Engineers meet AI Intelligence · Connection · Possibilities
A pharmacist re-typing a prescription. A finance team re-keying invoices for a week. A family relaying the shopping through a group chat. We build software that takes those jobs over — and hands the decision back to a person the moment it is not certain.
01 — Deterministic where it matters
Rules, schemas and reference data decide the things that have a right answer. The model is used where judgement is genuinely required, and nowhere else.
02 — A person at every uncertain edge
When the system is not sure, it does not guess and it does not quietly proceed. It stops and asks a person, showing exactly what it was looking at — so the check takes seconds, not a phone call.
03 — Every decision reconstructable
For every field we keep what the system saw, what it concluded, how sure it was and who signed it off. Months later, any record can be reconstructed exactly as it was decided.
What we build
Three business platforms, already used by a few early customers. One drone survey product, in a private pilot. Two apps for everyday life, still being built. Every card says where it really is — we would rather tell you a product is unfinished than let you find out.
For business · early customers
Healthcare & Pharmacy
Reads a prescription the moment it arrives, checks it against live clinical and stock data, and writes the confirmed record straight into dispensing.
See how it runsFinance & Business
Reads financial documents as they arrive, codes every line against one consistent rule set, and turns month-end into a review of exceptions.
See how it runsLogistics & Operations
Compares carriers automatically, books once against the right API, and watches tracking streams so exceptions surface before a customer calls.
See how it runsOn site, and in everyday life
Connection · consumer app
Find the people near you who are into the same thing you are, and somewhere real to meet them — a library, a coffee shop, a pub.
In active development. Not open to the public yet, and we are not collecting sign-ups until it is worth your time.
Family coordination · consumer app
One shared place for a family spread across the day — what the house needs, who is where, who is picking it up, and what it is costing.
In active development, being tested inside our own households before anyone else’s.
Construction intelligence · drone hardware and software
We build the drone, fly the site, and turn what it sees into an honest answer to one question: is this build where the plan says it should be?
Running as a private pilot on real sites. We build the airframe and the software; we are not reselling someone else’s drone.
How it works
This is the engine behind Quill, Keystone and Atlas, our three business platforms. Follow one prescription through it: read, checked, flagged for a pharmacist where the number was unclear, then written into dispensing. The steps and checks are the real ones; the document is a made-up example.
Ingest
Extract
Validate
Checkpoint
Commit
Input
prescription_scan.pdf
inbound fax · 1 page · 240 dpi
Validation
Interaction screen
No conflicts · 4 active medicines
Stock
In stock · 3 packs at this branch
Formulary
Covered · no substitution needed
Structured output
threshold 0.90
Below threshold — queued for a pharmacist, never auto-filled.
Commit
holding
No record written — extraction and validation still running.
Nothing is written until every field clears.
Trace
What it replaces
The usual manual flow: every field re-typed from a fax by hand, and a phone call to the prescriber for anything unclear.
Integrations
FHIR endpoints, dispensing systems, national drug reference data and branch stock records.
Where a person stays
Any field the system is not sure about, plus final sign-off. Nothing reaches dispensing unsigned.
A walkthrough, not a live log. The steps, checks and audit record work this way in the product; the document and its values are made up.
Architecture
Quill, Keystone and Atlas share one foundation. Reading, checking, the hand-off to a person, and the record of who decided what are built once — so a new industry needs a new layer, not a new system.
A fax, an email attachment, a phone photo or a system feed — everything arrives in one place, and we keep a record of where each one came from.
Each field goes to whatever answers it most reliably. A fixed rule handles it where a rule is enough; AI is used only where real judgement is needed.
Before anything is trusted, it is checked against your own data and the industry references — and against the other documents it should agree with.
Anything the system is unsure about goes to a person, with the exact spot on the page highlighted. What they decide, it remembers.
The result is written into the software you already use — once, never twice — with a permanent record of how every field was decided.
Engagements
The same team that built the six products above takes on client work — designing, modernising or shipping software you own outright at the end.
AI systems
We find the places where AI genuinely helps, build it into how the work actually runs, and tell you plainly where it would not help at all.
An honest answer before any big spend
Custom software
Web, mobile and cloud applications built around how your business really works, rather than bending your process to fit someone else’s product.
Code and documents you own
Architecture
Restructuring a system that has become slow to change, expensive to run, or risky to touch — without stopping the business to do it.
Change it without stopping the business
Cloud engineering
Infrastructure that deploys the same way every time, so releases stop being an event and nobody wonders what happens at 2am.
Infrastructure as code, set up the same way every time
System integration
Getting the systems that should talk to each other actually talking, so nobody is left copying data between them by hand.
One source of truth, watched end to end
AI automation
Automating the repetitive work that is too variable for rigid rules — with a clear line between what it decides and what it hands to a person.
Hands off to a person, never guesses silently
Stack
We build on established cloud platforms and proven architecture, and we use AI only where we can show it does the job better than a simple rule would.
Language models and retrieval applied where they measurably improve a workflow.
Managed cloud services chosen for reliability, scale and predictable operating cost.
Proven languages, frameworks and data platforms for systems built to last.
Cross-platform mobile applications that share a single, maintainable codebase.
We build the drones we fly, so the airframe and sensors can be specified for the site rather than the survey being shaped around an off-the-shelf unit.
Technology is selected per product. Not every Arclen AI product uses every technology listed here.
Tell us about the problem you are trying to solve and we will tell you whether technology is the right answer.
