SAYNIS AI Suite
A workspace that reads institutional documents, automates the routine steps around them, and runs AI assistants — including in Somali, Amharic and Arabic.
Institutions in the region run on documents — letters, forms, reports, registers — most of them scanned, many of them in Somali, Amharic or Arabic. General-purpose AI tools handle these poorly, and sending institutional documents to a third party is often not permitted.
AI Suite is being built as a workspace where reading, sorting and assistants all run against an organisation's own documents. One deployment option keeps those documents inside the organisation.
It is not a chatbot wrapper, and it is not a general model. It is the workflow layer around whichever model is appropriate for the task.
What SAYNIS AI Suite is being built to do.
Document ingestion
Designed to take scanned material from existing scanners and file shares.
Field extraction
Will pull structured fields from forms and letters into records.
Multilingual handling
Somali, Amharic, Arabic and English, tested against real documents.
Scoped assistants
Answer from one department's own documents, not the open web.
Routing and triage
Will classify incoming correspondence and route it to an owner.
Human escalation
Low-confidence results go to a person rather than being guessed.
Decision log
Every automated action recorded and reviewable after the fact.
Four steps, end to end.
- W1Connect
Connect a document source
A scanner output folder, a mailbox or an existing file share.
- W2Understand
Extract and classify
Fields are read from the document and the type is identified.
- W3Act
Route or answer
The document reaches its owner, or an assistant answers from it.
- W4Escalate
Escalate what it is unsure about
Anything below the confidence threshold goes to a person.
- Cloud, or on-premise where documents may not leave the building.
- Organisations with continuous document flow, not occasional files.
- Mixed-language material, including scanned handwriting.
- Existing scanners, mailboxes and file shares — no new hardware.
- Custom models where they earn their cost; established providers where they do not.
- PostgreSQL with object storage for source documents.
- Docker, with cloud or on-premise installation.
- Held-back evaluation sets, confidence thresholds, monitoring from day one.
What people ask about SAYNIS AI Suite.
We are not publishing a launch date, because we do not have one we would stand behind. Design partners get access first, and we will say plainly where the product is when you ask.
Wherever you require. An on-premise deployment keeps documents inside your infrastructure. It costs more to run, and for some institutions it is the only acceptable answer — so we will price both.
English is strongest. Somali, Amharic and Arabic are uneven, particularly on scanned material. We test against your actual documents before making any claim about accuracy.
It escalates. Every automated decision has a confidence threshold and a human route, and every action is logged so a wrong result can be found and corrected.
Pricing is not set. Design partners are not charged during the programme, and there is no obligation to buy afterwards.
Join the early-access list
Design partners get early access and direct influence over what we build. We can only take a few at a time, so we will tell you quickly either way.