There’s nothing that gives me greater joy than meeting our customers in person. The obvious reason is that it’s good for business: relationship building is foundational to growing an enterprise software business. The non-obvious reason is that customers are often more open with their thoughts and ideas when they’re not limited to shared Slack channels or 30-minute Zoom meetings. It’s those moments on the ride to dinner or the hang after the meeting when they share kernels of insight that can be more valuable than the stated agenda of the visit.

On several of these recent visits, customers have pulled me aside to tell me how much Gigi has positively changed the way they work and how she’s unlike any product they use today, and to implore us to rapidly expand to other channels and capabilities. Naturally, I ask why. Why has Gigi been the only new AI product that your team has not only adopted but scaled across the agency? Why do you think the legacy incumbents have failed so mightily in their AI bets?

There are three common themes in their responses that I thought would be helpful to humbly share. My hope is that by writing them down, we’ll use them as a guide for the future products we build for our customers, without losing sight of what’s allowed us to win on our first media channel, the Amazon DSP.

Gigi Wins on Simplicity

There are no sub-agents exposed to our customers. It’s just Gigi. Leaders don’t have to worry about handing work off to other agents, or about whether a given problem calls for the pacing agent or the optimization agent. That’s not how human teams are formed, so why should agentic teams work this way? A media manager doesn’t invoke a budgeting skill and then a pacing skill. She opens the campaign, looks at what’s happening, and makes a call. “The cognitive mechanics underneath are invisible. They don’t need names and they definitely don’t need their own interfaces.” The same principle applies to the customizations our customers build into Gigi. They don’t need to intentionally invoke them. Gigi just adapts to their way of working.

Gigi’s Scope of Work Is Clear

A scaled ad tech leader recently threw me into his company’s marketing leadership meeting as they were bringing their agentic AI product to market, and on the spot asked me for some wisdom to impart. I told him that in a world where everything in ad tech is AI, it’s important to very clearly identify the job your agent does and align it to a job humans do today. Invariably, budgets for agentic AI will initially need to come out of labour budgets, so if you can do that, you likely have a clear fit on value and budget.

Gigi’s scope is clear: she is (at the moment) the AI media manager for the Amazon DSP. We did a painstaking amount of work to go a mile deep on the Amazon DSP so our customers could trust that Gigi will do all of the work of a media manager they’d otherwise hire. “You hire an agent the way you hire a teammate, and a teammate who does 60% of the job doesn’t shrink your hiring problem.” We will expand, and the scope of Gigi’s work will expand with it, but the clarity on what that scope is and the job Gigi is hired to do will remain.

Investing in Gigi Is an Investment in Us

One of the early design principles of Gigi’s agentic harness was allowing our customers to bring their ways of working, knowledge base, standard operating procedures, and context to Gigi. We allowed them to codify their secret sauce, containerized to their agency, and invest in new IP blended between human capital and token capital. The days of off-the-shelf legacy SaaS, deployed the same way at every agency, were gone. Not only would Gigi be customized. She would be personalized, and increasingly so with every action taken, every piece of context provided, and every team member onboarded. “It’s how you’d onboard a new hire. You don’t own the person. You own the training, the standards, and the institutional knowledge you put into them.” Our customers saw that Gigi was a force multiplier for their secret sauce, and we will continue to make investments that allow for increased personalization and newly created, customer-owned IP.

Notice that none of the three is a model capability. One teammate instead of a committee. A clear job description. Someone who learns how you work and gets better at it.

Agencies have always known what they want from a hire, and it turns out they want the same from an agent.