Hundreds of millions in enterprise value has been created by companies building products on Amazon's ad tech. Companies like Pacvue and Skai have emerged as clear leaders with enterprise agencies and brands. Upstarts like Xnurta and Intentwise have built valuable solutions for 3P sellers and small, independent agencies - and many of our team at Gigi were fortunate to be early employees of Perpetua, which was acquired by Ascential for a 9 figure outcome in 2021.

All of these companies have built scaled solutions for Amazon Sponsored Ads, Amazon Commerce, and Amazon Marketing Cloud.

Interestingly, when we talk to customers of these companies, none of these solutions are actually being used for the Amazon DSP. Every customer cites finding the notoriously clunky and laborious native UI of the Amazon DSP more effective than the solutions offered by these companies. There's a reason for this.

As a technology partner to walled gardens like Amazon, there is a constant race to "keep up with the APIs." Every time Amazon releases new functionality, a technology partner needs to urgently assess whether this new functionality is valuable to its customers and prioritize integrating with the APIs and building workflows in its app to support it. If they choose not to do so, they are tacitly making a choice that their customers will continue to see value in their product without this offering, and begrudgingly encourage them to complete this workflow outside of their product.

The Amazon DSP, in particular, is incredibly feature-rich, and that's exactly what makes it so hard to build for. The API integration itself takes days. Everything after takes months: designing the user journey, accounting for every edge case, building the front-end workflows that make it usable. And you have to do that from scratch for every single feature Amazon ships. The surface area keeps growing, the UX gets heavier, and at some point you're spending more energy mirroring what Amazon built than improving on it. For legacy SaaS companies, that debt is nearly impossible to repay. So they make rational tradeoffs, supporting common workflows for their customer segment, typically basic bottom-of-funnel strategies, and brands and agencies end up back in the Amazon DSP console’s native UI anyway. This is not meant to be an indictment of any of these companies. Rather, it is an assessment of the challenges of building an Amazon DSP product for legacy SaaS.

With agentic AI, the way we build software and interact with it is fundamentally different. The distinction comes down to this: legacy SaaS is prescriptive and rigid, agentic AI is declarative. For a company like Gigi to integrate with new Amazon functionality, we don't need to undergo the hardest part of these integrations — building prescriptive UI workflows — because there are no UI workflows to build. Instead, we expose the API endpoint as a "tool" that Gigi can call. In plain terms: Gigi reads the API's schema, understands what it does, and can invoke it directly without any front-end wrapper needing to be built first. That cycle drops from months to days. The surface area stops being a liability and becomes an asset.

Take, for example, analyzing audience reporting and making optimizations from those reports — functionality that no other partner aside from Gigi provides. In legacy SaaS, you would need the ability to ingest these reports, a filtering UI to analyze them, and a workflow to make audience changes to existing campaigns. Building all of this could take 2 to 3 months, and it's unclear how to do so in a manner that's better than the DSP. With Gigi, we simply expose audience reporting and audience editing within line items as tools that Gigi can call. Gigi then analyzes the data, provides suggestions, and takes action on audiences — all within a single prompt or "task." In fact, one of Gigi's most popular scheduled tasks is asking her to identify underperforming audiences and suggest the most performant audiences (based on purchase rate) to swap into a campaign in a single click. Gigi currently exposes hundreds of API endpoints as callable tools, and each new one we add goes live in days. This sort of functionality is only possible when you move from prescriptive and rigid software to declarative agentic software.

When talking to customers, many ask us: "Why hasn't anyone else built a great Amazon DSP product?" This is the reason: it is impossible to do so with legacy SaaS. The only way to provide demonstrable value to customers and "keep up with the APIs" to meet the feature richness of the Amazon DSP is with agentic AI.

Note: This first appeared in our Cherry Picked newsletter. You can subscribe here 🍒