$100k+ MRR
From an AI concept to a launched consumer product.
I joined with no stable end-to-end product and turned an ambitious conversational video concept into a public iOS subscription launch.
By Andrei VolkRole at the company: Head of Product · Elva · 2026
ScopeDefined the product metrics, product strategy, experience design and feature roadmap, using customer discovery, UX interviews and feedback from paying users to set priorities.
ResultA live AI video editor used by paying customers.
- FROM CONCEPT TO FIRST PAYING CUSTOMERS
- 3 months
- EXPORTS / WAU
- 2.5
- EXPORT-BASED D7 RETENTION
- ~25%
About
A compelling idea, but no stable end-to-end product.
Users were meant to turn raw photos and videos into a finished story through conversation. The team had to solve the customer journey, agent orchestration, media processing, generation quality, monetisation and launch readiness at the same time.
01 / 07 · Start a conversation
What it took to launch
- User journey
- AI agent coordination
- Photo and video processing
- Output quality
- Subscriptions and payments
- Release preparation
Product architecture
Define the complete journey before expanding the feature set.
I mapped the flow from onboarding and voice or text prompting through media upload, generation, preview, refinement and export. Home became the navigation portal; Agent Space became the core creation environment — voice-first, but not voice-only.
The path to a finished story
- 01OnboardingEnter the product
- 02MediaUpload source material
- 03PromptVoice or text
- 04GenerationCreate the first result
- 05PreviewReview the output
- 06Refinement / editingAdjust the result
- 07ExportReach the value moment
- 08RetentionLearn from behaviour
Preview and editing repeat until the story is ready to export.
View original artifact
MVP trade-offs
Protect the core value moment: a complete, reliable export.
Camera capture, video zoom and crop, and text-to-speech voiceover were removed from the launch MVP. The product was split into explicit creation, clarification, variation, refinement and failure states so mobile, backend, ML and agent workflows could operate against the same model.
- Camera capture
- Video zoom and crop
- Text-to-speech voiceover
After launch
Use behaviour and support signals to shape what came next.
We added photo and Live Photo support, built analytics and experimentation foundations, and prioritised contextual manual editing tools when data showed that AI output alone was not consistently sufficient.
- 01Support patterns
- 02Product priorities
- 03Photo & Live Photo support
From a first result to a finished story.
Conversation, direct manipulation and export — complementary ways to finish the job.
Refine the result in context.
Reorder clips on the timeline.
Take the finished video with you.
Onboarding experiment
- 01
- 02
- 03
- 04
- 05
