Their Finance Team Was Living in 47 Spreadsheets. We Fixed That in Weeks.
"Not a finance system: a single point of failure with a salary attached."
Table of Contents
- The Call That Started Everything
- Step 1: Mapping the Bleeding Before the Bandage
- Step 2: Killing the Spreadsheet Graveyard
- Step 3: A Dashboard That Tells You What is About to Go Wrong
- Step 4: Connecting Finance to the Business
- Step 5: Automating HR and Recruitment
- Step 6: Making Meetings Stop Disappearing
- The Numbers: What Changed
- The Honest Part: What Was Actually Hard
- The Bigger Picture
1. The Call That Started Everything
The finance manager had a system. Forty-seven Excel files. Colour-coded tabs. Formulas that only she understood. A payroll process that required cross-referencing five documents simultaneously and took two full days every month: and still produced errors that took another day to find.
She wasn't bad at her job. She was exceptional at her job. She had to be, just to keep the whole thing from falling apart.
When she went on holiday, the business held its breath.
When this UK-based organisation reached out to Gezora.ai, they described their situation the way most businesses do: carefully, professionally, understating the chaos by about 60%. What they actually had was manual payroll built in Excel, credit control tracked across disconnected sheets, and zero real-time visibility.
2. Step 1: Mapping the Bleeding Before the Bandage
The first thing we did was nothing. No tools deployed. No automations built. Just listening. We mapped every manual process across finance, HR, recruitment, and operations.
What we found was predictable and brutal:
- Payroll data lived in Excel but had to be manually checked against timesheets, overtime logs, and absence records.
- Overdue invoices were tracked in a sheet that got updated when someone remembered.
- HR documents were being drafted from scratch every single time.
- Onboarding had no automation: new starters got sent a PDF and wished good luck.
3. Step 2: Killing the Spreadsheet Graveyard
Forty-seven files became one system. We built a unified payroll and finance automation layer that connected the data sources that had previously been living separate lives.
Errors that used to take a day to find now surface in seconds. The system knows what the numbers should look like and highlights anything that deviates. Duplicate records: gone. Overtime miscalculations: gone.
The payroll process that took two days every month? Now takes under two hours.
4. Step 3: A Dashboard That Tells You What is About to Go Wrong
The question you actually want answered isn't "what's overdue?": it's "what's about to be overdue, and who needs to hear from us today?". We built a real-time credit control dashboard with three lanes:
- On Watch (30 days): Automated reminder sent. No human action needed yet.
- Escalation (60 days): Personalised follow-up drafted and queued for human approval.
- Critical (90+ days): Flagged to finance manager with full history surfaced.
Within the first month, outstanding receivables dropped by 34%.
5. Step 4: Connecting Finance to the Business
Automation's biggest gains come from connecting systems that were never talking to each other. Payroll now talked to HR. When a new starter was added to the HR system, the payroll record was created automatically: no separate data entry, no risk of the finance team not being told.
6. Step 5: Automating HR and Recruitment
HR document creation had been a blank-page problem. Every policy, risk assessment, and disciplinary letter started from scratch. We built an AI-powered document generation system that created consistent, ACAS-compliant documents from structured inputs.
Recruitment tracking moved out of email threads and into a structured pipeline. CV screening against role requirements and candidate communication was drafted and queued: never needing to be written from scratch.
7. Step 6: Making Meetings Stop Disappearing
We integrated AI transcription and summarisation into their Zoom workflow. Every recorded meeting automatically produced a summary, a decision log, and a flagged action list: attributed to the right people and stored in the right place.
The number of times "I didn't know that was my action" happened in the two months after deployment: zero.
8. The Numbers: What Changed
Three months post-deployment, here's what the data showed:
| Area | Before | After |
|---|---|---|
| Monthly payroll processing | 2 days | Under 2 hours |
| Overdue receivables | Baseline | Down 34% |
| HR document drafting | 2-3 hours | 15-20 minutes |
| Recruitment admin per hire | ~18 hours | ~6 hours |
| Payroll error discovery | End of pay run | Real-time flagging |
| Onboarding completion | ~60% | 97% |
9. The Honest Part: What Was Actually Hard
Migrating historical data from 47 spreadsheets with inconsistent formatting is exactly as unpleasant as it sounds. We spent significant time on data cleaning before a single automation could go live.
Change management was real work. Bringing people along: showing them that the new system made them more valuable, not less: was as important as the technical build.
10. The Bigger Picture
Gezora.ai builds systems that remove the drag. Not by replacing your people: by removing the parts of their jobs that were never worthy of their abilities in the first place.
The finance manager who ran 47 spreadsheets now runs a business intelligence dashboard that updates itself. She spent last Tuesday on a strategic project she'd been pushing back for eight months.
That's what automation is actually for.
Ready to find out what's compounding in your business? Let's talk.