
Forty Bank Accounts, Nine That Matter
Most finance teams have more bank accounts than anyone can justify. Here is how to inventory them, prove which ones earn their keep, and close the rest without breaking a single reconciliation.
Bank account rationalisation is the exercise of listing every account the company holds, scoring each one on whether it still serves a purpose, and closing the accounts that do not. The usual outcome in mid market groups is that roughly a quarter of accounts carry almost no activity but still generate fees, reconciliation work and fraud surface. The work is data assembly first and banking second.
- Bank accounts accumulate through acquisitions, entity changes, one off projects and legacy relationships, and almost nobody removes them afterwards.
- Every open account costs three things: maintenance and analysis fees, monthly reconciliation effort, and one more place a fraudulent payment can originate.
- The inventory is the hard part. Signatory lists, mandates and statement feeds usually live outside the ERP, in spreadsheets and portals.
- Score accounts on volume, purpose, entity, currency and control coverage before deciding anything. Closure order matters more than closure speed.
- ChatFin reads the statement and fee data, builds the activity picture per account, and prepares the ledger cleanup entries for a person to approve.
Nobody sets out to run forty bank accounts. They arrive one at a time, each with a reason that made sense on the day.
An acquisition brings six. A new country needs a local collection account. A lender requires a segregated debt service account. A project team opens one for a grant. A payroll bureau asks for a dedicated funding account. Ten years later the treasury inventory is a list nobody has fully read, and several of those accounts have not moved in eighteen months.
The cost is quiet, which is why it survives. This page walks the rationalisation end to end: how to build the inventory, how to score each account, the closure sequence that avoids breaking anything, and where an AI layer takes the assembly work off the team.

How forty accounts happen
Account counts grow through events, not decisions. Each event has an owner who needed a bank account quickly, and no owner who is responsible for removing it later. Acquisitions are the largest single source, because the acquired group arrives with its own banking relationships and those relationships are rarely unwound in the first year when everyone is busy elsewhere.
Entity changes are the second source. A subsidiary is renamed, merged or dissolved, and the account attached to it stays open under the old mandate because closing it means finding the signatories. Project accounts are the third: opened for a grant, a construction draw, an escrow or a one off programme, and left behind when the project ends.
What an idle account actually costs
The direct fee is the smallest part. Account maintenance and analysis charges on a low activity account are usually modest, and that is exactly why the account survives every cost review. The larger costs sit in operations and in risk, and neither shows up as a line item anyone owns.
Reconciliation effort
Every open account is a monthly reconciliation, a statement feed to maintain, a mapping to keep current and a set of unreconciled items to chase. That effort scales with account count, not with balance.
Fee leakage
Low activity accounts rarely get their account analysis statements read. Charges drift, negotiated rates lapse, and nobody notices because the amounts are small per account.
Fraud surface
Each account is another set of credentials, another mandate, another payment origination point. Dormant accounts are attractive precisely because nobody watches them.
Control coverage
Positive pay, dual authorisation and daily balance monitoring are configured on the accounts people think about. The forgotten ones often have none of it.
Put together, an account that costs very little in fees can cost a real fraction of an analyst month each year, and it carries risk that is not proportional to its balance at all.

Building the inventory nobody currently has
The blocking step is almost always the inventory. Bank account data is scattered by design: the ERP knows the GL accounts, the treasury system knows the ones that were onboarded to it, the bank portals know the rest, and the mandates live in a folder with the company secretary.
A usable inventory needs one row per account with enough context to make a decision. Anything less turns the review into a series of email threads.
| Identity | Bank, branch, account number in masked form, currency, country, legal entity, date opened |
| Purpose | Why it exists in one sentence, plus the person or process that depends on it |
| Activity | Credits and debits by month for the last 24 months, peak balance, average balance |
| Cost | Maintenance and analysis fees for the last 12 months, taken from the account analysis statement |
| Control | Signatories, mandate date, positive pay status, dual authorisation, statement feed type |
| Ledger | The GL account it maps to and whether that GL account reconciles cleanly today |
Two of those rows are where teams stall. Twenty four months of activity means pulling statements in several formats from several portals, and fee history means reading account analysis statements that are designed to be difficult. Both are document reading problems, which is why they are a natural fit for an agent.
The software that helps, and where it stops
Treasury management systems are the right home for the ongoing bank account inventory. Kyriba, GTreasury and Nomentia all maintain account registers, signatory records and bank connectivity, and larger groups usually already own one. Trovata and Agicap sit lower in the market and are strong on aggregated balance and cash visibility through bank APIs.
The gap is the same in all of them, and it is not a criticism of the category. A treasury system tells you what it has been told. It holds the accounts that were onboarded, with the fee data somebody keyed in and the purpose somebody typed. The accounts that most need rationalising are the ones that were never onboarded in the first place.
Building the first complete inventory is a document and reconciliation exercise across all four. That is the part ChatFin automates, and it is why teams often run the rationalisation on the same agent layer they already use for autonomous reconciliation.

Closing in the right order
Closure is where rationalisation programmes go wrong. An account closed before its dependencies are moved creates returned payments, failed direct debits and a payroll incident. The order matters far more than the pace.
Freeze new dependencies
Stop anyone attaching new mandates, direct debits or customer remittance instructions to any account on the closure list. This is a policy step and it costs nothing.
Redirect the inflows
Move customer remittance instructions, card settlement deposits and interest credits first. Inflows are the slowest to move because they depend on third parties acting.
Redirect the outflows
Move direct debits, standing orders, payroll funding and tax payments. These are within your control and move faster once the inflows are proven.
Run a dormancy window
Leave the account open with zero expected activity for a full cycle, usually 60 to 90 days, and watch for anything that still lands in it.
Close and clear the ledger
Close with the bank, obtain written confirmation, and post the final clearing entry so the GL account is zero and can be deactivated in the chart of accounts.
The dormancy window is the step most often skipped and the one that catches the annual charges, the quarterly rebates and the one customer who still pays into the old account every March.
How ChatFin runs the assembly work
ChatFin sits on top of the ERP you already run and reads the documents that hold this data. The document processing agent parses bank statements and account analysis statements in whatever format the bank supplies them, including scanned files, and turns them into structured activity and fee records per account.
From there the autonomous reconciliation agent maps each bank account to its GL account and shows the reconciliation position over the whole review window, which is what tells you whether an account is genuinely dormant or simply unreconciled. The conversational finance AI layer lets a treasurer ask plain questions against that data without waiting for a report.
| Document processing | Reads statements and account analysis charges per account, including scanned and portal exports |
| Reconciliation agent | Maps bank to GL, shows unreconciled items per account across the full 24 month window |
| Data query agent | Answers questions such as which accounts had no third party activity in twelve months |
| Pattern recognition | Flags accounts where fee charges continue after activity stops, the clearest closure candidate |
| Governed write back | Prepares the final clearing entries and GL deactivation requests for a person to approve |
The last row is the one that matters for control. ChatFin does not post anything on its own. It prepares the entry, routes it through the approval chain you already use, and posts through the ERP interface only after a person releases it, which is the same governed pattern described in adding AI without replacing your ERP.

A four week version of this project
Rationalisation is usually scoped as a six month programme and then abandoned. A tighter version works better because it produces a decision list quickly.
Week one, gather
Pull twelve months of statements and account analysis for every account anyone can name, from every portal. Do not filter yet.
Week two, score
Build the one row per account view, score each on activity, purpose, control coverage and fee cost, and split into keep, review and close.
Week three, agree
Take the close list to the entity controllers and the lender or regulator owners. Expect a third of the list to come back with a dependency you did not know about.
Week four, sequence
Publish the closure sequence with the dormancy windows dated, and start redirecting inflows.
The output at the end of four weeks is not a set of closed accounts. It is a defensible list with owners and dates against it, which is what makes the closures actually happen over the following quarter.
An account that costs almost nothing in fees can still cost a real fraction of an analyst month, and carries risk that has nothing to do with its balance.
ChatFin
Frequently Asked Questions
How many bank accounts should a company have?
There is no correct number, only a correct test. Each account should have a named purpose that cannot be served by an existing account, an owner who depends on it, and control coverage proportionate to the payments that can originate from it. Applied honestly, that test usually removes a meaningful share of the accounts a mid market group holds, most often the ones inherited through acquisitions and entity changes.
What is the risk of closing a bank account too quickly?
Returned inbound payments and failed outbound ones. Customers pay into accounts long after being told to change, and direct debits, standing orders and tax mandates can sit against an account without appearing in any current process document. Running a dormancy window of 60 to 90 days with zero expected activity before closing is what catches these, along with annual and quarterly items that would otherwise be missed entirely.
Can ChatFin connect to bank data as well as the ERP?
ChatFin reads bank statements and account analysis statements as documents and reconciles them against the ERP ledger, so it works with whatever statement formats and portal exports your banks provide. It sits as a governed layer above the ERP you already run, reads through the interfaces that ERP exposes, and writes back only entries that a person has approved.
Does this work if our bank accounts are not in a treasury management system?
That is the more common case and it is the one this approach is built for. The accounts most worth rationalising are usually the ones that were never onboarded to a treasury system, so an inventory built from statements and fee documents will find accounts the treasury system does not know exist. Once the register is complete it belongs in the treasury system permanently.
See ChatFin build your bank account inventory from source documents.
ChatFin reads your bank statements and account analysis statements, builds the activity and fee picture per account, and reconciles each one against its GL account across the full review window.
Every closure entry, clearing posting and chart of accounts change is prepared as a draft and posted only after a person approves it, through the ERP interface you already use.
Book a DemoBring your statements. We will show the inventory.
Send a sample of bank statements and account analysis statements and ChatFin will assemble the per account activity and fee picture that the rationalisation decision needs.