Platform

An operations panel instead of Excel: when the sheet lies and a system may cost real money

Author: Marcin Bochenek · Updated: 2026-08-24

Excel is not the enemy. The enemy is Excel pretending to be ERP, CRM, calendar and till at once while five people sit in the file and nobody knows which copy is true. An operations panel I ship is the Platform package — from PLN 25,000, because it is not a prettier table in the browser. Below: thresholds where I say keep the sheet — and where the sheet already steals hours.

Signs Excel stopped being a tool

• Two “final” files, one in mail, one on disk.
• Someone is afraid to filter because formulas will break.
• The same fact typed in three places (calendar, invoice, WhatsApp).
• A new hire learns the process from a person, not from the system.
• Your report is built by hand on Sunday evening.

If you recognise two points, we talk. If one — often a tidier sheet or an off-the-shelf tool is enough. I do not sell a panel out of habit. Audit Sprint (from PLN 1,500) exists so we do not code regret for 80k.

What a panel is, and what is just another SaaS with a card

An operations panel is an app around your flow: roles, statuses, one event log, integrations you already have (invoices, locks, calendar). Off-the-shelf SaaS is cheaper when your process is a market standard. At Mint I did not write a PMS — I used Previo. Plumm is a product because sole-trader accounting + e-invoices + an assistant is our process, not a marketplace template.

A price a company website cannot impersonate

Platform: from PLN 25,000. In that range: a panel or a team app, connections to payments / calendars / e-invoices, automation with limits and a log, staging, training, post-launch measurement. That is not a Start brochure at PLN 2,000. Someone who “drops Excel into Airtable over a weekend” ships a different product — often without roles, audit trail or escalation.

Before code I name before/after metrics: hours on the report, error count, time from event to invoice. If we cannot name them, I do not start. A pretty UI with no result at this budget wastes both sides.

People, not “digital transformation”

1. A test version before the team loses the old file.
2. 1–2 hours of training, not a PDF on “how to click”.
3. Code and config after invoices — yours.
4. NDA. No theatre of fake testimonials from a panel rollout.

If the team sabotages the tool, the problem is not React. It is an unwritten process. A panel will not fix that. Map it in the audit first.

Neighbouring decisions

Off-the-shelf vs custom: a separate article. Accounting instead of Excel: Plumm as my own product, not “I rolled it out to twenty offices”. I add HITL when repetitive steps can be limited to an allow-list — not when you want “ChatGPT inside Excel”.

Symptoms that the sheet already manages you

An operations panel instead of Excel does not start with “we want to look modern”. It starts when two people overwrite the same row, a macro dies on one laptop, and “final_v7_really” lives in email. When a process repeats every week and touches money or occupancy, the cost of a mistake stops being “we will fix it by hand”. Then we talk about a system, not a prettier table.

I leave Excel when one person handles low volume, there are no parallel edits, and there is no audit trail. An awkward truth: many firms do not need a panel. They need file-name discipline and one person who does not leave without a deputy. The Platform package at PLN 25,000+ makes sense when that cost is cheaper than monthly chaos, or when Excel blocks growth you no longer pretend not to have.

• Several roles on one file: front desk, accounting, owner.
• A change history you cannot reconstruct from “someone deleted a row”.
• The same process in three locations or three properties.
• An integration Excel will not hold: KSeF, PMS, payments, a mail queue.

What is in scope, and what is “Excel+AI” fantasy

I build states, roles, step logs and allow-lists. I do not build “ChatGPT in a sheet that books the accounts by itself”. I add HITL when a repetitive step can be limited to an allow-list. If you want a model to guess VAT from a PDF and send it immediately, that is not a panel — that is risk. Online accounting in Plumm is my own product, not a claim that I rolled it out to twenty offices.

SaaS versus custom is a separate call. If the market has a tool covering 80% of the process, I take it — like Previo for Mint bookings. If the process is your edge, like e-invoices and the assistant in Plumm, I build. Off-the-shelf vs a panel is not a religious slide. It is lock-in arithmetic and the cost of change in two years.

Shipping a panel: data, training, the day Excel dies

Migration is not “we will paste CSV on Friday”. It is field mapping, junk in old columns, a decision about history. A 1–2 h training does not replace a process owner. If after launch everyone returns to the sheet “because it is faster”, the panel was theatre. So I write down which day Excel is switched off and who may open an exception.

I do not guarantee the team will love the UI. I do guarantee I do not leave you with a server password and a PDF. After invoices the code is yours, or the product is my SaaS — two different contracts. Mixing them is the usual source of grievance a year later.

When I decline a panel project

1. There is no process owner — only “a team that will gather requirements”.
2. The budget is “an MVP for a few thousand, like a template app”.
3. The goal is to replace people with a language model, no log, no escalation.
4. Excel hurts, but nobody can name one metric that should fall.

Then an Audit Sprint or a stop is more honest. An operations panel instead of Excel is expensive because a bad state is expensive. If you do not feel that price, keep the sheet and hire someone to guard it. That is also a decision. I do not sell shame for not being “digital” yet.

The day the sheet dies must be in the contract

Without a date Excel does not die. The team returns “for a moment”, a macro lives on one laptop, the panel becomes a museum. A 1–2 h training does not replace an owner. A Friday CSV paste does not replace mapping junk in old columns. Ops at PLN 60–180k makes sense when that cost is cheaper than monthly chaos or when the sheet blocks growth. If you cannot name a metric that should fall — we do not start. Audit Sprint or a stop. I do not sell modernity theatre. I sell states, roles and a log, or I tell you to keep the file with a named guardian.