What We Found When We Audited Finance Team Month-End Processes

The Problem With Month-End Close Is Rarely the ERP

Finance teams on modern ERP platforms, whether D365, SAP, Oracle, or Workday, often assume that if they have a system, they have automation. In practice, the ERP is usually configured to handle the core transaction processing, but the close process itself has grown up around it through a combination of workarounds, institutional habits, and steps that nobody has questioned in years.

When we run a close process audit, we are not looking at the system. We are looking at what the team actually does. Who opens which spreadsheet, what gets copied where, which approval gets chased over email, which reconciliation is done manually because the system matching logic was never set up properly. These are the steps that accumulate into a 6-day close when 3 days would be achievable.

The average mid-market finance team we audit is spending between 5 and 8 working days on close. Our audits consistently identify 2 to 3 recoverable days. That is not a technology limitation. It is a process and configuration problem.

What the Audit Actually Finds

A close process audit is a structured mapping exercise. We sit with the finance team, walk through every step in close sequence, and document what is happening, who is doing it, how long it takes, and whether it needs to happen at all or could be automated. The output is a list of manual steps ranked by time cost and automation potential.

Across the audits we have run, several patterns appear almost every time:

None of these are unusual. They are the default state of most finance functions that went live on an ERP without a post-implementation process review. The system handles transactions. The close process was never fully designed.

How AI Tools Address Each Bottleneck

Once the audit identifies where the time is going, the fixes become specific rather than general. Most do not require new software. They require either ERP configuration changes (fixing what was never set up properly) or AI tools working alongside the existing system.

Journal Automation

Standard recurring journals do not need to be prepared manually. Microsoft Copilot for Excel can generate journal templates from your prior period data and flag any period-on-period anomalies before posting. For teams on D365, the recurring journal functionality within the system handles standard entries without any manual input once configured. The audit establishes which journals are genuinely standard (and should be automated) and which require human judgment each month.

Approval Routing Through n8n

Replacing email approval chains does not require buying a workflow tool. n8n, which is an open-source workflow automation platform, connects to your ERP, captures the approval trigger, routes it to the right person via Teams or email, captures their response, and updates the system. The approver does not need to log into the ERP. The approval happens in whatever channel they are already in, and the system is updated automatically.

Claude Projects for Commentary and Review

Claude Projects is the most practically useful AI feature for month-end close work. A Project lets you build a persistent knowledge base for your specific close workflow. You add your chart of accounts, your standard commentary format, your prior three months of management pack text, and your close checklist as project knowledge. Claude then has full context every time you use it during close, without you re-explaining your cost centre structure or preferred commentary style each session.

Within a Project you can also set up Skills, which are custom tools for specific tasks. A “variance commentary” skill, for example, takes your actuals-versus-budget data as input and produces first-draft commentary in your standard format. The analyst reviews it, adds business context and any qualitative factors the numbers do not capture, and publishes. Commentary that previously took three hours takes forty minutes.

Claude Connectors extend this further by letting you connect the Project directly to the Google Drive folder or SharePoint site where your close packs and prior period files live. Claude reads the files directly rather than requiring you to paste data in manually.

Copilot for Excel Reconciliation Review

For teams doing reconciliations in Excel, Microsoft Copilot for Excel identifies unmatched items, flags statistical outliers, and produces a summary of exceptions without manual scanning. You still review and clear the exceptions. You stop spending time confirming the items that match.

A Real Audit: Scottish Manufacturing Business

A Scottish manufacturing business came to us 18 months after going live on D365. The system was running, but the finance team was still spending close to four days on each month-end close. The CFO knew something was wrong but had not had the capacity to map where the time was going.

We ran a two-day finance function audit, documenting every step in their close sequence. We found eight manual processes that the system either could automate or should have been handling already: journals being prepared in Excel outside D365, approval routing happening over email, 3-way matching being done manually because the goods receipt workflow had never been properly configured, and commentary being written from scratch each month with no templates or AI assistance.

We rebuilt the approval chain configuration inside D365, fixed the goods receipt and journal automation, and set up Claude Projects with their chart of accounts and close checklist for the commentary workflow. We ran hands-on training with the finance team so the manual workarounds did not return.

Month-end close dropped from four days to under two. The time recovered went back into analysis and business partnering work rather than being absorbed by the next manual task.

What to Do If You Have Not Run a Close Audit

If you have not mapped your close process step by step in the last two years, you almost certainly have recoverable time in it. The question is how much and where.

The fastest way to find out is a structured audit. You can do a lightweight version yourself by asking your team to log every close task and how long it takes for one full cycle, then reviewing where the manual steps are. For a more thorough review, our AI finance audit includes a close process review as a core component, with a written output identifying what is recoverable and what to prioritise.

For teams who want to learn the tools and build the workflows themselves, our AI for Finance Leaders training covers Claude Projects setup, Copilot for Excel, and n8n workflow automation in a finance context. The training is built around the specific workflows finance teams run, not generic AI instruction.

Frequently Asked Questions