Who it is for
Businesses with a repetitive task that scales badly — data entry, cataloguing, answering the same question, sorting the same documents.
The problem
Most AI projects fail because they start from the technology. The ones that work start from a task somebody is dreading, and remove it.
What you end up with
A job that took a week of typing takes an afternoon of reviewing.
What we build
- Catalogue and content generation — descriptions, categorisation, image sourcing
- Document and image understanding, including OCR in Arabic, Hebrew and English
- Computer vision where the camera can be the input instead of a keyboard
- Assisted replies with a human able to take over at any point
Proof
BookNest photographs a shelf and reads each book's title in three scripts. TorchCommerce writes an entire catalogue. Both are shipped, not experiments.
Common questions
Is this just ChatGPT with our logo?
No. Some of it is a language model doing what those do well. Some of it is a vision model we trained for a specific job. The right answer depends on the task.
Will it make things up?
That is the real risk, and the reason a person reviews generated content before it is published. We design for review rather than pretending the output is always right.
Can it answer from our own documents?
Not today — we do not ship retrieval over your own knowledge base, and we will not claim it until we do. What works now is assisted replies plus handover to a person.
Talk about your system
Two questions and a phone number. A real person replies — usually the same day. We are two people in Jerusalem, not a queue.
Tell us about your business