00:20:59.470 --> 00:21:02.550
I'm sure, is wildly understaffed.
347
00:21:04.870 --> 00:21:08.230
Anyone who touches the legal team,
348
00:21:08.290 --> 00:21:11.050
probably feels like they're not being prioritized fast enough.
349
00:21:11.110 --> 00:21:15.310
And what they've done is actually used Claude to help anyone who interacts with
350
00:21:15.330 --> 00:21:17.810
them, understand where you stand in the queue,
351
00:21:17.870 --> 00:21:21.230
exactly who's going to help you and about how long it will take. It's amazing.
352
00:21:21.390 --> 00:21:23.790
It makes the whole experience so much more user-friendly.
353
00:21:24.830 --> 00:21:28.150
I had someone present on my team just yesterday, actually,
354
00:21:29.030 --> 00:21:33.370
a woman came and she's a front-line leader in our commercial
355
00:21:33.450 --> 00:21:38.350
organization and she has about 2,000 accounts in her
356
00:21:38.410 --> 00:21:41.110
book and she has a version of Claude,
357
00:21:41.170 --> 00:21:45.730
an agent that she herself built managing about 800 of those accounts entirely on
358
00:21:45.790 --> 00:21:47.190
its own.
359
00:21:47.190 --> 00:21:50.470
She was a little nervous to tell me that she was effectively outsourcing a lot
360
00:21:50.510 --> 00:21:52.550
of her job, but at the same time,
361
00:21:52.610 --> 00:21:56.790
it was just unbelievable to see what someone who is a salesperson could actually
362
00:21:56.810 --> 00:21:58.710
go build on their own and how they're doing that.
363
00:21:58.910 --> 00:22:02.390
Yeah. I'm curious, because I think at Stripe,
364
00:22:02.830 --> 00:22:07.110
we're a culture where AI is everywhere, we use it all the time, but
365
00:22:09.440 --> 00:22:10.273
how do you...
366
00:22:10.390 --> 00:22:13.310
Obviously you mentioned the top-down part where executives are sort of pushing
367
00:22:13.330 --> 00:22:15.190
it down, but I'm curious,
368
00:22:15.810 --> 00:22:18.810
how does Anthropic create the culture and the freedom where people are willing
369
00:22:18.850 --> 00:22:19.750
to just try AI?
370
00:22:19.810 --> 00:22:22.430
Even if it maybe doesn't work the first time and it produces worse results than
371
00:22:22.490 --> 00:22:23.490
had they done it manually.
372
00:22:24.810 --> 00:22:28.350
How do businesses create that sort of culture with their employees?
373
00:22:28.830 --> 00:22:31.690
I think culture is so much in large part tone at the top,
374
00:22:31.930 --> 00:22:35.470
but then being very deliberate about what you hope to get out of the technology.
375
00:22:36.130 --> 00:22:41.110
You asked me earlier what a company's doing well, and this is very general,
376
00:22:41.170 --> 00:22:45.050
but we see companies bucketing their AI investments into one of three buckets,
377
00:22:45.350 --> 00:22:46.910
sometimes doing all three at the same time.
378
00:22:47.790 --> 00:22:50.130
One is just give tools to every person in the company.
379
00:22:50.430 --> 00:22:53.430
Claude Code mostly for developers, though not exclusively.
380
00:22:53.430 --> 00:22:57.070
Claude Cowork increasingly for every other knowledge worker.
381
00:22:57.810 --> 00:23:01.050
The second big bucket is how do you change how you operate internally?
382
00:23:01.830 --> 00:23:02.770
So you think about,
383
00:23:03.050 --> 00:23:07.370
we published a study where Goldman Sachs has completely rewritten their KYC
384
00:23:07.570 --> 00:23:11.450
process that touches Stripe in a lot of ways. You are too.
385
00:23:11.950 --> 00:23:13.350
Stripe is doing a whole lot of this.
386
00:23:13.430 --> 00:23:15.570
How do you think about risk management yourselves?
387
00:23:15.630 --> 00:23:17.390
How do you think about merchant onboarding?
388
00:23:17.390 --> 00:23:20.890
How do you operate the teams that look at all of those things?
389
00:23:21.310 --> 00:23:25.970
Can you completely rewrite what maybe used to be 200 steps and make it 30 and
390
00:23:26.010 --> 00:23:27.610
have agents do 15 of them?
391
00:23:27.970 --> 00:23:31.150
There's so many opportunities for companies to be run more efficiently.
392
00:23:31.810 --> 00:23:34.890
And then the third, and this is something Stripe is also investing in a lot,
393
00:23:34.970 --> 00:23:37.210
is how do you build AI into the products that you're shipping?
394
00:23:38.150 --> 00:23:42.870
And my answer to how does it work well kind of depends on which one of those
395
00:23:42.910 --> 00:23:45.730
you're doing. In the first bucket, it's find the champions,
396
00:23:45.790 --> 00:23:50.050
find the people who are just amazing at this and really adopt it and celebrate
397
00:23:50.190 --> 00:23:51.190
how creative they're being.
398
00:23:51.950 --> 00:23:55.650
The second is have the right organizational leaders really running it and
399
00:23:55.670 --> 00:23:59.270
learning the results. And the third is just try it and experiment.
400
00:23:59.510 --> 00:24:01.040
Yeah. So something you just said there,
401
00:24:03.370 --> 00:24:07.710
I think we're seeing consumers clearly more and more familiar with AI.
402
00:24:07.770 --> 00:24:10.550
It's no longer a thing they don't understand.
403
00:24:10.610 --> 00:24:13.110
It's like they're using it every single day, everywhere.
404
00:24:14.550 --> 00:24:15.630
I guess what I'm curious about is like,
405
00:24:16.790 --> 00:24:21.630
how are the best businesses and brands leveraging AI and their products in a way
406
00:24:21.650 --> 00:24:24.590
that feels right to consumers?
407
00:24:25.590 --> 00:24:29.210
Whenever I come to a chatbot and it feels very AI scripted,
408
00:24:31.510 --> 00:24:33.910
it's personally not my favorite thing. So I'm curious,
409
00:24:34.510 --> 00:24:38.770
how do the best businesses build AI experiences in a way that consumers actually
410
00:24:38.830 --> 00:24:40.390
feel like this is good?
411
00:24:41.950 --> 00:24:46.030
I think the thing that is so exciting about the technology is how fast you can
412
00:24:46.110 --> 00:24:46.943
iterate.
413
00:24:47.750 --> 00:24:51.970
So the best companies that we're working with who are doing this really well
414
00:24:52.810 --> 00:24:55.490
aren't planning for three months and then shipping something.
415
00:24:56.430 --> 00:24:59.830
They're planning for three days and then shipping a hundred things over the next
416
00:24:59.890 --> 00:25:03.950
three months. And you get user feedback really quickly when you reach consumers.
417
00:25:04.030 --> 00:25:07.410
You get hundreds of data points a day, if not thousands, if not millions,
418
00:25:07.470 --> 00:25:09.630
depending on what you're shipping. Use that,
419
00:25:09.730 --> 00:25:12.350
change what you're doing and then figure out how to make it great.
420
00:25:12.490 --> 00:25:15.490
And the technology today allows you to do that more easily than ever.
421
00:25:15.930 --> 00:25:17.190
Yeah. I mean, certainly on Link,
422
00:25:17.270 --> 00:25:22.070
I think we see the same thing as Stripe's consumer-facing brand. Thanks to AI,
423
00:25:22.150 --> 00:25:27.030
we can build things in a day or week and get that feedback loop
424
00:25:27.650 --> 00:25:28.450
really going.
425
00:25:28.450 --> 00:25:31.590
And you don't necessarily have to build one thing and ship it anymore.
426
00:25:32.530 --> 00:25:34.330
If you used to have 10 user profiles,
427
00:25:34.390 --> 00:25:38.190
now you can have a thousand and build for all of them and manage it all because
428
00:25:38.210 --> 00:25:39.490
it's a little bit easier to do that.
429
00:25:40.750 --> 00:25:44.530
So if you fast forward a year, and to your point,
430
00:25:44.590 --> 00:25:46.670
we should only be thinking a couple of months ahead because it's where the
431
00:25:46.690 --> 00:25:49.990
future goes. But if you think ahead a year,
432
00:25:51.290 --> 00:25:53.250
what do the companies that have really
433
00:25:54.970 --> 00:25:57.890
sort of got exponential benefit from AI look like?
434
00:25:58.510 --> 00:26:03.430
And what separates them from just the averagely better company with AI?
435
00:26:03.610 --> 00:26:03.810
Well,
436
00:26:03.810 --> 00:26:06.810
what are the things that really differentiate those who will really succeed,
437
00:26:06.870 --> 00:26:09.870
especially the ones that are building consumer products, but just in general?
438
00:26:11.530 --> 00:26:14.190
I think it goes back to that culture question you were asking.
439
00:26:15.090 --> 00:26:17.790
Don't be too conservative about what you're letting people do.
440
00:26:18.390 --> 00:26:22.210
Celebrate if someone is really mailing a lot of jokes,
441
00:26:23.810 --> 00:26:24.790
maybe after 5:00 p.m.,
442
00:26:25.070 --> 00:26:29.270
but celebrate if someone's really mailing a lot of jokes from their account.
443
00:26:29.670 --> 00:26:34.090
How do you go and make sure that that desire to be really creative and to really
444
00:26:34.170 --> 00:26:37.970
push the limits of the technology is infused throughout the entire organization?
445
00:26:38.550 --> 00:26:40.950
And be very deliberate about those three different buckets.
446
00:26:41.010 --> 00:26:42.710
Make sure you're doing something with all of them.
447
00:26:43.370 --> 00:26:48.210
So it's less about trying to predict exactly what AI or the technology
448
00:26:48.270 --> 00:26:52.810
will be in a year and more a culture of rapid iteration, adaptation, learning,
449
00:26:53.250 --> 00:26:55.770
and being able to embrace it when it comes.
450
00:26:56.050 --> 00:26:59.630
I think so. The companies that I think are moving a little bit too slowly
451
00:27:01.350 --> 00:27:04.390
are exactly the ones that you might expect, but in some ways,
452
00:27:04.530 --> 00:27:09.390
those companies that are generally traditionally more old school enterprises,
453
00:27:09.450 --> 00:27:12.550
heavily regulated enterprises that have long procurement cycles,
454
00:27:13.030 --> 00:27:16.810
some of them are actually best positioned to go be very creative because they
455
00:27:16.850 --> 00:27:19.230
have so much information about their consumer bases.
456
00:27:19.730 --> 00:27:24.190
Where I find that companies are partnering with us very well is when they're
457
00:27:24.290 --> 00:27:26.610
asking us, "What do we think is coming?" You're right,
458
00:27:26.670 --> 00:27:29.250
we have no idea what's going to be released in the next year,
459
00:27:29.410 --> 00:27:31.910
but we do have an idea of what will come out in the next month or two,
460
00:27:32.630 --> 00:27:33.630
and they're building for that.
461
00:27:34.250 --> 00:27:34.410
Great.
462
00:27:34.410 --> 00:27:37.230
They're always trying to stay one step ahead. Stripe does this very well.
463
00:27:37.570 --> 00:27:41.970
Great. Well, in the final minute, any other thoughts, words of wisdom,
464
00:27:42.170 --> 00:27:42.770
ideas,
465
00:27:42.770 --> 00:27:46.130
things that you think would be useful for folks in this room to take away?
466
00:27:47.250 --> 00:27:50.330
I was at Stripe for eight years. I left three years ago.
467
00:27:51.290 --> 00:27:55.130
I'm very familiar with Sessions. I know this is a very, very technical audience.
468
00:27:55.190 --> 00:27:57.070
It's funny, two years ago when I was here,
469
00:27:57.150 --> 00:28:00.390
I asked people to raise their hand who was using AI at work?
470
00:28:01.210 --> 00:28:04.290
And it was about 20% of the room. Now, I guess if I asked you this now,
471
00:28:04.390 --> 00:28:07.890
please just humor me and do it. It's almost all of you at this point, right?
472
00:28:08.770 --> 00:28:12.830
And it's amazing how quickly the technology's being
473
00:28:13.450 --> 00:28:14.283
adopted.
474
00:28:14.750 --> 00:28:19.130
The one thing I would challenge each of you though is I bet your jobs day to day
475
00:28:19.270 --> 00:28:21.010
haven't fundamentally changed.
476
00:28:21.790 --> 00:28:24.310
And I think that that will not be true a year from now.
477
00:28:25.370 --> 00:28:29.070
So the challenge I have for everyone here is go figure out how to make that true
478
00:28:29.350 --> 00:28:32.190
sooner rather than later, so it's not really decided for you.
479
00:28:33.150 --> 00:28:35.730
Thank you so much. Thank you, Kate, for joining us.
480
00:28:35.930 --> 00:28:36.763
Thanks for having me.
481
00:28:37.550 --> 00:28:40.750
And my final thought, if you've been inspired,
482
00:28:40.810 --> 00:28:45.230
want to give your agents a way to pay link.com/agents to get started. Thank you.