AI cataloguing reads photographs, old catalogue cards and accession ledgers and proposes descriptions, transcriptions and controlled-vocabulary terms for each object. A cataloguer accepts or corrects every one before it joins the record.
All Museums, Heritage & Auction →
Collection backlogs are rarely a lack of information. It is usually there, in a photograph, on a handwritten card from 1962, in an accession register, in a curator’s file. It has simply never been turned into a structured, searchable record. That transcription and description work is exactly where models help.
The system reads each source and proposes fields mapped to your collection management system: object name and type, materials and techniques, a physical description, inscriptions, dimensions read from cards, and terms from the Getty Art and Architecture Thesaurus or your own authority lists. Handwritten cards and ledgers are transcribed with the original image beside the text.
Everything lands in a review state in TMS, Axiell Collections, CollectiveAccess or whichever system you run, marked as machine-generated. Attribution, dating and significance are left for the curator, because those are scholarly judgements a model should not make.
The registrar picks the test set, deliberately including the difficult material: damaged cards, multilingual inscriptions, composite objects, culturally sensitive items. We run the system blind against records already catalogued by hand and report accuracy per field, so the review threshold can be set per field rather than once for everything.
Some collections need extra care. Descriptions of human remains, sacred objects or material from colonial contexts follow the institution’s ethical guidance and, where relevant, consultation with source communities. Those records are routed to a named curator and never auto-described.
photos · cards · ledgers
vision and transcription
mapped to your CMS
AAT · ULAN · TGN
logged by name
sensitive material
It can propose it. Vision and language models describe objects from photographs, transcribe legacy cards and suggest terms from controlled vocabularies such as the Getty AAT. The suggestions are useful drafts, not authoritative records, so we write them to a review state marked as machine-generated, and a cataloguer accepts or corrects each one.
Yes. We integrate with the collection management system you already run, through its API or import tools, and map suggestions to its fields and vocabularies. The record stays in your system; nothing is moved to a separate platform.
It varies sharply by field and by material. Transcribing typed cards and suggesting materials tends to go well; reading damaged handwriting or describing composite objects is harder. We measure accuracy per field on your own records before rollout and set review thresholds accordingly.
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 sample of records, including the awkward ones; we will show the suggestions field by field.
30-minute technical call · no deck