AI-native workflows

Automating campaign production

Exploring how AI could cut time and cost by turning fragmented campaign inputs into draft content automatically.

Customers saw Figma handoffs as the problem. Research showed the real cost was duplicated effort: campaign information was repeatedly re-entered across systems.

Organisation: Amplience · 2025-26
Role: Principal Product Designer
Contribution: Product R&D · Product strategy · AI-assisted prototyping
Scope: Campaign production · AI orchestration · CMS automation

The brief

Cut time or cost by 50%

A Product Owner and I took on a brief originating with customer CTOs, calling for cuts to the time and cost of producing retail campaigns. The assumed problem was the Figma-to-CMS handoff. Campaign production teams had started telling us, ‘We need to do something with Figma.'

Research collage showing customer pressure to cut campaign production time by 50%, fragmented campaign tools and manual CMS data entry. — The request sounded like a Figma integration problem.Designs created in Figma were manually rebuilt as structured CMS content.

Workflow research

Figma maturity varied widely

Around 80% of customers used Figma, validating the premise. But key campaign data such as SEO and alt text still lived elsewhere.

Research synthesis showing high Figma adoption, varied Figma maturity and campaign information that lived outside Figma. — How customers used FigmaAt one end, teams eyeballed CMS output against Figma. At the other, Figma libraries mirrored CMS schemas, React props and variants 1:1.

AI-assisted prototyping

AI as the bridge

A technical POC showed AI could translate a Figma component instance into structured CMS content. Customers already shared Figma links for QA, so I designed around that behaviour and prototyped it in Figma Make. 2 engineers turned the prototype into working hackathon code.

Figma link-sharing research alongside an AI-assisted prototype that translates a selected Figma component instance into CMS content. — Research simplified the interactionThe POC initially supported both browsing or pasting a direct link. Research showed links were common, so we dropped the browse flow.

Deep discovery

The cost of data entry

Data entry was a recurring source of duplicated effort. Campaign data lived in tickets and linked files, was partially recreated in Figma, then assembled again in the CMS. Customers wanted AI to assemble the first draft — but no single system held everything needed to build a campaign.

Customer research showing demand for draft content that is 80 percent complete, alongside evidence of repetitive manual campaign assembly and data entry. — The source of truth movedVisual campaign data lived in Figma, but the rest of the campaign was distributed across tickets, assets, metadata and the CMS.

Product direction

Move the CMS upstream

Figma offered 2 opportunities: mapping campaign designs into CMS content, and defining CMS schemas from mature design libraries. The bigger opportunity was to use the campaign ticket as the orchestration point, connecting AI to briefs, designs, assets and metadata to assemble a draft automatically.

Product direction showing tactical and strategic Figma opportunities, then a larger opportunity to move CMS responsibility upstream from campaign assembly into AI-assisted draft creation and human review. — The ticket as orchestration pointThe ticket already connected the source material. AI could use those links to assemble a first draft.

As AI increases content volume, the CMS shifts from production to curation: deterministic work is automated, AI interprets source material, and people review, govern and exercise judgement at scale. My time at Amplience ended here, so the case study closes on product direction rather than a shipped outcome.