Denis Matveev

Lakku.ai

Lakku is an AI marketing platform for small businesses in Indonesia. It turns product photos into advertising images, videos and marketplace content. I joined when the team had an early technical prototype and helped develop the product through launch and later iterations.

Duration
2024–2026
Role
Product Designer
Team
9 people
Lakku’s Create AI content menu on a phone held against a sunlit tropical background

My contribution

As the sole product designer, I worked across user research, product architecture and interface design. I took part in user interviews, synthesized findings, ran first-click tests and analyzed product usage to identify friction and test design assumptions. I translated that evidence into simpler creation flows, then reviewed post-release data to assess the changes.

Share of videos using a style
6.72% → 16.19%
Share of content from Product Images
67% → 72%

How Lakku works

The core flow starts with a product photo. Users upload an image, choose a scene, adjust the settings and generate visuals for their ads or online listings. They can then review the results, save them or try another variation.

Image generation was the starting point. We later added video generation and other AI tools. As the product grew, new challenges emerged: making creation easier to reach, helping people choose the right tool, and explaining how styles and model choices affect video quality and cost.

The core creation flow: select a product photo, choose a scene and generate visuals

Help people choose the right tool

Users confused ready-made product scenes with prompt-based image creation. I replaced abstract illustrations with examples of each tool's output and clarified the descriptions. A first-click comparison gave an early positive signal, usage after release shifted towards Product Images.

Before and after: the Create menu replaces abstract icons with output examples and clearer tool descriptions

Make video styles easy to discover

Only 6.72% of video generations used a style. I added a clearer entry point and live previews inside the existing creation flow. In the subsequent measurement period, 16.19% used a style.

Video style discovery with a prominent Choose style entry point and live previews

Explain quality and cost

Small first-click tests revealed that people did not clearly connect video models with quality or cost.

I presented models as understandable quality options, with capability labels and credit costs. The update shipped, but its effect could not be measured before further development paused.

Before and after: technical model names become quality options with capabilities and credit costs