Gigi

"Human in the loop" agentic AI isn't about reviewing every AI decision—it's about surfacing the right moments for human judgment. Safeguards are critical for this, but balance matters. Too much human review and your team drowns in approvals. Too little, and hands come off the wheel. The sweet spot: automate routine optimizations with AI, and surface only the decisions that truly need a human eye like: 📈 Bids pushed past a defined ceiling 💰 Material budget changes mid-flight ⚖️ Modifier shifts outside an approved band Adam digs deeper into this in his latest guest post for U of Digital. Check it out 👉 https://lnkd.in/gxQCyE8Q

The best AI products feel like magic. But magic rarely comes from the feature that shines in a demo. It comes from the hard problems customers never see you solve. Two months of engineering time to hit a 30-second response target. Vision zero for task failures. Running to magic means being just as intentional about what you're not building as what you are. In this week's Cherry Picked, Adam breaks down Gigi's 2026 theme “Run to Magic”, and why the hard part about running toward magic is often invisible. Read this week's full edition ➡️ https://lnkd.in/epyZhhRy Subscribe to Cherry Picked now ➡️ https://lnkd.in/eFbjzKBZ

There’s still time to register for tomorrow's session with Adam Epstein & Russ Dieringer at Stratably - 12pm ET 👉 https://lnkd.in/dQTfgaKb They'll be covering what excellent Amazon DSP management actually looks like, and what brands should be holding their agencies accountable to.

"Set it and forget it" isn't a strategy. It's what happens when execution is constrained by human bandwidth rather than performance potential. Budgets get spent. ROAS targets get met. But brands are left with no visibility into what's actually happening under the hood. Adam Epstein is joining Russ Dieringer at Stratably on Tuesday at 12pm ET to break down what excellent Amazon DSP management actually looks like—and how agentic AI is changing what's operationally possible. Register Now 👉 https://lnkd.in/dQTfgaKb

Nobody wants to talk to their data. They want intelligence. "Talk to your data" has become the default AI pitch in advertising. We think it's the wrong use case—and we almost shipped it ourselves. This week in Cherry Picked Adam wrote about why we paused a release mid-flight, what we got wrong, and how we repositioned Gigi to act on data autonomously rather than just surface it. Read this week's full edition ➡️ https://lnkd.in/eVzQYmb3. Subscribe to Cherry Picked now ➡️ https://lnkd.in/eFbjzKBZ

UI isn't where value compounds. It's in the layers underneath: the memory, the skills, the context that stays consistent no matter where your agent shows up. Gigi is going headless. That means showing up in buyer agents, in Slack, wherever our customers already work—not waiting for them to log in. In this week's Cherry Picked, Adam breaks down what that shift looks like in practice and what it means for how we build. Read this week's full edition ➡️ https://lnkd.in/ePwgTpQG Subscribe to Cherry Picked now ➡️ https://lnkd.in/eFbjzKBZ

Amazon's creative review process has no institutional memory. Every rejection lives in a black box, isolated, with no record of what came before. So Adam vibe-coded a fix—solo, no engineering support, nights and weekends. In this week's Cherry Picked, he breaks down what he built, why it couldn't wait for the roadmap, and what vibe coding a compliance tool taught him about what it means to lead an AI company right now. Read this week's full edition ➡️ https://lnkd.in/eCXshUFE Subscribe to Cherry Picked now ➡️ https://lnkd.in/eFbjzKBZ

Sometimes a client's monthly budget shifts at the last minute—and you're suddenly left scrambling to reallocate across every campaign and get it all executed before flights start. For Joe Massey at The Bluebird Group, this was a recurring reality. A process that used to eat up well over an hour of his day. Now it's just one prompt to Gigi. Every campaign adjusted proportionally. Executed in minutes.

Automation in adtech has been a double-edged sword. Done right, it drives better performance. Done wrong, it creates adverse edge cases that erode trust fast. In this week's Cherry Picked, Adam Epstein breaks down how trust in automation is built, not assumed, and what it actually takes to get there. The path from assisted AI to autonomous strategist doesn't skip the guardrails. It runs through them. Read this week's full edition ➡️ https://lnkd.in/e6DszJgn Subscribe to Cherry Picked now ➡️ https://lnkd.in/eFbjzKBZ

Jensen Huang recently said "Every company in the world needs an OpenClaw strategy. This is the new computer." Most AI waits for you to ask. OpenClaw doesn't. It runs 24/7, watches for what needs attention, acts on it, and reports back. It's the difference between a calculator and an accountant. In this week's Cherry Picked, Adam unpacks what makes OpenClaw different and how we're building those same principles into Gigi. Read this week's full edition ➡️https://lnkd.in/e42tn8hT. Subscribe to Cherry Picked now ➡️https://lnkd.in/eFbjzKBZ