WEBVTT
1
00:00:09.280 --> 00:00:13.760
How is everybody doing? I'm Kalyani Koppisetti.
2
00:00:13.920 --> 00:00:16.380
I'm a partner solutions architect at AWS.
3
00:00:17.140 --> 00:00:20.340
I have been in the financial services industry for 25-plus years.
4
00:00:21.140 --> 00:00:22.280
In the last couple of years,
5
00:00:22.340 --> 00:00:26.440
I've been focused on payments and recently agentic commerce.
6
00:00:27.140 --> 00:00:31.600
Has anybody heard the term "agentic commerce"? Okay, good.
7
00:00:31.900 --> 00:00:36.100
I'm glad. Before we get started, so how can you introduce yourself,
8
00:00:36.160 --> 00:00:41.040
tell me what you do, and everything that we need to know about you?
9
00:00:41.320 --> 00:00:43.220
Yeah. Hello, everyone. I'm Soha.
10
00:00:43.820 --> 00:00:47.680
I'm a global leader for Amazon Agentic Payments.
11
00:00:49.180 --> 00:00:52.700
I've been nine years now working for Amazon.
12
00:00:53.580 --> 00:00:58.100
My team basically sits at the intersection of the agentic
13
00:00:58.140 --> 00:01:02.860
experiences we're building on Amazon and the infrastructure that
14
00:01:02.920 --> 00:01:04.220
makes them work.
15
00:01:05.500 --> 00:01:09.900
One of the examples we have now live is Shop Direct.
16
00:01:10.160 --> 00:01:15.040
It's a feature we recently launched in beta in the US within the
17
00:01:15.100 --> 00:01:16.600
Amazon Shopping app.
18
00:01:16.680 --> 00:01:20.520
It's an AI-powered experience on Amazon and in the app,
19
00:01:20.620 --> 00:01:25.600
which enables customers to discover and purchase products from stores across the
20
00:01:25.640 --> 00:01:26.473
web.
21
00:01:26.580 --> 00:01:29.720
Awesome. So let's actually go to the beginning.
22
00:01:30.060 --> 00:01:34.620
And at what moment it was a signal that the
23
00:01:34.720 --> 00:01:39.640
agentic commerce is not a hype and actually it is a way people would stop.
24
00:01:41.400 --> 00:01:42.560
How did this journey begin for you?
25
00:01:44.000 --> 00:01:44.260
Yeah,
26
00:01:44.260 --> 00:01:49.220
so I think the shift started to become real.
27
00:01:50.580 --> 00:01:53.480
It wasn't because of a specific product launch.
28
00:01:53.680 --> 00:01:58.040
I think it was through a change in the conversations we were having with the
29
00:01:58.100 --> 00:02:00.960
merchants. They stopped asking,
30
00:02:01.040 --> 00:02:03.180
"So what is agentic?
31
00:02:03.300 --> 00:02:08.140
What is this?" And starting asking, "So how do we get ready to it?
32
00:02:08.200 --> 00:02:10.200
This is meaningful." Signal,
33
00:02:10.300 --> 00:02:13.520
it means the question moved from curiosity to speed.
34
00:02:14.440 --> 00:02:16.080
From the payment perspective,
35
00:02:16.140 --> 00:02:20.860
the payment's tell was when merchants stopped asking about conversion rate in
36
00:02:20.920 --> 00:02:25.500
agentic flows and started asking more about authorization
37
00:02:25.560 --> 00:02:26.393
models.
38
00:02:26.480 --> 00:02:31.420
So who authorizes a transaction when a customer is not in the checkout
39
00:02:31.480 --> 00:02:35.900
screen? This is genuinely a new question,
40
00:02:35.900 --> 00:02:39.340
and one the industry does not really have a
41
00:02:39.440 --> 00:02:44.140
clean answer if I'm being totally candid. At Amazon,
42
00:02:44.320 --> 00:02:47.900
we are building agentic experiences for customers,
43
00:02:48.060 --> 00:02:50.460
and we are building the payment infrastructures.
44
00:02:50.860 --> 00:02:55.520
Awesome. At Amazon, at AWS, I should say. At AWS,
45
00:02:55.600 --> 00:02:58.460
we have every tool that you need to build an agent, run an agent,
46
00:02:58.520 --> 00:03:02.820
and scale an agent. And I've been talking to some customers and partners,
47
00:03:02.880 --> 00:03:07.140
and I'm hearing different definitions of agentic commerce. So at Amazon,
48
00:03:07.280 --> 00:03:09.500
how do you define agentic commerce?
49
00:03:09.780 --> 00:03:13.840
Yeah, that's an interesting question. It's actually something we've debated.
50
00:03:15.540 --> 00:03:16.140
For us,
51
00:03:16.140 --> 00:03:20.700
agentic commerce means an AI agent acting on customers' behalf to
52
00:03:20.760 --> 00:03:25.120
complete a shopping task. The task could be understanding their intent,
53
00:03:25.340 --> 00:03:27.180
discovering the right product,
54
00:03:27.520 --> 00:03:30.880
completing the purchase with varying degrees of autonomy,
55
00:03:31.540 --> 00:03:34.820
depending on how much the customer wants to delegate.
56
00:03:36.120 --> 00:03:39.640
The key word here is "act.". And one of the analogy that
57
00:03:41.380 --> 00:03:42.940
for me that speaks a lot
58
00:03:44.740 --> 00:03:49.620
is basically if you compare, let's say a GPS to a self-driving car,
59
00:03:50.260 --> 00:03:52.200
the GPS is informing you. However,
60
00:03:52.260 --> 00:03:56.400
the self-driving car is taking you from place A to place B.
61
00:03:57.500 --> 00:04:01.700
Agentic commerce is the self-driving car moment for shopping.
62
00:04:02.880 --> 00:04:07.820
And this really changes a lot how we think about the whole
63
00:04:07.860 --> 00:04:11.920
shopping experience. On the maturity curve,
64
00:04:12.760 --> 00:04:16.600
I believe, if you ask me this three, four months ago,
65
00:04:16.700 --> 00:04:18.640
I would tell you we're still in experimentation.
66
00:04:18.760 --> 00:04:20.720
I think now we're past experimentation,
67
00:04:20.780 --> 00:04:25.180
but that also speaks a lot about the speed at how things are moving.
But,
68
00:04:25.360 --> 00:04:27.840
however, we are not yet at mass adoption.
69
00:04:29.400 --> 00:04:30.400
As I mentioned before,
70
00:04:30.540 --> 00:04:35.460
we have at Amazon some live experiences that customers are actively using.
71
00:04:36.360 --> 00:04:37.193
Yeah.
72
00:04:37.420 --> 00:04:41.320
A concrete example, and you have it here is what I mentioned earlier,
73
00:04:41.480 --> 00:04:43.840
Shop Direct. It's currently in beta.
74
00:04:44.000 --> 00:04:48.620
It enables brands to surface their products directly within the Amazon shopping
75
00:04:48.680 --> 00:04:49.513
experience.
76
00:04:52.980 --> 00:04:56.440
When customers search for a product or brand,
77
00:04:56.580 --> 00:04:58.680
they can discover relevant items,
78
00:04:59.140 --> 00:05:03.120
and they are then redirected to the brand's own website to complete the
79
00:05:03.200 --> 00:05:04.033
purchase.
80
00:05:04.320 --> 00:05:08.260
This is very important for us because this allows merchant to reach Amazon
81
00:05:08.380 --> 00:05:12.520
customer base while maintaining full cover control over
82
00:05:12.640 --> 00:05:15.720
pricing, fulfillment, and customer service.
83
00:05:17.620 --> 00:05:21.880
It is powered by direct product feed from merchants,
84
00:05:22.180 --> 00:05:23.620
so that we ensure accurate,
85
00:05:23.820 --> 00:05:28.480
up-to-date product information and a seamless discovery to purchase journey.
86
00:05:29.600 --> 00:05:31.880
This is a clear journey, example,
87
00:05:31.960 --> 00:05:36.740
of how customer experiences are already evolving beyond
88
00:05:36.800 --> 00:05:41.680
traditional marketplaces that were more distributed yet still connected
89
00:05:42.240 --> 00:05:43.073
models.
90
00:05:43.500 --> 00:05:47.880
Awesome. I love the analogy of self-driving cars versus GPS.
91
00:05:48.640 --> 00:05:49.620
We've all been there.
92
00:05:49.920 --> 00:05:54.620
We are all comfortable with Waymos or self-driving cars that get us from one
93
00:05:54.640 --> 00:05:55.240
place to another.
94
00:05:55.240 --> 00:05:59.420
So I'm sure we're going to see the same thing in agentic commerce as well. But
95
00:06:01.120 --> 00:06:04.420
one of the things, we talked about the merchant side of product field thing.
96
00:06:04.540 --> 00:06:08.520
I think for a consumer to go from GPS to self-driving car,
97
00:06:09.020 --> 00:06:12.240
there needs to be a consumer confidence and authentication.
98
00:06:12.460 --> 00:06:17.180
I feel like agentic commerce is a place where that needs to be explored a
99
00:06:17.240 --> 00:06:20.500
little bit more. So,
100
00:06:21.580 --> 00:06:24.560
what does it take to capture what consumer wants?
101
00:06:25.440 --> 00:06:29.900
How does merchant know if the agent is acting on behalf of the
102
00:06:29.940 --> 00:06:33.220
customer, and how does the merchant fulfill that transaction?
103
00:06:33.460 --> 00:06:34.180
Yeah.
104
00:06:34.180 --> 00:06:39.040
So this is one of the challenges we have in solving basically for
105
00:06:39.140 --> 00:06:42.500
agentic commerce. From a payments perspective,
106
00:06:42.880 --> 00:06:47.800
capturing consumer needs is all about how customer authorize agents
107
00:06:48.160 --> 00:06:50.320
with specific spending parameters.
108
00:06:51.000 --> 00:06:54.800
So this is fundamentally different from traditional
109
00:06:56.160 --> 00:06:59.340
payment authorization and how we think of authorization and payment.
110
00:06:59.840 --> 00:07:03.840
So let me give you an example. If a customer tells an agent, "Oh,
111
00:07:03.840 --> 00:07:03.840
go buy me running shoes under $150." Or they tell them,
112
00:07:03.840 --> 00:07:04.273
"Can you optimize my weekly grocery spend?"
113
00:07:14.140 --> 00:07:18.380
From a authorization perspective, these are two different scopes.
114
00:07:19.420 --> 00:07:22.620
The first one is bounded and with specific
115
00:07:23.880 --> 00:07:28.420
parameters while the second one requires ongoing authority
116
00:07:28.840 --> 00:07:31.920
and judgment calls from the agent.
117
00:07:32.020 --> 00:07:35.960
So you need to be really trusting the agent to make these judgment calls on your
118
00:07:36.000 --> 00:07:40.660
behalf. We're not there, I think. I think, what is working?
119
00:07:41.240 --> 00:07:46.220
I would love for the agent to optimize spending of a lot of things that we
120
00:07:46.320 --> 00:07:47.153
go and shop.
121
00:07:47.480 --> 00:07:49.860
Yeah. And I think it's coming. At the moment though,
122
00:07:49.920 --> 00:07:53.460
what's working is more the bounded authorization models.
123
00:07:53.860 --> 00:07:55.920
Customer can set spending limits,
124
00:07:56.100 --> 00:07:59.320
they can define the product categories they are interested in,
125
00:07:59.500 --> 00:08:00.820
approval thresholds,
126
00:08:01.520 --> 00:08:05.080
similar to basically when you have a corporate card with rules,
127
00:08:05.560 --> 00:08:09.960
but you can have it very granular rules. So that's really working.
128
00:08:10.740 --> 00:08:11.573
I think today,
129
00:08:11.580 --> 00:08:15.380
one of the biggest challenges in agentic commerce is when you have
130
00:08:15.500 --> 00:08:18.980
cross-merchant agent shopping. Basically,
131
00:08:19.220 --> 00:08:22.560
if an agent wants to compare-which is the use case I think we all would love to
132
00:08:22.620 --> 00:08:26.580
have-is wants to compare prices and purchase across multiple
133
00:08:26.640 --> 00:08:28.160
retailers, today,
134
00:08:28.220 --> 00:08:32.820
the customer has to authorize the agent separately with each
135
00:08:32.920 --> 00:08:37.780
merchant. There is no common framework yet to make this seamless.
136
00:08:39.080 --> 00:08:43.840
Awesome. We talked about consumers. Let's talk about merchants. This morning,
137
00:08:44.140 --> 00:08:46.180
I had a session on agent-to-agent,
138
00:08:46.180 --> 00:08:48.560
preparing for agentic commerce or agent-to-agent transactions.
139
00:08:48.940 --> 00:08:53.080
One of the things that we said is most merchants are not even aware of consumers
140
00:08:53.160 --> 00:08:57.160
coming as agents. They block it on robots.txt,
141
00:08:57.520 --> 00:09:01.760
they don't have manifestation, they're still preparing themselves.
142
00:09:02.240 --> 00:09:05.120
But all of those are on the technical side of it,
143
00:09:05.400 --> 00:09:08.080
but how do they prepare for product?
144
00:09:08.510 --> 00:09:12.320
What do you think merchants should be doing from a product catalog perspective,
145
00:09:12.480 --> 00:09:14.680
from a data perspective, and
146
00:09:16.880 --> 00:09:18.880
what should they be thinking about?
147
00:09:18.940 --> 00:09:23.440
Yeah. And so we've been having a lot of discussion in the past few days with
148
00:09:23.600 --> 00:09:28.450
merchants. And I think one, in my point of view,
149
00:09:28.460 --> 00:09:31.720
one of the biggest opportunities in agentic commerce is personalization.
150
00:09:31.920 --> 00:09:36.320
This is what really makes it really adding value to the buyer experience,
151
00:09:37.000 --> 00:09:41.760
and how agents can act on individual customer behavior
152
00:09:42.000 --> 00:09:46.880
versus any customer behavior. The merchants,
153
00:09:47.000 --> 00:09:51.440
when we talk to them today, at least their perception is,
154
00:09:51.600 --> 00:09:56.320
that they have to face a choice. Do they optimize for agent discoverability,
155
00:09:56.480 --> 00:10:00.040
or do they continue with human-centric merchandising,
156
00:10:00.280 --> 00:10:01.720
what they have been doing for years?
157
00:10:02.920 --> 00:10:06.400
Because current systems make it hard to do both at the same time.
158
00:10:06.560 --> 00:10:09.560
So you need to make a choice.
159
00:10:11.080 --> 00:10:15.960
What is working is merchants who expose structured product data,
160
00:10:17.220 --> 00:10:21.880
pricing, inventory, specification, shippings, via APIs.
161
00:10:22.960 --> 00:10:26.680
Agents can query this data programmatically without screen
162
00:10:26.760 --> 00:10:29.560
scraping and make smarter recommendations.
163
00:10:31.360 --> 00:10:34.720
The challenge is that most systems were not built in this way.
164
00:10:35.760 --> 00:10:40.120
So this data is really scattered. It's at different places.
165
00:10:40.360 --> 00:10:45.220
It's not necessarily data that can be exposed tomorrow through an
166
00:10:45.280 --> 00:10:47.420
API. However,
167
00:10:47.740 --> 00:10:51.800
agents need real-time inventory, dynamic pricing,
168
00:10:51.940 --> 00:10:53.000
frictionless payments.
169
00:10:53.160 --> 00:10:56.980
And also they don't like CVV prompts,
170
00:10:57.880 --> 00:11:01.220
3D Secure pop-ups, all of that interrupts the flow.
171
00:11:01.980 --> 00:11:05.620
And so there's really rethinking of how is it?
172
00:11:05.800 --> 00:11:10.520
So if I'm being candid, most merchants today,
173
00:11:11.000 --> 00:11:15.120
they tell us we're not there. It's hard because, how do I optimize for agents?
174
00:11:15.180 --> 00:11:19.780
So do I remove my CVV, which I still need? So it's not easy,
175
00:11:22.400 --> 00:11:26.860
but they are starting and they are starting to build the first layer,
176
00:11:26.940 --> 00:11:31.420
which is agent-first interfaces, structured data, APIs,
177
00:11:31.860 --> 00:11:36.100
programmatic payment acceptance. This will help them in the agentic commerce.
178
00:11:37.880 --> 00:11:42.780
The tension is real though, because investing now takes effort.
179
00:11:43.160 --> 00:11:47.260
ROI is not clear. Agent-driven volume is small.
180
00:11:47.360 --> 00:11:51.200
It's growing immensely. We're seeing this also on Amazon.
181
00:11:51.600 --> 00:11:55.420
There's a lot of growth from the agent-driven volume. However,
182
00:11:55.560 --> 00:11:58.000
relative to other channels, it's still very small.
183
00:11:58.500 --> 00:12:02.420
So some merchants are taking a wait and see approach.
184
00:12:02.520 --> 00:12:05.780
Some of them are starting by structuring their data.
185
00:12:06.740 --> 00:12:07.573
Awesome.
186
00:12:09.400 --> 00:12:14.220
So we've talked about consumer needs, merchant readiness.
187
00:12:14.340 --> 00:12:17.280
Let's get into nuts and bolts of things.
188
00:12:17.920 --> 00:12:22.680
How do we ensure agent-driven transactions are actually work end to end?
189
00:12:23.700 --> 00:12:26.180
We have some recommendations that we give to our merchants,
190
00:12:26.240 --> 00:12:28.200
but I would love to hear from your perspective,
191
00:12:28.620 --> 00:12:32.300
what do you think we should be thinking from authentication to payment?
192
00:12:32.460 --> 00:12:34.560
Are there any frictions? Are there any...
193
00:12:36.120 --> 00:12:36.953
A few.
194
00:12:38.660 --> 00:12:42.160
Is it ready? Is it a matter of investing,
195
00:12:42.260 --> 00:12:46.060
or are there technologies that are still lacking, or is... What do you think?
196
00:12:47.380 --> 00:12:51.520
So if we break down the purchasing journey today,
197
00:12:51.680 --> 00:12:54.020
and what's happening, and where we see... By the way,
198
00:12:54.080 --> 00:12:57.100
it's an opportunity for us to review all of this.
199
00:12:57.680 --> 00:13:02.100
The first thing is our traditional fraud models.
200
00:13:03.180 --> 00:13:07.940
They will go and probably flag agent transactions as
201
00:13:08.020 --> 00:13:09.760
suspicious. Why?
202
00:13:09.980 --> 00:13:13.900
Because agents really behave very differently from humans,
203
00:13:14.860 --> 00:13:19.100
and these fraud models have been trained for years on human behavior.
204
00:13:19.500 --> 00:13:22.440
So for example, agents will click very, very quickly.
205
00:13:23.020 --> 00:13:26.540
They have a very high velocity. Usually the fraud models,
206
00:13:26.600 --> 00:13:29.890
if there's a lot of clicking, or clicking in a very short period of time,
207
00:13:29.940 --> 00:13:30.773
they will say,
208
00:13:36.850 --> 00:13:36.850
"Okay, this is fraudulent." They also can act at odd hours,
209
00:13:36.850 --> 00:13:40.860
3:00 am in the morning making a purchase because your sneakers now is at $100.
210
00:13:41.990 --> 00:13:44.510
And so they do not follow human patterns.
211
00:13:44.610 --> 00:13:47.590
And I think this is the first challenge is like,
212
00:13:47.710 --> 00:13:52.090
how do we retrain these fraud models?
The second one is
213
00:13:52.210 --> 00:13:56.010
authentication protocols. We just talked about it. Today,
214
00:13:56.210 --> 00:13:59.450
for me to know that Kalyani, it's you,
215
00:13:59.450 --> 00:14:01.590
usually I will send you an SMS one-time
216
00:14:02.800 --> 00:14:07.270
password, biometrics, CAPTCHA, all of that.
217
00:14:07.390 --> 00:14:12.110
And these things will block an agent.
And then the third one is
218
00:14:12.170 --> 00:14:16.090
settlement and reconciliation. I think this is also very important.
219
00:14:17.410 --> 00:14:20.950
Our current models, they assume human-readable receipts,
220
00:14:21.030 --> 00:14:23.330
you receive the PDF in your email,
221
00:14:23.750 --> 00:14:27.010
you read it very well, agent doesn't know how to read it.
222
00:14:27.070 --> 00:14:30.910
It needs to be data that they can parse. And so,
223
00:14:31.570 --> 00:14:35.970
executing an agent transaction requires
224
00:14:36.650 --> 00:14:41.250
agent authentication protocols that verify both the
225
00:14:41.350 --> 00:14:46.210
payment method and the agent authority. So instead of saying, okay,
226
00:14:46.390 --> 00:14:48.530
is your card valid? Now we need to do, "Is your card valid?
227
00:14:48.530 --> 00:14:48.530
And is the agent authorized to do this?" So add another step to the
228
00:14:48.530 --> 00:14:48.530
authorization.
229
00:14:48.530 --> 00:14:48.530
The fraud models need to distinguish between authorized merchant agents and
230
00:14:48.530 --> 00:14:48.530
compromised ones. And again,
231
00:14:48.530 --> 00:14:48.530
this is through training LLM models-which AWS knows a lot about-and then
232
00:14:48.530 --> 00:14:48.530
real-time authorization,
233
00:14:48.530 --> 00:14:48.530
because also one of the things that agent is very important,
234
00:14:48.530 --> 00:14:49.283
they cannot wait for batch approvals. Things need to happen right away.
235
00:15:21.410 --> 00:15:26.010
Yeah. I can imagine if I was seeing shoes that I want to purchase,
236
00:15:26.090 --> 00:15:29.390
I probably would do one more time research before I buy it,
237
00:15:29.470 --> 00:15:32.650
but if I told an agent to buy it, it would probably buy it immediately.
238
00:15:32.710 --> 00:15:33.390
Yeah, exactly.
239
00:15:33.390 --> 00:15:36.950
It won't even wait. And I think I'm assuming
240
00:15:39.310 --> 00:15:39.930
with agents,
241
00:15:39.930 --> 00:15:44.750
we also have to think about disputes and chargebacks and all of those things
242
00:15:44.870 --> 00:15:45.703
as well.
243
00:15:45.790 --> 00:15:49.950
Yeah. And today, there are no standards for that either. It's by the way,
244
00:15:50.030 --> 00:15:51.550
an industry-wide challenge.
245
00:15:51.610 --> 00:15:56.310
I think this is the nice thing about this is that actually it's not one
246
00:15:56.390 --> 00:16:00.870
player. I think it's all of us that need to work together to solve it.
247
00:16:01.870 --> 00:16:05.770
There's a lot of layers to the agentic stack. You have identity,
248
00:16:05.890 --> 00:16:10.630
authentication, payments, real-time settlement, dispute resolution.
249
00:16:11.710 --> 00:16:16.070
All of this needs to be developed in an integrated whole, but again,
250
00:16:16.190 --> 00:16:20.090
through everyone-industry and regulators working together.
251
00:16:20.350 --> 00:16:20.590
Yeah.
252
00:16:20.590 --> 00:16:25.590
And I'm assuming cross-border becomes another aspect that we focus on as
253
00:16:25.630 --> 00:16:26.030
well.
254
00:16:26.030 --> 00:16:26.863
Yes, exactly.
255
00:16:26.930 --> 00:16:27.763
Awesome.
256
00:16:28.590 --> 00:16:33.510
So we broke down a lot of areas where still work needs to be
257
00:16:33.730 --> 00:16:34.563
done,
258
00:16:35.110 --> 00:16:39.010
but this reinforces something that I hear from our partners and our customers as
259
00:16:39.070 --> 00:16:43.530
well, that the infrastructure decisions that we are doing right now,
260
00:16:43.830 --> 00:16:46.350
the things that we are doing in terms of how you're modernizing your
261
00:16:47.310 --> 00:16:48.870
architecture or infrastructure,
262
00:16:48.990 --> 00:16:52.630
these are going to drive what's going to come after for the next 10 years or
263
00:16:52.690 --> 00:16:55.070
next 15 years. So we are at that pivotal moment,
264
00:16:55.090 --> 00:16:59.790
and I think we kind of move towards actual action rather than just
265
00:17:00.090 --> 00:17:04.930
wait and see. Exactly. So let's close on what that means practically.
266
00:17:05.090 --> 00:17:07.770
And can you share some tips for the audience?
267
00:17:08.770 --> 00:17:11.830
Yeah. So here you go. All right.
268
00:17:13.910 --> 00:17:14.790
So in my opinion,
269
00:17:14.850 --> 00:17:19.750
there are three main things that merchants need to invest on
270
00:17:20.150 --> 00:17:24.150
if they want to really start working on agentic commerce.
271
00:17:24.950 --> 00:17:29.850
The first one, which we talked about, is focus on foundation and infrastructure.
272
00:17:30.390 --> 00:17:35.230
So do not get ahead of yourself with very sophisticated use
273
00:17:35.270 --> 00:17:38.230
cases. For merchants really,
274
00:17:38.330 --> 00:17:42.170
that means exposing structured data, via APIs.
275
00:17:42.630 --> 00:17:46.970
Start with basic product information, pricing, inventory,
276
00:17:47.550 --> 00:17:49.930
and that agents can query programmatically.
277
00:17:49.990 --> 00:17:54.750
This will help you a lot in getting the right information in front of the
278
00:17:54.810 --> 00:17:58.270
customers when the agent is trying to get this information.
279
00:17:58.330 --> 00:18:02.510
So this is very important. The second one is,
280
00:18:03.830 --> 00:18:08.250
start with simple, test-friendly, agent-friendly payment flows.
281
00:18:08.330 --> 00:18:12.410
Don't start with very complicated. Remove friction points.
282
00:18:12.520 --> 00:18:16.510
So you don't advise merchants to open up $10,000 purchases for agents yet?
283
00:18:16.630 --> 00:18:17.463
Not right away.
284
00:18:18.330 --> 00:18:18.330
Okay.
285
00:18:18.330 --> 00:18:21.170
It will come. I really believe it's coming.
286
00:18:21.270 --> 00:18:21.530
Yeah.
287
00:18:21.530 --> 00:18:25.370
It's just, we don't have, today, the infrastructure that is there. So,
288
00:18:26.610 --> 00:18:31.590
remove friction points like CAPTCHA, multifactor authentication,
289
00:18:32.450 --> 00:18:35.950
all of these. All of these block agent transaction.
290
00:18:36.350 --> 00:18:37.850
Start simple and iterate,
291
00:18:37.910 --> 00:18:41.710
see how it goes.
And then the third one is,
292
00:18:42.010 --> 00:18:44.190
begin with contained use cases.
293
00:18:45.130 --> 00:18:48.810
Think like subscription optimization. This is easy.
294
00:18:49.350 --> 00:18:52.410
Reorder automation, basket building.
295
00:18:53.070 --> 00:18:57.310
Do not try to solve right away cross-merchant agent shopping on day one.
296
00:18:57.690 --> 00:19:02.030
Start small, learn fast, expand as you gain confidence.
297
00:19:02.730 --> 00:19:04.910
Awesome. I think in the earlier session,
298
00:19:04.950 --> 00:19:09.790
we were talking about use cases across all industry verticals, including retail.
299
00:19:10.210 --> 00:19:14.970
And I think where we are seeing a lot of traction is with the SaaS
300
00:19:15.110 --> 00:19:19.550
API monetization customers, but you're absolutely right.
301
00:19:19.630 --> 00:19:20.810
Retail is also right there.
302
00:19:20.870 --> 00:19:25.770
There are some use cases that will be amazing to have an agent take care of end
303
00:19:25.770 --> 00:19:26.043
to end.
304
00:19:29.590 --> 00:19:32.960
Is there anything else you would like to add or...
305
00:19:34.190 --> 00:19:37.610
So for me, agentic commerce is here.
306
00:19:37.850 --> 00:19:42.830
It's not something like that is a future use case, et cetera. I think
307
00:19:45.350 --> 00:19:49.630
the websites or the merchants or the ecommerce that will win in the future are
308
00:19:49.670 --> 00:19:53.430
those who are really starting today.
309
00:19:54.810 --> 00:19:59.590
What I keep coming back to is that merchants and payment providers who win
310
00:20:00.230 --> 00:20:04.530
are the ones who treat trust as a product. This is super important.
311
00:20:04.590 --> 00:20:05.830
You just said that, you asked me,
312
00:20:05.830 --> 00:20:06.623
"So how does a buyer will trust like a self-driving car?"
313
00:20:10.830 --> 00:20:15.550
It's the same. We need to build this trust. It's not a feature. It's a product.
314
00:20:16.850 --> 00:20:21.070
The technology is moving very fast. The standards will catch up,
315
00:20:22.610 --> 00:20:26.350
but the trust infrastructure, the customer relationships,
316
00:20:26.890 --> 00:20:29.130
the networks that you have,
317
00:20:29.630 --> 00:20:34.050
those take years to build and cannot be replicated quickly.
318
00:20:34.730 --> 00:20:39.370
So if you have them, lean into them. If you are building them, start now.
319
00:20:39.870 --> 00:20:43.630
Awesome. So we have quite a few audience here.
320
00:20:43.730 --> 00:20:45.670
So let me ask the audience this question.
321
00:20:46.190 --> 00:20:50.870
How many of you are working for a company or know of a
322
00:20:50.950 --> 00:20:54.370
company which is ready for agentic transactions?
323
00:20:55.210 --> 00:20:57.850
You can raise your hands, you can clap,
324
00:20:59.800 --> 00:21:03.870
you can wake up if you're sleeping. There's a few.
325
00:21:07.610 --> 00:21:09.530
I see few hands.
326
00:21:09.790 --> 00:21:14.330
I'm glad there is some experimentation happening in this space.
327
00:21:15.630 --> 00:21:17.890
Have you used AI agents?
328
00:21:20.190 --> 00:21:24.110
So I did use our own, the one that we have live,
329
00:21:24.230 --> 00:21:27.850
the Amazon one that is live on Amazon.
330
00:21:29.050 --> 00:21:32.990
I think the experience is very smooth. Again, there's a bit,
331
00:21:33.130 --> 00:21:35.250
like I think like everyone of like,
332
00:21:36.330 --> 00:21:41.130
is it going to be exactly what I thought I was ordering? It was, by the way,
333
00:21:41.270 --> 00:21:45.370
so that was the good news. But it's like the trust layer, again,
334
00:21:45.430 --> 00:21:48.150
what we talked about is so important, so foundational.
335
00:21:48.250 --> 00:21:52.990
And I think these first experiences for customer will determine as well whether
336
00:21:53.070 --> 00:21:55.850
they will trust this new world that we are describing,
337
00:21:55.990 --> 00:21:58.150
or whether they will be suspicious of it.
338
00:21:58.470 --> 00:21:59.303
No, it's important.
339
00:22:01.310 --> 00:22:04.810
The merchants can provide the 1P agents on their protocol,
340
00:22:04.850 --> 00:22:08.680
like Rufus is amazing, by the way. I use it every day, but
341
00:22:10.350 --> 00:22:11.530
we are also seeing,
342
00:22:11.930 --> 00:22:16.370
we have technology where consumers can pretty much automate
343
00:22:16.530 --> 00:22:18.410
everything that they want, right?
344
00:22:18.770 --> 00:22:22.670
So you get lot of these agents that are coming through the doors,
345
00:22:23.130 --> 00:22:27.950
which you may or may not recognize because these are the agents that the
346
00:22:27.990 --> 00:22:31.810
consumers are developing using open source software that is available.
347
00:22:31.870 --> 00:22:34.450
It's pretty easy for anybody to get hands on,
348
00:22:35.050 --> 00:22:38.330
and everybody's trying out automating their own workflows.
349
00:22:38.470 --> 00:22:43.230
And I can see some use cases where I would rather have an agent do end to
350
00:22:43.350 --> 00:22:44.183
end
351
00:22:47.070 --> 00:22:51.570
than me actually spending that time or doing the context switching thing.
352
00:22:53.610 --> 00:22:58.370
So in my opinion, I don't think "agentic commerce is coming"
353
00:23:01.150 --> 00:23:06.010
is a good statement anymore. I think it's already here. What have you been,
354
00:23:07.890 --> 00:23:11.690
from the perspective of where you are, what are you seeing with companies?
355
00:23:11.910 --> 00:23:16.770
So on Amazon, we're seeing a lot of uptake. Rufus,
356
00:23:17.250 --> 00:23:19.710
there were some numbers that were shared yesterday,
357
00:23:20.870 --> 00:23:25.330
like there is a hundred and something percent year on year, more usage of Rufus.
358
00:23:25.690 --> 00:23:29.250
So the experiences we have on Amazon are really taking off.
359
00:23:30.310 --> 00:23:35.170
And again, the merchant conversation we're having is not like, should I do this?
360
00:23:35.230 --> 00:23:36.750
It's like, when should I do this?
361
00:23:36.830 --> 00:23:41.790
And I think this is really a big change from where we were
362
00:23:42.050 --> 00:23:44.910
maybe really just four or five months ago.
363
00:23:46.150 --> 00:23:48.550
Things evolved. I think also
364
00:23:51.010 --> 00:23:54.110
our mental model around agentic commerce has evolved,
365
00:23:54.170 --> 00:23:58.950
and I think that is also creating some trust in the
366
00:23:59.050 --> 00:23:59.883
industry.
367
00:23:59.990 --> 00:24:02.970
Okay. Yeah, no, that makes sense.
368
00:24:05.890 --> 00:24:08.090
I have one last question
369
00:24:10.570 --> 00:24:13.850
from an agentic commerce, we've talked about a lot of things.
370
00:24:14.490 --> 00:24:18.330
So when you ask the merchants to go slow and get the product ready,
371
00:24:20.350 --> 00:24:24.190
do you see that all the merchants are ready with most of these things,
372
00:24:24.910 --> 00:24:29.490
or is it just a matter of return on investment use case,
373
00:24:29.590 --> 00:24:34.410
or is it that they don't even have the tools and they don't even know where
374
00:24:34.570 --> 00:24:39.510
to begin? Is that more of a true statement than return on investment?
375
00:24:39.850 --> 00:24:44.370
Yeah, It's a very good question. I think it really depends.
376
00:24:45.170 --> 00:24:46.003
So
377
00:24:46.990 --> 00:24:50.430
you have some merchants who have already started.
378
00:24:51.210 --> 00:24:54.850
I was talking to a merchant already two months ago,
379
00:24:55.590 --> 00:24:59.370
and they had already started by structuring their data, API.
380
00:24:59.930 --> 00:25:04.650
They actually noticed that when customers were querying on
381
00:25:04.750 --> 00:25:08.530
ChatGPT, they were never coming up. And they looked into it,
382
00:25:08.610 --> 00:25:10.350
and it's because they were blocking the-.
383
00:25:11.240 --> 00:25:12.350
That's what I was talking about.
384
00:25:12.410 --> 00:25:17.070
-The agent. And so they opened that and they understood as well that,
385
00:25:17.690 --> 00:25:22.370
for the agent to be able to really query them in a very precise way,
386
00:25:22.470 --> 00:25:26.150
they need to expose their catalog API.
387
00:25:26.510 --> 00:25:29.910
So they exposed the catalog API. So you have these merchants,
388
00:25:31.490 --> 00:25:34.690
even those ones though, for them, this is a pilot.
389
00:25:34.930 --> 00:25:36.210
So they treat this as "We're testing, we're seeing what works,
390
00:25:36.210 --> 00:25:36.210
what doesn't work." And then on the other side of the spectrum,
391
00:25:36.210 --> 00:25:36.210
you have the merchants, and rightly so, who are wondering,
392
00:25:36.210 --> 00:25:36.943
"So what does it mean for my brand?"
393
00:25:51.690 --> 00:25:54.390
If I have agents coming and like buying,
394
00:25:54.490 --> 00:25:58.730
and the buyer might never be on my website.
395
00:25:59.250 --> 00:26:01.130
And so this is why for us,
396
00:26:01.290 --> 00:26:04.910
the Shop Direct experience is actually a good one because
397
00:26:06.750 --> 00:26:10.190
it's a mix of an agent driving the discovery,
398
00:26:10.770 --> 00:26:15.590
but then you go on the merchant website and the merchant keeps all of the
399
00:26:15.750 --> 00:26:17.670
branding, et cetera, on their website.
400
00:26:18.050 --> 00:26:20.830
But this tension is real as well for merchants.
401
00:26:20.890 --> 00:26:23.710
So what happens with everything that I've invested on?
402
00:26:24.150 --> 00:26:28.230
Yeah. We hear that from the merchants that I work with as well.
403
00:26:29.010 --> 00:26:30.810
They want to enable agentic commerce,
404
00:26:30.870 --> 00:26:33.620
but they want to also keep the consumer relationships to...
405
00:26:33.970 --> 00:26:37.530
They don't want to lose the consumer relationship,
406
00:26:37.850 --> 00:26:41.570
and it is their relationships that they want to protect as well.
407
00:26:42.510 --> 00:26:46.830
And also more importantly, they want to preserve that their brand,
408
00:26:47.470 --> 00:26:50.890
how they present their brand to their consumers. They know their customers well.
409
00:26:50.950 --> 00:26:55.610
So I can see that being the case with this.
410
00:26:57.190 --> 00:26:58.110
So I think,
411
00:26:58.210 --> 00:27:02.470
is there any other last questions from you or from the audience?
412
00:27:02.950 --> 00:27:05.930
Thank you, Kalyani. Thank you. Thank you, everyone.