Runway's new Agent 2.0 builds entire ad campaigns from a prompt

Anyone who runs digital ads knows that generating a flashy clip was never the real bottleneck. The pain is figuring out what to say, slicing that idea into half a dozen aspect ratios, and constantly bouncing between analytics dashboards and creative tools.

Runway just launched Agent 2.0, an AI workspace designed to turn barebones text prompts into full-fledged marketing campaigns and multi-platform assets.

From brief to multi-model execution

Agent 2.0 isn't just another standalone video model. Instead, it sits one layer above, acting as an interactive creative partner inside a single conversation.

You can feed it an initial concept, a product launch brief, a target audience, or even an underperforming ad. Before generating a single frame, the agent asks probing questions to pin down your angles, hooks, and core messaging.

Once the strategy is clear, it handles production across video ads, social posts, image ads, and commercials. Crucially, it manages the tedious adaptation work: generating 9:16 vertical assets for Reels and Stories, 16:9 for YouTube, and 1:1 for social feeds, while localizing copy and visual cues for different regional markets.

Under the hood, Agent 2.0 can automatically orchestrate various models depending on the task, tapping into Runway's Gen-4.5 as well as models like Aleph 2.0, Seedance 2.0, Nano Banana Pro, or Kling 3.0 Pro.

Ingesting ad data without the guesswork

Runway is also taking aim at post-campaign analysis. Agent 2.0 can ingest creative and performance data from Meta, YouTube, TikTok, and Google campaigns to determine what is working and what to test next.

Instead of making marketers piece together correlations across spreadsheets, the agent can spot practical patterns—like noting that variants with a direct demo in the first two seconds converted best—and immediately draft the next batch of creatives.

Still, marketers need to keep their guard up. Sensitive user data needs to be stripped before uploading, and correlation isn't causation: a high click-through rate means nothing if the landing page flops. Plus, casually asking the agent to "try another variation" can quickly turn into an expensive token burn without improving the underlying marketing thesis.

Why it matters

We think this is one of Runway's smartest pivots to date.

Instead of staying trapped in the brutal race of pure video foundation models, Runway is pushing upmarket into end-to-end creative workflows where marketing budgets actually live. Orchestrating multiple third-party and proprietary models inside a single brief-to-asset pipeline turns Runway from a video generation tool into a specialized operating system for growth teams.

For lean startups and solo operators, it substantially lowers the barrier to running multi-variant, localized tests across channels without a huge production agency.

That said, the dream of a fully autonomous marketing machine that auto-deploys and optimizes ads on live ad networks is still on Runway's future roadmap. For now, Agent 2.0 is an assistive co-pilot, and human marketers still need to verify the strategy, manage brand safety, and keep the token meter from running wild.