RM48 Million of Software. The Month Still Closes in Excel.

RM48 million of enterprise software, and finance still closes the month in Excel. A comedy about why finance still uses Excel — and what finally fixes it.

🎂 There Was a Cake

We've been in a lot of rooms with C-level executives lately, and they keep telling us the same story. Different industries, different vendors, same story. So let's just tell it.

One of them put a number on his: RM48 million. A multi-year enterprise application programme. Board-approved. A steering committee so large it had its own steering committee. Delivered. Live. Officially transformed.

There was a cake. Project logo printed on top in edible ink. By all accounts the cake was excellent, and it was also the only deliverable that arrived on schedule.

And every month since go-live, his finance team closes the books in Excel.

Not as a backup. Not as a "transition period." As the actual process. The one the auditors get shown. RM48 million later, the beating heart of a large Malaysian enterprise is a file called CLOSE_AUG_FINAL_v9_LATEST(2)_USE-THIS-ONE.xlsx.

He told us this story laughing.

That's the part you should worry about.

🏆 Why Finance Still Uses Excel (A Love Story)

Before we mock the spreadsheet, let's be fair to it — because nobody else ever is, and honestly, it's earned better.

Consider the track record. Excel has a 100% adoption rate. A training budget of zero. No licence negotiation, no vendor lock-in workshop, no eighteen-month roadmap to value. Its implementation timeline is double-click. If Excel had an enterprise sales team, it would be the most feared software company on Earth. It doesn't need one. It already won.

And here's why finance still uses Excel after every implementation you'll ever fund: your ERP knows what the process should be. The spreadsheet knows what the process is. One of them is lying, and it isn't the spreadsheet.

Somewhere in your organisation right now, the consolidated position of an entire group of companies depends on one workbook, maintained by one person we'll call the Keeper of the Macro. The Keeper is brilliant, underpaid, and currently on annual leave. Nobody else opens the file. Opening the file is considered bad luck.

Is this fine? It is not fine. Research out of the University of Hawaii found roughly 88% of spreadsheets contain errors (Source: Ray Panko, University of Hawaii) — and frankly we assume the other 12% just haven't been audited yet.

But "error-prone" and "irrational" are different accusations. Excel keeps winning for a reason. To see it, you have to visit month-end.

🌙 Month-End: Season Seven

Ask for a description of the month-end close and you'll get the policy-manual version. Clean. Sequential. Fits on one slide. This is the version the RM48 million was scoped against.

Now ask the person actually doing it at 11pm on Day 3, somewhere between their second teh tarik and a small crisis of faith.

You'll meet the real cast. The intercompany eliminations between two charts of accounts that were never harmonised, because harmonising them was "Phase 2," and Phase 2 is a place projects go to rest. The accrual that arrives — as tradition demands — as a photo of a handwritten note, taken at a slight angle, posted in a WhatsApp group called URGENT (do not reply). The FX revaluation for three legacy entities the migration quietly skipped. And the mysterious reclassification an auditor requested in 2019 that nobody has dared remove since, because nobody remembers why it exists and nobody is brave enough to find out.

None of this was in the RFP. All of it is the job.

The expensive system was built for the process. The spreadsheet absorbs the exceptions — and the exceptions, it turns out, are where the real work lives. This is precisely the neighbourhood where large IT projects go to underdeliver: the famous McKinsey–Oxford study found they run 45% over budget while delivering 56% less value than promised (Source: McKinsey & Oxford University). The missing 56% doesn't vanish. It moves to Excel and stops paying rent.

💸 A Change Request Costs RM40,000. Excel Costs a Teh Tarik.

Here's the quote we can't stop thinking about, from a finance director (details blurred, wording intact — it's a composite of several conversations, and every one of them said some version of this):

"A change in the system costs RM40,000 and takes eleven weeks. Excel costs nothing and takes an afternoon. What exactly do you want me to do?"

He's right, and that's the punchline of the whole genre. The system change needs a request form, a business case, a vendor quote, a release window, UAT sign-off, and a small ceremony. The Excel change needs the Keeper of the Macro and thirty minutes of privacy.

As long as that arithmetic holds, the spreadsheet is the correct decision. No change-management poster in the lift lobby will out-argue it. People didn't reject your system out of stubbornness — they did the math you accidentally set up for them, and the math said column F.

Meanwhile, the licences from the last transformation sit exactly where the industry average says they sit: roughly half of purchased SaaS seats are never used (Source: Zylo SaaS Management Index). Shelfware: the only enterprise asset class with negative occupancy.

🤖 Great News, Everyone: We're About to Do It Again — With AI

And now, the sequel. Same plot, new fonts.

Right now, companies everywhere are buying AI exactly the way they bought ERP: with a big announcement, a bigger invoice, and a quiet unspoken agreement to keep doing everything in Excel. Early returns suggest the sequel is on script — MIT researchers found 95% of enterprise GenAI pilots deliver no measurable P&L impact (Source: MIT NANDA, via Forbes). The copilot goes in the front office. The close stays in the workbook. Eighteen months from now: two monoliths, one desk, same cake supplier.

Okay. One sincere paragraph now, because this is the bit that's actually new. Ahem.

Every previous generation of enterprise software needed the mess cleaned up before it could help. Harmonise the charts of accounts first, structure the data first, standardise the process first — then the system works. Agentic AI is the first technology that can be deployed inside the mess. An agent can read both unharmonised charts of accounts and propose the mapping, showing its working. It can read the warehouse WhatsApp photo, extract the number, and log where it came from. It can be told about the 2019 reclassification in one sentence — like a human, except it never goes on leave and never forgets why the reclassification exists.

Which flips the whole question. The highest-return AI project in most Malaysian enterprises is not a chatbot on the website. It's an agent pointed directly at the workbook the CFO has been quietly maintaining since 2019.

The RM48 million wasn't wasted, by the way. It bought a very solid 80%. The comedy lives in the other 20%.

So does the ROI.

Meanwhile, LHDN Would Like a Word

There's a distinctly Malaysian reason the joke is expiring.

LHDN's MyInvois e-invoicing mandate has been rolling out in phases by turnover, pulling transaction reporting toward near-real-time (Source: LHDN — e-Invoice). Read that again slowly: your invoice now reaches the tax authority faster than it reaches your own general ledger. Your invoice has a better relationship with LHDN than with your ERP. Add the amended PDPA and, for financial institutions, BNM's RMiT expectations (Sources: PDP Department; Bank Negara Malaysia), and the era of the artisanal, hand-crafted, single-origin month-end close is politely being shown the door.

Because regulators have started asking a very un-funny question: how was this number derived, by whom, from what source, and when did it change?

"Janet knows" is many wonderful things. It is not an audit trail.

A spreadsheet can't answer that question. A well-built agent — reasoning logged, sources cited, human approval gates — answers it by default, every time, without sighing.

💎 The Symprio Approach: We Interview the Spreadsheet First

Symprio builds AI products — for clients and in our own portfolio — and we start these engagements from the opposite end to a traditional implementation. Four principles, jokes free of charge.

We interview the spreadsheet first. Respectfully — it has seniority. Before designing anything, we map the shadow layer: every workbook, who its Keeper is, which exception it absorbs, and what breaks if that person resigns. That inventory is the most accurate description of how your business actually runs — the requirements document nobody thought to write.

We automate the exception, not the happy path. The happy path already has RM48 million of software on it; it's fine, leave it alone. Our agentic AI products and platforms → go after the 20% that never fitted — the reconciliations, the judgement calls, the WhatsApp accruals, the one-off adjustments that became permanent residents.

We measure before we build. Days to close, manual journals, hours per exception — baselined before a single line of code. Without a baseline, "faster" is a vibe. With one, it's a number your CFO can take to the board without doing improv.

We co-build, so the capability doesn't leave with our invoice. Our adopt-and-build model pairs Symprio engineers with your team through enterprise AI development →, transfers the architecture, and certifies your people to run it. Because if changes still cost RM40,000 and eleven weeks after we leave, we haven't fixed the trap — we've just rebuilt it in a newer language.

💬 Over to You

Quick self-assessment. No one is watching. Be honest:

Where does your month actually close — in the system you paid for, or in a file with FINAL in its name at least twice?
How many spreadsheets stand between your ERP and your board pack? (If you just thought "I'd have to check a spreadsheet to answer that" — yes, exactly.)
If the Keeper of the Macro resigned tomorrow, is your recovery plan a document or a prayer?

If any of that made you laugh in a slightly haunted way — you're in excellent company. Some of the biggest, best-implemented companies in the country are laughing right beside you.

📞 Let's Build Something Real

The cake was great. Now let's finish the last mile. It's time to build AI that works where the work actually happens — in the exceptions, with an audit trail, owned by your team.

Reach out to the Symprio team today — and yes, bring the workbook. We won't judge it. We want to meet it.

👉 Talk to the Symprio team → — 30 minutes, no slide deck, just whiteboard thinkin

👉 Explore our Agentic AI products & platforms →

❓ FAQ: Why Finance Still Uses Excel

Why does finance still use Excel after an ERP implementation?

Because implementations are scoped around the documented process, while month-end is mostly exceptions — unharmonised charts of accounts, manual accruals, audit-driven adjustments. Excel absorbs those instantly and free; a system change costs real money and takes weeks. The team isn't being stubborn. They're doing the arithmetic.

What is the enterprise software adoption gap?

The gap between a system being implemented and the business actually changing how it works. The software goes live, passes UAT, and gets a cake — while the real process keeps running through spreadsheets and workarounds, so the investment delivers a fraction of its business case.

Can AI actually fix the month-end close?

Agentic AI is the first enterprise technology that can work inside the exception layer instead of demanding it be eliminated first — reading unstructured inputs, proposing mappings with its reasoning shown, and escalating to a human when unsure. It works best pointed at the spreadsheets around the system, not sold as a replacement for the system.

Does Malaysia's e-invoicing mandate affect spreadsheet-based closes?

Yes. LHDN's MyInvois moves transaction reporting toward near-real-time, and the amended PDPA tightens data governance. An undocumented spreadsheet close stops being just an efficiency problem and becomes an evidence problem: you need to show how a number was derived, by whom, and when it changed. "Janet knows" is not an audit trail.

📚 Sources & Further Reading

#Symprio #EnterpriseAI #AgenticAI #Automation #DigitalTransformation #MyInvois #Malaysia

Symprio builds AI products that work where the work actually happens — which, statistically speaking, is somewhere in column F. Find us at symprio.com.