The finance team grows with volume
Each new market or entity means another hire, and the team spends its days preparing numbers.
Solutions · AI for finance
AI for finance: agents, vetted partner software and a CFO who runs them
AI for finance is the use of AI agents and AI-native software to do finance work such as the month-end close, reconciliations, reporting and forecasting, with a finance leader accountable for the result. Numminen Advisory brings AI for finance to international startups and growing technology companies.
The work pairs Joanna Numminen’s experience as a CFO with two kinds of AI: software from partners we have vetted, and agents she builds on the company’s own system of record.
- Bank reconciliationDone
- Intercompany matchDone
- Variance commentaryDrafted
- CFO approvalOpen
What AI for finance means at Numminen Advisory
A CFO’s judgement. AI changes who prepares the numbers. It does not change who answers for them. Joanna decides which work moves to AI and signs off what the board and investors see.
Vetted AI-native partners. Where a specialist product already does a job well, such as spend management or payables, we recommend it and connect it. We work with a small group of AI-first finance software companies whose products we have tested in live finance operations.
Agents we build. Work that depends on how the company operates, such as reconciliations or intercompany matching, runs on agents Joanna builds on the company’s ERP and ledgers. For many growing groups that means Oracle NetSuite, with Procountor or Netvisor for Finnish entities.
AI in finance
How AI changes the work of a finance team
AI now prepares most of what a finance team used to assemble by hand, from reconciliations to the first draft of the board commentary. At Numminen Advisory, Joanna Numminen sets this up on the company’s own ledger and runs it with the team, so people spend their time on review and decisions. Nothing posts to a ledger until the CFO approves it.
| Finance task | What AI does now | What stays with the CFO |
|---|---|---|
| Month-end close | Ties bank and sub-ledgers to the general ledger every morning and lists the breaks. | Judgement on accruals and estimates, and sign-off of the close. |
| Payables | Reads each invoice and codes it against the chart of accounts. | Approval of payments and of anything outside policy. |
| Reporting | Assembles the management pack and drafts first-pass variance commentary. | The story the numbers tell, and accountability for them. |
| Forecasting | Rolls the cash forecast forward from the bank feed and flags movements against plan. | Scenario choices and the conversation with the board. |
How we run it
- Each client gets its own project in the workspace Joanna built, with its context and systems connected read only.
- Partner software and our own agents read from the same system of record, so there is one version of the numbers.
- Every run is logged, and a figure in a board pack can be traced back to its source.
When a growing company needs AI for finance
The close depends on one person’s spreadsheets
Reports arrive late, and nobody else can run the month-end when that person is away.
The team is already trying AI tools
People use chat assistants on company data with no controls, and results differ from one person to the next.
An investment round is coming
Diligence will test whether every number can be reproduced from the systems each time a request comes in.
CFO Operating System
AI agents in finance: what they do and what they run on
An AI agent in finance is software that carries out a defined piece of finance work on its own, such as matching intercompany lines or drafting variance commentary, and reports what it did. Agents are fast at volume and repetition. What they lack is the company’s own context, and without it every agent starts from zero.
In a growing company that context usually sits in a few people’s heads and spreadsheets. The close depends on whoever built the workbook, the same judgement call is made differently from month to month, and a figure in the board pack is hard to reproduce when an investor asks where it came from.
Numminen Advisory solves this by building a CFO Operating System for the finance function. It is a deliverable the company owns: a structured set of plain-text files that people and agents both work from, describing how the company keeps its books, the written procedure for each recurring task and where every process stands. Agents run from it on request or on a schedule, and they stop for the CFO’s approval before anything changes a ledger.
- Context. The group structure, accounting policies, account mapping and close calendar are written down once and kept current, so the knowledge no longer lives with one person.
- Skills. Each recurring task becomes a written procedure with its tolerances and the point where the agent must stop and ask. Judgement turns into something the team can inspect and correct.
- Two ways to run. The CFO or the team starts an agent for a board pack or an investor question, while scheduled agents handle the morning reconciliation and the month-end match. Both read the same files, so there is one version of the numbers.
- Approval. Agents prepare and people decide. Nothing posts to a ledger or leaves the company without sign-off, and every run is logged so a figure can be traced back to its source.
Build, buy or connect
Connect the system of record
The ERP and local ledgers stay the source of truth. Everything else reads from them.
Buy from AI-native partners
When a specialist product already does the job, buying it is faster and cheaper than building it.
Build your own agents
When the work depends on how this company operates, agents built on its own ledger fit better than a generic product.
How we vet AI-native partners
A partner product has to read from the company’s existing system of record and respect its role-based permissions. We also check its audit trail and where client data is stored. When we recommend a partner we name it, and we disclose any commercial relationship with it.
What we are not
Numminen Advisory is not a software company and does not resell licences. Nor are we an AI agency building chatbots. We are a finance advisory led by a CFO who uses these systems in her own daily work and is accountable for the numbers they produce.
Who does this work
Joanna Numminen, the founder of Numminen Advisory, was the NetSuite product owner across Finland, China, Japan, Korea and the US at NAPA and led the NetSuite implementation at ONEiO. At Silo AI she built the global finance function and took the group through acquirer diligence. She is ACCA qualified and trained in audit at Deloitte. Today she builds and runs AI agents in her own finance operations.
How to engage
AI for finance review
A fixed-scope look at the finance function, its systems and the AI tools already in use, ending in a written plan for the AI setup. Quoted against scope before the work starts.
Build project
Agents and partner software set up on the company’s system of record, then handed over to the team. Quoted after the review.
As part of a fractional CFO retainer
For companies that want the person who sets it up to keep running it. See fractional CFO.
Amounts exclude VAT. See project packages and pricing.
Common questions
What is AI for finance?
AI for finance is the use of AI agents and AI-native software to do finance work such as the close, reconciliations, reporting and forecasting. In a well-run setup the AI prepares the work, and a finance leader approves it and answers for the numbers.
What are AI agents in finance?
Software that carries out a defined piece of finance work on its own, using the company’s records and written procedures, and reports what it did. Typical examples are reconciliations, invoice coding, intercompany matching and first-draft variance commentary.
Where does an AI for finance engagement start?
With a finance outcome, not a tool. Joanna starts from the work that matters most to the company right now, such as a faster close or a board pack investors trust, and traces who prepares it and from what. The first agent goes where that trace shows the most manual effort.
What is a CFO Operating System?
The CFO Operating System is the deliverable at the centre of an AI for finance engagement. It is a set of plain-text files, built for the company and owned by it, that records how the company keeps its books, the written procedure for each recurring task and where every process stands. Numminen Advisory builds it at the start of the work and hands it over documented, because an agent is only as reliable as the context it is given.
How do you decide what an agent may do on its own?
Judgement is written down as rules before an agent gets the task, including tolerances, the evidence a match needs and the conditions under which it must stop and ask. Agents prepare and draft. Estimates, payments, postings and anything outside policy stay with people, and the CFO signs off what the board and investors see.
What happens when an agent gets something wrong?
Every exception is logged and reviewed. When a rule turns out to be wrong, Joanna corrects the written procedure, reruns the affected step and compares the result. Only logic that has held up in review becomes standard procedure, so the setup grows more accurate each month it runs.
Do we need to change our accounting software?
There is no need to. The work builds on the systems the company already uses, and they stay the source of truth. Where a specialist product already does a job well, we recommend one of our vetted AI-native partners, and custom agents are built only where the company’s own way of working calls for them.
Is our financial data safe with AI agents?
Agents get the least access a task needs and are read only by default. Nothing changes a ledger without approval, and every run is logged, so any figure in a report can be traced back to its source.
Does AI replace the finance team?
No. It changes what the team spends its time on. Preparation moves to agents, and people move to review and decisions. For a growing company the effect is usually that more volume no longer means another hire.
What does the company keep after the engagement?
The company keeps everything we build, in its own environment, starting with the CFO Operating System. All of it is documented so the finance team can run and change it without us, or Joanna can keep running it as part of a fractional CFO retainer.