Rembrand AI — In-Content AdTech Platform
Mykola Popov
Role: Sr. Product Designer · Led editor and platform UI
Timeline: Jul 2022 – Jan 2024 (~18 months) · Industry: AdTech · Gen-AI video
Scope: Creator platform, brand dashboard, video editor, design system
The problem
Rembrand places brands directly inside video using spatially-aware AI — a can on the desk, a logo on the wall — instead of a pre-roll nobody waits through. The hard part wasn't the placement. It was building a product three different audiences could trust.
Creators feared a placement would cheapen their video, so they hesitated to monetize at all.
Brands feared context they couldn't see in advance — the placement looking wrong, or landing in an unsafe scene.
Advertisers needed to run and target campaigns at volume without hand-reviewing every frame.
Research
Interviews and flow-mapping across all three user types, plus walkthroughs of how each decided to approve or reject a placement.
Realism mattered less than expected. The decision hinged on preview and control — people rejected anything they couldn't see and adjust before it went live, not things that merely looked fake. Confidence, not realism, was the conversion lever.
The bet
In-content advertising only works if every party trusts what the viewer will see. Preview, control and approval are the product. The AI placement is the engine, not the experience.
AI proposes → everyone previews and vetoes → it goes live.
Design principles
Never auto-publish. Every brand-in-scene is previewed and approved by the people who own the risk.
One platform, three lenses. The same placement reframed for the creator (authenticity), the brand (safety), the advertiser (scale).
Make the AI legible. Show where it placed a brand and why, so people can trust it or override it.
Three strategic decisions
Editor: automation with a veto. Full automation is the pitch, but it terrified the people who own the risk. Built the editor around review-and-approve: AI proposes, humans inspect in context and accept, adjust or reject.
One platform, not three products. Creators, brands and advertisers have opposite priorities. One platform with role-specific lenses on a shared model, unified by a single design system across all three surfaces.
Brand safety: expose control, don't hide it. Surfaced context and approval controls where the risk lived, with sensible defaults — clarity over a falsely simple feel.
Results
38% → 72% placement acceptance once everyone could preview and veto
~3 days → under 10 min approval time, end to end
+44% creator opt-in after hesitation, the top blocker, dropped
3 surfaces, 1 design system — faster, more consistent releases
North-star: placements that shipped because all three parties trusted them — not placements the AI generated.
What I'd carry forward
Make the model legible from day one. The AI's "why it placed here" explanation was central to trust, not a polish-phase addition.
Like this project
Posted Aug 8, 2026
Led product design for an AI platform that places brands inside video. Preview, control and approval became the product — placement acceptance 38% → 72%.
