The Agentic Wars Have Begun: Why AI Agents Will Disrupt Retail Like E-Commerce In The 1990s

Foreigning us once again for five insightful minutes is one of our favorite recurring guests, David Dorff.

David is the head of retail industry solutions at AWS and he always regales us with something fascinating about AI.

David, welcome back to omnitalk.

And let's get started with this.

When you were here in April, you talked about AI agents and how you foresaw them changing retail over the next five years.

What has changed, if anything, in the last six months?

Yeah, I'll tell you what, one thing that's changed is it's gone a lot faster than I imagined.

There's been so many crazy things happening.

Case in point, Amazon came out with the buy for me.

Google's got a shopping agent and OpenAI has got a shopping agent.

Now it seems like they're all over the place.

We have.

The agentic wars have begun I think.

So it's a race right now and I think this holiday season we're going to see how this stuff really works in action.

I think a lot of search is going to emanate from these gen AI platforms and I think Adobe was showing like 1000% year over year growth with gen AI leads coming in for retail purchases.

So I think it's still a small slice of the pie, but I think it's going to grow quite a bit.

And then I think there's a lot of standards that are emerging that are very cool.

So.

So I saw Fortner has a trusted agentic commerce Protocol and Visa, MasterCard and PayPal have got come up with their own standards around payment for agents.

So there's a lot of change out there.

Then David, what happens for if I'm a retailer right now, how do I prepare for a holiday season where people are going to be asking these agents, you know, what do I get from my mother in law who hates the color green and everything I've given her in the history of Christmas?

Yeah, yeah.

So I think success increasingly depends on optimizing your website for both humans and agents.

I know that sounds weird.

We've all been doing search engine optimization and now we're starting to switch to generative engine optimization geo.

So really there's three things that I think retailers need to think about.

Number one is what is your strategy for inbound agents?

Some retailers may choose to block them.

Understood.

Some may want to accept them but make it easier for agents to actually get the data that they need to execute a product lookup and an eventual checkout.

The second thing they need to think about is what's your outbound Agent strategy.

So do you want to extend your assortment, kind of like you would do with a marketplace, but by creating agents that go out to other platforms and buy on behalf of your customers?

So that may work in some cases, in some it may not.

And then the third thing they need to look at is what do you want to do with on site?

Do you want to have an agent on your site like Amazon Rufus, for example, that can help make those recommendations so that you're capturing the customers and not, you know, just letting them go off to some LLM platform?

So I think if they look at inbound, outbound and on site, those are two really big things for agents in commerce today.

Oh my God, David, I have so many questions, but we only have five minutes, so.

All right, I'm going to try to pull my.

To try to boil down everything in my head into this.

So.

So if, if that's the case, what technical infrastructure do retailers need to be successful given whatever approach they want to take?

Yeah, so right now the way a lot of these agents execute checkout is through the browser.

And so they're sort of emulating a person clicking and all that kind of stuff.

They can use something like Amazon Nova act, which is a browser based LLM, but that can be unreliable and tricky.

Lots of JavaScript can mess it up and things like that.

What retailers need to think about, can I avail my product catalog to these LLMs?

So you may want to create some APIs to make it easier.

We really need to digitize that interface.

That way agents can find the products that they're looking for and execute those checkouts a little bit easier.

And the other thing we need to keep an eye on, like if you're having an on site LLM for conversational search, we're starting to see domain specific LLMs come to market.

So we've seen some for healthcare and finance and I think retail is not far behind.

We'll have large language models that are trained specifically on product catalogs to make really clean recommendations.

So there's some cool stuff coming around the corner.

Wow.

I don't even know if I can ask you this without having more questions, like Chris said.

But if you had a crystal ball, what do you think it would reveal about the future of agentic commerce?

David?

So just like the Chief Digital Officer kind of ushered in E commerce for retailers, I think we're going to see more chief AI Officers.

Officers popping up.

You probably saw Lululemon.

Yeah, yeah, they just named wow.

Okay, Officers.

And I think that's important.

You really need somebody that's tasked with getting AI in the whole business.

It's not just about commerce.

There's a lot of really cool stuff you can do on the back end, too.

So automating decisions around forecasting, pricing allocations, marketing.

There's a lot of really good things that agents can help with on the back end.

So I think we're going to see a lot more of that as well.

And that's all going to lead to better customer experiences in the long run.

And if anybody's interested, Granger, Cisco and Fanatics are going to be talking about their use of agents at Re Invent, which is our user conference coming up in November in Las Vegas.

So that said, I really think agent commerce is going to disrupt the retail industry in much the same way E commerce did back in the 1990s.

So this is a retailer's opportunity to sort of stay ahead and understand how this is going to work and make it a competitive advantage or risk falling behind.

Wow.

Great stuff, David.

Oh, my gosh, great stuff.

Thank you.