Every ad platform's agentic AI bets converge on building easy buttons.

Google has Performance Max. Meta has Advantage+. TikTok has Smart+. Amazon has Performance+, Brand+, and Full Funnel Campaigns. Set an objective, hand over the catalog, launch in minutes.

Here's what I think is true about easy buttons: they are excellent, but they cannot produce alpha.

Alpha is a term often used by investors. In its simplest form, it means the edge that allows an investor to beat the market. An easy button is the market. Every advertiser in your category presses the same button, on the same model, trained on the same signals, bidding into the same auction. When inputs converge, outputs converge. Index funds are easy buttons for investors, but you don't pay a hedge fund to invest in index funds. An enterprise agency hired to get an edge for its clients cannot build its media strategy on one either.

Easy buttons thus far have been built for humans. Automated campaigns like Performance+ and Brand+ replace or augment campaigns that humans would historically need to build. But Amazon is quickly realizing that humans will no longer be the primary user of its ad properties. Agents will. And they're investing in easy buttons for agents.

The new easy button is for builders.

The most eye-opening update at this year's unBoxed was to Amazon Ads' MCP, an easy button for people building agents. Amazon's MCP server continues to expand its capabilities. It now exposes 115 tools. A new LightMCP endpoint collapses them into three: search, describe, invoke. And a new class of expert tools moves the reasoning inside Amazon's tool.

This is Performance+ for vibe coders. SMBs will build thousands of agents in Claude or ChatGPT, and every one of them will reason the same way, because the reasoning is Amazon's. Which is why Amazon's product team implied it may not be an optimal fit for enterprises. When a partner asked how to think about MCP versus the API when designing their architecture, the answer was that the two are complementary access patterns. The API remains the right choice for deterministic, repeatable, high-scale integrations. MCP exists to make Amazon Ads capabilities easier for agents and AI workflows to discover and use. Partners were told to pick the pattern that fits each use case, and explicitly not to redesign existing production integrations just to adopt MCP.

The pricing is a confession.

Existing MCP tools are free. Expert tools will be priced at cost, and Amazon's stated reason is that the price reflects the agentic reasoning happening inside the tool. Sit with that for a second. The free layer does not reason. The moment reasoning happens, a meter starts. This answer and these pricing decisions are a clear demarcation of who these products are built for: vibe coding SMBs creating thousands of low-cost, low-compute agents interacting with Amazon's ad tech. Inversely, enterprises who need deterministic, repeatable, high-scale integrations, and who want to arm the agents they hire with abundant intelligence, should rely on a pre-existing robust API infrastructure with enterprise-grade compute.

This week's announcements at unBoxed crystallize Gigi's positioning in market: a highly customizable AI agent purpose-built for enterprise agencies and brands. Whereas the Amazon Ads Agent is limited to ad hoc prompts, enterprises can assign Gigi end-to-end work and bend their operating models around the operational efficiencies. Whereas the Amazon Ads Agent can provide automated suggestions for targeting parameters, Gigi becomes encoded with an enterprise's newfound best practices to exponentially increase optimization cadences. Whereas the Amazon Ads Agent admits in its pricing that intelligence is limited, Gigi operates with abundant intelligence, running background enrichment jobs daily with always-on campaign performance monitoring. When an agent only works when you're typing, it's a convenient copilot. It's an easy button.

That clarity is exactly why we're excited to partner and go to market with Amazon from here. Amazon's agentic AI efforts are making advertising easy for the widest possible base of advertisers. Ours are raising the ceiling on what the largest and most sophisticated advertisers can do, and turning what they learn into IP they own. Those are not competing bets. They are two ends of the same market, and together they grow the pie and meet our shared customers wherever they are on that spectrum.

Amazon is building the index fund, and it is their place to do so. Most advertisers will be better off for it. But enterprises don't buy the index. The enterprises who properly invest in agentic AI and complement their human capital with token capital will create compounding IP that will enable them to leap past their peers relying on easy buttons. Gigi's place in the market is for these enterprises. We give them alpha.