The Client Portal and the Advice Desktop are what you see. This is what runs underneath.
An adviser-connected intelligence layer needs more than a good model. It needs agents that share context, follow your firm’s rules, and act as one system. This page describes that layer, what it will do for Australian firms, and how it is governed.
A human workforce, with an agentic one behind it.
An Agentic OS is a unified layer that coordinates AI agents across workflows — so they share context, follow rules, and act as one system instead of isolated tools. In the US, TIFIN.AI runs this in commercial production. In Australia, it is the roadmap we are building with our founding firms, released only as the harness holds for each capability.
Growth workflows
Agents that help product and advice providers execute distribution and sales efforts more efficiently.
Investment workflows
Agents that handle investment tasks, reporting, portfolio management, and other investment objectives.
Operational workflows
Agents that automate and execute middle and back office operational functions such as onboarding, monitoring and servicing.
Where agents do the work, and who they answer to
The Advice Desktop is where advisers already meet this layer: AI prepares, drafts and ranks; the adviser reviews, approves and advises. As capability widens, the same rule holds for every role.
Financial advisers
Financial agents that prepare and propose on behalf of clients under defined risk, compliance and fiduciary guardrails — with the adviser approving what reaches the client.
Operations teams
Streamline compliance, account opening, financial planning and ongoing account monitoring workflows.
Investment teams
Scale high-touch portfolio construction, analysis and communication while delivering personalised portfolio insights at speed.
End clients
Agents let advisers reach mass-affluent clients, replacing advice that goes quiet between reviews with continuous, adviser-governed attention.
Context is the hard part. That’s what we’ve built.
Some assume a powerful enough model will figure out context and coordination. It won’t. The model doesn’t know your client just lost a parent, has a tax liability this quarter, or panic-sold in 2020 — or what is relevant for which workflow. Four layers do that work, and every one of them is auditable.
Assembled, enterprise grade
Including privacy, compliance, permissioning, observability and explainability.
Human–system experience layer
Interface workflow layer where humans and systems interact through build-your-own-agent experiences, enabling latency-aware execution paths across multilingual, multimodal inputs and outputs.
Agent orchestration layer
System that coordinates multi-agent workflows across native and third-party agents, with optional evaluation-gated tuning to optimise performance and outcomes.
Agent library layer
Model-agnostic agent layer that enhances context extraction and model accuracy through workflow schemas, knowledge graphs, hallucination guardrails, personalisation, context planning, SLMs and self-learning feedback loops.
Context layer
Unified data and execution layer that combines knowledge graph–based data unification with tool access, business goal alignment, and rule/exception handling to enable intelligent action.
Founding firms decide what we build next.
Which workflows come first is a decision we make with the firms in the room. A conversation, not a commitment.
Become a founding firm