PolyAI

Wren went live on 9 September 2026 across the PolyAI site, blog and social, with a launch film, a product page, in-product motion and a demo series. Leadership credited the film publicly on the day. Numbers go in here once the campaign reporting closes.

The brief

Wren is the agent behind PolyAI's agents. It watches live conversations, works out what's going wrong and proposes fixes. That's hard to explain in a sentence and harder to explain in a way that makes an enterprise buyer lean in. The brief was a launch that made the product feel like the obvious next step, not a clever trick.

What I owned

Everything visual, from sourcing the film crew to the last pixel on the product page. I scoped and briefed Ideal Insight, storyboarded the shoot around the London office, handed static UI frames to the crew, then took the footage into motion graphics, an end bumper and the loops that live inside the product. In parallel I designed the six use-case visuals, the workflow diagram and the hero prompt box for the product page.

The call

The obvious route was a fully animated explainer. I pushed for live film of real people in the real office with the product's motion language over the top, because what Wren is selling is trust: an agent that watches your agents. That lands better on a face than on a diagram. The animation carries the how. The film carries the why.

Where the agents came in

The product page needed assets that matched components already in the site's Storybook. Rather than export from Figma and rebuild them, I pointed Claude Code at the Storybook repo and had it generate the assets in place from the Figma frames, so what shipped was the component itself. Metadata went in through Sanity's MCP. I directed, reviewed and corrected. The agents did the transcription work that used to eat the last week of every launch.

Play

Outcome

The Agent Development Kit launched on 22 April 2026 with this thirty-second film at the centre of it. The product design team called it out internally on the day, and product marketing asked to reuse it for organic social, which is usually the sign a launch asset has done its job.

The brief

ADK lets developers build PolyAI agents in code. The audience is technical and allergic to marketing. The film had to show the product doing real things, at a pace that respects a developer's attention, and still feel like a brand moment.

The process, honestly

There was no script when I started, only a messaging doc. I worked with Claude to turn that into a script that hit the points in order, then generated a rough storyboard from the script so I could see the pacing before opening After Effects. The storyboard was mostly wrong in the details, which is fine. It let me make the structural decisions early and spend the craft time on the frames that mattered. The icons were rebuilt on a baseline grid, the IDE scene became a working prototype rather than a mock, and the glow end frame turned into the launch's signature.

The call

Version one had the product in a browser window. Version two removed the window and let the interface float in the brand's dark field, because the thing being sold wasn't a website, it was the feeling of building. That came from watching the first draft with the product team and noticing where their eyes went.

Dialog RSN-1 is PolyAI's new reasoning model. It launches on 30 September 2026 with its own page, a film and a paid campaign.

The brief

A model isn't a product with a screen. It has a name, some numbers and a claim about how it behaves on a call. The job was to give it an identity that sits under the PolyAI brand, speaks to developers and enterprise buyers in the dark, technical register they respond to, and stretches from a 4K launch frame to a 300 pixel ad without falling apart.

What I did

Set the identity: a dark palette, a generative particle system for all of the imagery, and the type and layout for the launch page. Directed the film shoot with Ideal Insight, a table, a laptop and the people who built the model, so the launch has real people showing real behaviour rather than a voiceover over stock. Built the model frames at 4K and the ad sets that went to demand generation.

The imagery

The particle field isn't a still I made once. It's a WebGL2 shader that renders any image or video as glyph particles, so the launch page, the film inserts and the ads all come from one system at whatever size they need. I wanted this model to look different from the rest of the range: the same system, tuned to read as circular where the others read as fields. How the shader works is written up in the lab.

Where it went

Launch page, film, model frames, social and paid. The paid sets are in Demand generation.

Most of what a brand does in public is ads, and they're the first thing to drift when a team is moving fast. So the work here is systems rather than one-offs: a master frame with one headline, one proof point and a piece of the product, then the sizes that fall out of it. Highlight one thing, make it legible, make it adaptable to motion.

Dialog RSN-1

Two audiences, two sets. Buyers get the outcome and a photograph: faster calls, fewer escalations. Developers get the benchmark and the waveform. Same palette, same type, so they read as one launch.

The second wave leads with a single number and an annotated transcript, so the ad shows what the model does on a call rather than telling you.

Wren webinar

A featured frame with the four speakers, a card for each of them to post themselves, and email banners that carry the two lines from the launch: the agent behind your agents, and your agent, better every day.

A product sizzle for PolyAI's simulation testing launch. Problem, product, resolution in thirty-five seconds, for two people at once: the buyer who cares about risk and the builder who's actually shipping the agent. Built from an animatic in a fortnight, exported in landscape for launch and reworked in portrait for demand generation.