4 roles
The AI powered stack
Not a stack so much as a filter. AI in broking is worth paying for where it removes typing or removes waiting, and is worth nothing where it produces text somebody then has to check line by line against a compliance obligation.
For: A brokerage of any size deciding where AI earns its subscription and where it does not.
The roles this stage has to fill
One tool per role, no overlap. We have not named the tools, because naming them is a recommendation five deep and no product has been tested yet — what follows is the shape, and the whole field for each part of it.
Meeting capture and file notes. The clearest win: it removes typing from the part of the job with the highest note-taking burden.
19 products we track →
Chat and voice agents. Worth it for triage and booking; not for advice, and not for anything a client could act on.
23 products we track →
Open banking and categorisation. This is arithmetic rather than judgement, which is exactly what to automate.
13 products we track →
Automation between tools. Most of what gets sold as AI in broking is this, and it is genuinely valuable.
17 products we track →
What this stage leaves out
A stack is defined as much by what is not in it. Every one of these is a real category a broker could buy from today, and at this stage it does not earn the money.
- Anything generating client-facing credit advice — the obligation stays with the broker either way
- Anything that writes a file note nobody reads before it is saved
Before this names a single product
Every tool in each role has to have been run in a real account for a working week and scored against the published criteria, and its list price confirmed in AUD rather than converted. Until then this gives you the shape and the field, which is worth more than a stack naming three products nobody has used.