Across dozens of Gigi onboardings this year, some pilots have turned into smashing successes. Others have stalled mid-pilot. What determines which is which?

Historically our product was the bottleneck for success. When we first onboarded customers in summer 2025, the scope of Gigi's capabilities were limited and we weren't able to deliver on the promise of what Gigi could be. That's changed over the past couple months, and we've now found an agency's team structure can be a bottleneck in success.

Gigi cuts the operational lift of managing the Amazon DSP by at least 50% and immediately improves advertising performance. That much is established across the implementations we've run. The question we now get from agency partners, including ones who have seen Gigi running for months, is structural: "I get it. I see the value. How should we architect our teams to maximize it, keep operations efficient, and get the best results for our clients?"

This post is our answer to that question.

The agencies we work with fall into two buckets: programmatic media agencies and Amazon / retail media agencies. The team structures look different, but the principles for getting Gigi to work are the same.

The Amazon DSP demands expertise

The Amazon DSP is not a channel you operate part-time. It has too many knobs and levers, too much surface area, too many edge cases. When an agency allocates a small portion of a media manager's time to this channel, the media manager never builds the expertise the channel demands.

This matters for Gigi, because Gigi is downstream of that same expertise.

Bidirectional expertise is the unlock. Know the channel deeply enough to interpret client objectives. Know the agent deeply enough to wield it against those objectives. The media manager who allocates 10% of their time to Amazon DSP develops neither. That's the single biggest predictor of whether Gigi works inside an agency.

The generalist athlete model breaks

Across our retail media agency partners, a common operating model is the generalist. One media manager runs every channel and every retail capability for a book of clients. That model worked in legacy tech because the work was forms and dashboards. The job was data entry with judgment attached. A generalist could grind through it.

The model breaks in agentic AI. The job changes from pushing buttons to directing an agent that pushes the buttons. Directing well requires deep knowledge of what the agent is doing and why. The generalist who knows a little about a channel is limited in their ability to direct an agent with conviction. Yes, Gigi can make the generalist incrementally more efficient in their work. But agents bend an enterprise's operating model with exponential productivity gains. We believe we can enable 10x media managers, we've seen it, and the path to achieving this is through centralized expertise.

What works: an Amazon DSP expert who wields Gigi

The agencies getting the most out of Gigi have an Amazon DSP expert (or several) who wields Gigi to the fullest extent across a collection of clients, or in some cases a single very large client. Because this person lives in the DSP, they know exactly how they want Gigi to operate. Gigi is replacing the buttons and tasks they would otherwise do every day. They have an instinctive sense for what to delegate, what to review, and what to override.

For programmatic media agencies, we've seen collections of teams own individual tasks within the media channel. Since Gigi can now execute many of these tasks autonomously, the staffing requirements for scaled execution cease to exist. Similar to retail media agencies, a single media channel expert with Gigi as a team member is the optimal model for exponential productivity.

The traditional bottleneck in this model is intra-agency context. One expert serving many clients means everyone else on the team routes through that person for status, reports, and updates. That's where this structure historically breaks.

Gigi resolves the bottleneck. Anyone on the team, in Slack or in the Gigi app, can ask Gigi for context and basic reports on demand. The expert no longer fields requests for information. They focus entirely on strategic execution and orchestration of the agent.

The 10x media manager: a job description

This is a suggested optimal way to work with Gigi, not a prescription. We've seen Gigi work across a range of team structures. But based on patterns across our customer base, we’ve seen this model drive meaningful value.

We thought it would be helpful to draft a mock JD for this role:

What you own

You are the strongest operator of Gigi at your agency. You take strategic direction from leadership and clients, then wield Gigi to execute against it.

• Campaign and line item setup through Gigi: ad tag configuration, 3P creative setup, audience targeting, line item construction.

• Pacing and performance management through Gigi: pacing data review by line item, base and max bid adjustments, frequency tweaks, budget shifts across line items.

• Capabilities Gigi unlocks that weren't happening before: line item minimum enforcement, inventory QA at scale, programmatic bid adjustments across the full campaign.

• QA and intervention: review Gigi's recommendations and execution, intervene where judgment is required, escalate the few decisions that warrant a human in the loop.

• Strategic planning: use Gigi to translate client objectives into structured data-driven media plans, recommendations, and deal event strategies.

• Reporting through Gigi: generate custom scheduled reports across harmonized ADSP and AMC data that surfaces full-funnel performance with strategic insights and learnings.

• Subject-matter depth: build the deepest ADSP expertise at your agency. Every lever, every nuance, every edge case Gigi might encounter.

Profile

• 3+ years hands-on Amazon DSP. You've actually pushed the buttons.

• Strong judgment over task execution. The button-pushing is automated; the thinking is not.

• Comfortable with AI-native tooling. You prompt well, you review outputs critically, you know when to override.

• Performance-marketing instincts. You understand how pacing, bidding, frequency, and budget interact, not just how to adjust each in isolation.

• Strategic translator. You can take a directive like "grow reach while managing to optimal frequency to conversion" and turn it into a plan Gigi executes against.

The foundation is expertise

AI agents produce exponential productivity gains. The clearest evidence is in coding, where the best engineers are now 10-100x more productive than they were two years ago. The foundation is expertise and craft. We believe a similar foundation is required to empower the 10x media manager.