Seven production AI products, built in Perth on our own multi-signal fusion engine. Each one solves a specific real-world problem — and every one keeps a human in control of the outcome.
Most AI vendors sell you a model and leave the hard parts to you — governance, privacy, integration, and the awkward question of what happens when the model is wrong. We build the whole thing.
Every DOT product runs on the same foundation: multiple weak signals fused into one calibrated assessment, statistically validated before it reaches a person, and a governance layer that decides whether the system acts, escalates, or defers to a human. Sensitive data stays on your infrastructure by design, not by policy.
The underlying research is registered with AusIndustry under the R&D Tax Incentive and validated using published statistical methods — so when you stake your reputation on an output, there's evidence behind it.
Rule-based fraud systems catch yesterday's fraud and block today's genuine customers. TransACT learns each individual user's behavioural fingerprint — typing cadence, typical transaction size, usual login locations, session rhythm — and flags the moment that pattern breaks.
Real-time velocity checking catches micro-transaction clusters and card-testing sequences as they happen, not in the next day's batch report. Ensemble modelling combines fast numeric scoring with context-aware text encoders, which is what keeps false rejections down — genuine customers aren't blocked at checkout.
Mobile wallets, banks, payment processors and any organisation running high-volume transactions.
Edge inference — all scoring runs inside your environment. No transaction data leaves your infrastructure.
The gap between a question and an answer in most organisations is an analyst, a ticket queue, and three days. InsightDOT removes it. Executives type a question or a request in plain language and get a report or a dashboard back in seconds.
It works like a general-purpose AI assistant, except it's grounded entirely in your organisation's own data — not the public internet, not model memory. Every figure it quotes traces back to a source in your warehouse, which means the output is auditable and defensible in a board pack.
Executive teams, boards, and anyone who needs answers faster than the reporting cycle allows.
Microsoft Fabric, Power BI, Snowflake, Azure and Databricks — sitting on top of your existing estate.
Keyword search fails when you don't know the exact wording used in a document written six years ago by someone who has since left. LensDOT searches by meaning — ask a question the way you'd ask a colleague and it returns the ten closest-matching documents from across your entire library.
Contracts, policies, tender responses, incident reports, technical manuals, board minutes. Anything sitting in a shared drive or document management system becomes searchable in natural language, which turns a half-day of hunting into a ten-second query.
Legal, compliance, procurement and operations teams working across large document estates.
Indexes run in-country. Document contents never leave your environment.
Schools already hold the signals — attendance patterns, behaviour records, academic trend. They're just scattered across systems nobody reads together. ClassMate reads them as a set and gives each teacher a plain weekly view of every student: Settled, Watch, or Support.
It is deliberately not a score and not a diagnosis. It doesn't rank children or predict outcomes. It surfaces a pattern early enough for a teacher to have a conversation, and the teacher always decides what happens next. That constraint is a design principle, not a limitation.
Primary and secondary schools, school networks, and education departments.
Live pilot deployment. Active engagement with education systems across multiple regions.
Humanitarian field teams typically operate on lagging indicators — incident reports that arrive after the incident. EmpathAI applies the same multi-signal fusion approach to field data so teams can act on early indicators instead of after-the-fact reporting.
It runs on a field-partnership model: Data One contributes the technology in-kind, and the field partner contributes access, context and clinical oversight. Neither side is equipped to do this alone, and the governance layer means no automated system makes a decision about a vulnerable person without a qualified human in the loop.
Humanitarian agencies, NGOs and field programmes operating in low-resource settings.
In-kind technology partnership. Aimed at UNHCR and the UNICEF Venture Fund.
Most AI governance is a policy document. This is a working layer that sits between the model and the user, and every output passes through one of four decision states: Allow, Modulate, Escalate, or Override.
It was built in response to research on young people using AI for mental health support — a context where a fluent, confident, wrong answer causes real harm. The same architecture applies anywhere an AI system produces consequential outputs and a human needs to stay accountable for them.
Organisations deploying AI in mental health, youth services, healthcare or regulated environments.
In development. Independent evaluation pathway and clinical partnerships in progress.
Good prompts get lost. They live in old chat threads, scattered docs, and somebody's memory — rewritten from scratch every time. PromptEdge puts your saved prompt library one click away inside ChatGPT, Claude and Gemini, with variables so a template built once gets reused indefinitely.
It also tells you what a prompt is likely to cost before you run it rather than after. For teams running prompts at volume — content production, recurring workflows, outreach — that's the difference between managing a budget and discovering it.
Individuals and teams using AI tools daily who want consistency and cost visibility.
Browser extension and web app. Shared team libraries with role-based access.
Every system reads multiple weak signals before drawing a conclusion — the way an experienced analyst or clinician would, rather than reacting to one data point in isolation.
Privacy is built into the architecture through edge inference. Sensitive data doesn't leave your environment because the system isn't designed to send it anywhere.
Every AI output carries a governance state. The system provides the information; your people make the call on anything consequential.
R&DTI-registered with AusIndustry and statistically validated using published methods — so there's evidence behind the numbers when a board or regulator asks.