Clip Finder: AI search for employee video libraries
Background
The Martec enables enterprise companies to turn their people into the face of their employer brand. Employees record videos through the platform, which talent and employer branding teams use across careers sites, social and recruitment campaigns.
Customers can also upload video captured outside The Martec, such as professional shoots, agency content and long-form interviews. Over time, each account builds a library of thousands of videos, much of it used once and never again.
Context
Our customers measure employer branding content by cost per asset. New content is expensive: every video means recruiting employees, chasing responses, running approvals and editing, and professional shoots add production costs.
Reusing existing footage is far more cost-effective, because it has already been paid for and approved. In practice, reuse rarely happened. Finding the right moment meant scrubbing through every file, which took 3–4 hours for a batch of 10–12 videos before editing even started. It was often easier to collect new content. Customers were building workarounds, like pasting transcripts into ChatGPT to find highlights.
Process and actions
As VP of Product, I set the strategy and was accountable for delivery. A PM on my team led discovery and delivery day to day.
Unblocking the team: Design and feasibility were waiting on each other, so I set four requirements: natural-language search, a timestamped result at minimum, a simple path into Canva, and availability where customers already work.
Protecting launch quality: I held the team to the feature freeze process I’d introduced for enterprise support readiness, set a conservative standard for EU data residency, and coached our PM on triaging customer success feedback before go-live.
Shipping small, in a surface we own: Rather than a native Canva integration, I chose a ZIP export for the first release, to revisit once customer demand was clear. The team built Clip Finder into MAVEN, our AI assistant.
Outputs
Clip Finder launched to all customers in September 2026:
When a video is added, whether recorded in The Martec or uploaded, AI breaks it into short, self-contained moments and indexes them for search.
Users ask MAVEN in their own words, like “graduates explaining why they joined”. Using retrieval-augmented generation (RAG), MAVEN searches on meaning, so it also finds someone saying “what attracted me here was…”
MAVEN returns the best-matching moments across the whole library, each with recommended start and end timestamps, a headline, a transcript and a relevance score.
A one-click export of high-quality clips, ready for further editing.
Product launch video
Impact
In under two months, over half of customer accounts used Clip Finder.
Customers reported video editors saving up to 4 hours per day when creating content.
A global hotel group surfaced an unexpected “leadership” theme in an hour of unused professional-shoot footage.
At a global retailer, recruiters began pulling their own clips. As their employer brand lead put it, this gives “teams their own ownership.”