AI agents for finance.

WhatWill AI builds agents that do the reading, gathering and first drafting a finance business runs on. Research, due diligence documents, monitoring, briefing packs and recurring reports, written into your template with a citation against every claim. Your people review and sign instead of starting from a blank page. No views, no recommendations, no advice. Built on your own archive, priced in AUD, built by the founder.

or see how we build it

What AI for finance means here, and what we build

Most of the work in a finance business is not the judgement. It is finding the filings, reading the last four announcements, going through a data room before a deal, digging out what your own team wrote about this company eighteen months ago, pulling the client history together before a meeting, and reformatting all of it into the house template. Then the experienced person finally starts on the part only they can do, usually at six in the evening.

An agent takes the first part. It reads the sources you have approved, drafts into your format, and shows the source next to every line it has written. Your analyst, adviser or broker opens a draft with the gathering done and the workings visible, and spends their time on the judgement. Nothing is issued until a named person has checked it and signed it.

Research is usually the first job worth building, because it is the biggest pile of hours and the easiest to mark. It is not the only one. Document review, monitoring, briefing packs, meeting preparation, triage and reporting drafts all run on the same pattern, and they get added once the first one is producing work your team trusts.

It does not form a view, rate anything or recommend anything, and we would not build it that way if you asked. The view stays with your qualified people, where the responsibility already sits. We build AI systems. We are not a licensed adviser and we do not give financial advice.

Every job is led by the founder. No juniors on your account and nothing outsourced. You get the experienced person from the first call, and the system is yours to keep.

Where the senior hours actually go.

Three things eat the week in a finance business, and none of them are the analysis. These are the parts worth handing over.

01

Gathering and reading the same sources again

Filings, announcements, transcripts, a data room, product and lender documents, and whatever your own team wrote last time. Found by hand, every time, by whoever has the least on.

02

Reformatting into the house template

The content exists and then an hour goes into making it look like your work. An agent writes into the template from the start.

03

The turnaround you cannot make

An announcement lands, a client wants a read on it today, or a meeting moves up a week. The work is doable. The gathering is what makes it late.

The jobs we build agents for.

Research leads for most firms. We scope which of the rest earn their place for the work you actually produce before building anything, and we say so when the honest answer is none of them.

Research gathering from your list

Filings, ASX announcements, reports and transcripts pulled from the sources you have approved and are entitled to use, then drafted into your research format.

Source checking and a citation on every claim

Each line carries the source it came from and the figure is quoted and linked rather than recalled, so an unsourced sentence is obvious at a glance instead of buried in good prose.

Due diligence and document review

A data room, a contract set or a pile of statements read against the checklist you use, with what it found, where it found it, and what is missing or inconsistent.

Market and company monitoring

A watch on the names, sectors and themes you follow, with a short sourced summary of what changed, what it touches and which of your clients or holdings it affects.

Cited briefing packs

The weekly or monthly pack assembled from the week that happened, in your running order, with every point traceable back to where it came from.

Client and meeting briefs

The one-pager before the meeting: the history, the last file notes, what was promised, what moved since, and the questions worth asking. Assembled from your own systems.

Inbox and document triage

What arrived sorted, named, filed against the right client or deal, and a reply drafted where a reply is obvious. Nothing sends itself.

Reporting and update drafts

The recurring report or client update written as a first draft from the numbers and notes your systems already hold, so the work is checking rather than typing.

Your own archive, searchable

Past notes, models, file notes and CRM commentary made findable, so every agent builds on what your team already concluded.

The line we will not cross.

AI in finance goes wrong in one specific way: a system that looks confident, forms a view, and nobody checks it because it reads so well. We build against that on purpose. The agents gather, read and draft. Your people decide and sign. Licensee obligations are technology neutral, which ASIC said plainly in its 2024 review of AI governance in financial services, so none of this moves responsibility off your firm, and we would not sell it as though it did.

  • It does not form a view, rate, score or recommend anything
  • It does not produce personalised financial advice for a client
  • It does not trade, allocate, or act on anything by itself
  • It does not send, file or issue anything without a named person signing it
  • It does not read from sources your licence terms do not allow
  • We are not a licensed adviser and we do not give financial advice
  • Nothing we build makes your firm compliant. That stays your call and your governance

Pick the one that fits your firm.

Whether you want the agents run for you or handed to your own team, one of the three will fit.

01

Fully Managed

We build the agents and keep running them: maintaining your source list, tuning the drafts, and adding jobs as you need them.

02

Build & Hand Over

We build it on your sources and your templates and hand it to your team, with documentation and training.

03

AI Consulting

Not sure it is worth it yet? We look at the written work your firm produces and where the hours go, and tell you straight.

When this is worth it, and when it is not.

It earns its place when written work goes out on a rhythm and the gathering costs more than the thinking. When a better template or a tidier filing system is the honest answer, we say so.

  • You produce a couple of written pieces a year
  • Your past work is not kept anywhere you could point to
  • Nobody has time to review a draft before it goes out
  • What you want is the internal money function automated, so payables, payroll and the month-end close
  • The sources you depend on do not allow this use, and there is no substitute

Finance businesses this suits.

It pays off fastest in small financial services firms where written work goes out regularly and experienced people are doing the gathering.

  • Financial advice practices and private wealth teams
  • Mortgage, finance and insurance broking firms
  • Boutique research houses and broking firms
  • Corporate finance, M&A and transaction advisory teams
  • Family offices and investment committees
  • Commercial lenders and credit assessment teams

AI agents for finance: frequently asked questions.

What does an AI agent actually do in a finance business?

It does the gathering and the first draft of the written work. You give it a job, such as a research question on a company, a pile of documents to read before a deal, or the client pack due on Thursday. It pulls from the sources you have approved, such as ASX announcements, annual reports, your data subscriptions, your product research and your own past notes, then writes into your template with a citation against every claim. What comes back is a draft in your house format with the source next to each line, ready for one of your people to check, correct and sign.

Which jobs can you build agents for?

Research gathering with the sources checked, due diligence and document review, monitoring of the names and themes you follow, cited briefing packs, client and meeting brief preparation, inbox and document triage, and first drafts of recurring reports. Research is usually the one worth building first because it is the biggest pile of hours. The rest get added once the first one is producing drafts your team trusts.

Does it give investment advice or recommendations?

No, and we will not build it that way. It does not form a view, rate anything, score anything or recommend anything, and it does not produce personalised financial advice for a client. It gathers what the sources say and drafts the summary. The view, the recommendation and the responsibility stay with your qualified people, exactly as they do now. We also do not provide financial advice ourselves. We are an AI build shop, not a licensed adviser.

How do you stop it making things up?

Three ways. It only reads from the sources you have named, so it is not roaming the open web unless you want it to. Every sentence in the draft carries the source it came from, so an unsourced line is visible immediately rather than hidden in good prose. And where a number matters, the agent is built to quote and link the figure rather than restate it from memory. The point is to make checking fast, not to remove it.

Who reviews the output, and is that on us?

A named person on your side reviews and signs every draft before it goes anywhere. We build that step into the system rather than leaving it to good intentions: a draft is marked as a draft, it carries who it is waiting on, and it cannot be issued without that sign-off. If your firm is an AFS or credit licensee, your existing obligations and governance stay yours. ASIC has made the point directly, in its 2024 review of AI governance in financial services, that licensee obligations are technology neutral, so adopting AI does not change what you owe.

What sources can it use?

Whatever you are entitled to use and are comfortable putting into a system. Commonly that is company filings and ASX announcements, annual and half-year reports, transcripts, product and lender documents, your own research archive and CRM notes, industry bodies and regulators, and any data subscription whose terms allow it. We check the licence terms of a paid source with you before wiring it in, because that is a real constraint and not a detail.

Is our research archive and client data kept confidential?

Yes. It runs on business-tier AI that does not train on your inputs, and your archive stays in your own systems. The OAIC guidance on commercially available AI products recommends not entering personal information, particularly sensitive information, into publicly available generative AI tools, so we write down exactly which tools are approved and what may go into each one before anything real touches a system.

Is this for finance teams inside a business, or for financial services firms?

Financial services firms, and the professional teams around them. Advisers and advice practices, mortgage, finance and insurance broking, boutique research and broking houses, corporate finance and M&A, family offices and investment committees, and commercial credit teams. If what you mean by finance is the internal money function, so accounts payable, invoice coding, payroll and the month-end close, that is a different build and this is not the page for it. Accounting and bookkeeping practices are on our AI for accountants page.

What does it cost and how long does it take?

A first version covering one job usually takes a few weeks, most of it spent on your sources, your template and testing drafts against work you have already produced and can mark. We scope it on a short call and quote in AUD before any work starts, with no surprise add-ons. More jobs get added once the first one is earning its place.

Let’s talk about
the work you keep redoing.

Book a free 30-minute discovery call. We’ll look at the written work your firm produces and where the hours go, and tell you straight which agent is worth building first.

Book a callFree, 30 minutes, no jargon.