"But don't you think all of these legacy modes of agent interaction will be obsolete soon, as you'll just rely on a platform's MCP capabilities?"

An agency executive posed that question to me last week as I was explaining how Gigi now collaborates with media platforms. Gigi has browser use, so she can be trained to push buttons like a human where no API exists. In these instances, Gigi can also be trained to execute bulk CSV and structured data file uploads instead. Within weeks of the rollout, customers are already using it for YoY reporting, audit logging, and audience creation.

The MCP capabilities may get there. But that was never the question.

I wrote about this in February, when Amazon Ads launched its MCP server. The short version: Amazon has built API-first for over a decade, which is why an agent like Gigi can call most endpoints a human can click. The MCP server launched with a limited set of supported workflows, and I said then that parity with the API would take several years. Seven months later, there have been improvements, but the gap remains. What's changed is the shape of the question. It used to be a protocol question. Now it's a resourcing question: should an agency spend its own product budget working around the gap, or wait for Amazon to close it? "Wait for the platform" has quietly become the default answer to the build or partner conundrum.

Here's the point I'd make to any tech company or agency thinking about this: if you wait for MCP, you miss the value you could be driving right now. We chose not to wait.

We simply couldn't afford to wait.

We're an early stage startup. Two to three years for MCP to reach API parity isn't a roadmap we can plan around. Our customers needed value right away, and if Gigi couldn't deliver it, there would be no Gigi around by the time the protocol matured. Waiting wasn't a strategy. It was existential.

So we built browser use and bulk uploads because our customers asked Gigi for them, and like any teammate, it would have been insufficient for her to say she couldn't do the work because of an API limitation. Amazon is quickly evolving its ad tech infrastructure to be agent-first, but what's exposed through MCP still pales in comparison to what's available through the API. Our customers can't wait that long, and neither could we.

Our customers have been rewarded for it.

I said a version of this to the exec. His response: "That's exactly why we hired Gigi."

He'd come to the same conclusion months earlier. If he waited for Amazon, he could be waiting for years. So he hired Gigi, and his team got the immediate returns: exponential time savings, improved client performance, and an elevated level of service across all clients.

But the more interesting benefit is the second-order one. Because he didn't spend his product resources rebuilding a bridge Amazon will eventually ship, he can spend them on agentic bets that only his agency can make. His team is closer to the client than any platform will ever be. They have the calls, the communication, the strategic collateral, the reasons behind every decision a client made this year. None of that lives in an API, and no platform will ever expose it through MCP. We're now exploring product collaborations to bring that context into Gigi, enriching her beyond the media and commerce signals she already has access to. The agency owns the context. Gigi does the work. That is the compounding IP that an agency controls.

Waiting for the protocol isn't a neutral choice. It costs you the value on the table today, and it costs you the product cycles you could have spent on the layer no platform will build for you. So yes, these modes of interaction will probably be obsolete one day. By then, the agencies that didn't wait will have two years of compounding IP that no protocol will ever contain.