AI content tagging reads text, images and video and assigns subjects, people, places and events from your own taxonomy, so a media archive can be searched by meaning instead of by whatever keyword someone typed in 2004.
All Media & Publishing →
Every title has the same archive problem in a different shape. Copy migrated through three CMS generations, each with its own tag scheme. Photography sits in a DAM with captions but no subject codes. Broadcast footage is logged by programme and date, not by who appears in it. The material exists; the index does not.
We fix the index rather than the storage. Models read each item, whether article, image or clip, and propose tags drawn from a controlled vocabulary: your house taxonomy, IPTC Media Topics and IPTC Photo Metadata fields, and a people-and-places authority list. Alongside the tags, each item gets an embedding so the archive can be searched by meaning: “council housing disputes in the 1990s” finds stories that never used those words.
Tags are proposals until the archive team accepts the rules. Low-confidence suggestions, and anything naming a living person in a sensitive context, go to a review queue rather than straight into the record.
The archivists define the test. We take a sample they have already tagged by hand, run the system on it blind, and compare tag by tag. Precision and recall are reported per category, because a system that is excellent at sport and poor at politics needs a different review threshold for each.
Rights travel with every item. Search results carry the licence status recorded in your DAM, so syndicated wire photos and contributor work are visible as such and are not reused or offered for licensing by mistake. That same inventory is the starting point for content licensing.
article, photo or clip
text, image and speech
house taxonomy · IPTC
for search by meaning
to CMS and DAM fields
low confidence or sensitive
AI content tagging uses language and vision models to read an item, such as an article, photo or video, and assign descriptive labels: subjects, people, places, organisations and events. Done well, the labels come from a controlled vocabulary like your house taxonomy or IPTC Media Topics, so they are consistent and searchable, and low-confidence suggestions go to a person for review.
Yes. Speech is transcribed with timestamps, on-screen text is read, and scenes are described, so a clip can be found by what is said and shown in it. Results are written back to the fields your media asset management system already uses, so editors search where they always have.
It depends on the volume, the media mix and how much review the archive team wants. We measure throughput and review load on a representative sample before quoting a timeline, and index the most-requested years first so reporters see the benefit early.
What we built, what stage it is at, and a link to the live thing wherever there is one.
yourOS is an operating system with an AI at its heart. You talk to it like a person, out loud or typed. It makes your calls, plans your days and works on your projects with a crew of agents, on your own private server. Our own product, built from the first commit; the waitlist is open.
Website, backend and app assets for Sonora, a free AI sound-therapy app on iOS and Android published by Cognitive AI.
Brand, website, prototype, apps on five platforms, the AI backend and the algorithm-intelligence pipeline for Podswap, a social growth engine that runs without a human in the loop.
In closed beta: Roxy, a private AI assistant with permanent memory running on infrastructure the user owns. Brand, site, product, backend and conversational design.
Live at coherence.com: brand identity generated with AI agents we directed, the motion system and website for the group behind Sonora, AboutAI, WakeUp and MudShadows. Agent work remains in build.
Live at domainbooth.com for DomainBooth, a premium domain brokerage: a new animated mark, every page redesigned and rebuilt with designed mobile layouts, and a custom CMS for its sales record, featured names, press and enquiries.
Send a hand-tagged sample from your archive; we will run it blind and show you where the system agrees and where it does not.
30-minute technical call · no deck