---
title: "Our own systems instead of spreadsheets: why building pays off now"
description: "With AI, a custom system that used to cost licenses and years now pays off. How SUMAX is moving off spreadsheets and an old CRM, and what that decision costs."
url: https://carl-wuerfel.de/en/notes/custom-systems-instead-of-spreadsheets
author: Carl-Uwe Würfel
---

1. [Home](https://carl-wuerfel.de/en)
  2. [Notes](https://carl-wuerfel.de/en/notes)
  3. AI infrastructure

# Our own systems instead of spreadsheets: why building pays off now

With AI, a custom system that used to cost licenses and years now pays off. How SUMAX is moving off spreadsheets and an old CRM, and what that decision costs.

[Carl-Uwe Würfel](https://carl-wuerfel.de/en/about)First published November 18, 2026 · 3 min read[Deutsche Fassung](https://carl-wuerfel.de/notizen/eigene-systeme-statt-excel-listen)

## Contents

  1. Before: spreadsheets, an old CRM and knowledge in people's heads.
  2. With AI, the math flips.
  3. The decision: sales without the old CRM.
  4. We're not all the way out.
  5. What the decision costs.
  6. Further reading

For a mid-sized company, a custom system for clients, contracts and orders was too expensive for a long time. It would have cost licenses and years, so the spreadsheets stayed. AI flipped that math. At SUMAX we now build systems with the team that fit how we work, without endless licenses and without years of lead time. Since July 2026 our sales team no longer works with the old CRM. We're not done, and the decision has a price: if you build it, you maintain it.

Chapter 1 of 6

## Before: spreadsheets, an old CRM and knowledge in people's heads.

For years, a lot of our work lived in Excel spreadsheets. On top of that we had an old CRM and plenty of knowledge in individual people's heads. The overview of which services were running and who was working on what was spread across spreadsheets and heads.

That didn't always go well. More than once a cancellation or a renewal wasn't entered, and nobody in sales had it on their radar. Data was scattered, a lot of it was maintained twice and hard to trace. Check how many places the same piece of information is maintained in. That's where the errors come from.

Chapter 2 of 6

## With AI, the math flips.

I had wanted a better system for a long time. Buying one would have meant licenses and bending our workflows to the software. Building one would have taken years. Neither made sense for an agency our size.

With AI, building got cheap. Applications on top of existing AI models now take a fraction of the time, and they fit the way we work. Only the build got cheaper, though; running a system costs as much as before. The general trade-off is in [Build vs. buy AI](https://carl-wuerfel.de/en/notes/buy-or-build-ai). So a build only adds up once you count running costs over several years.

Chapter 3 of 6

## The decision: sales without the old CRM.

Clients and contracts sit in one dataset that our tools share.

_Figure: Before and today at SUMAX. Clients and contracts: before in spreadsheets, an old CRM and people 's heads, today in one shared dataset. Inquiries: before in several places, today in one inbox list. Vacation and sick days: before entered by hand, today nobody enters them anymore. Gaps in billing: before someone happened to notice, today a warning comes up first. A few spreadsheets are still running._

In July 2026 I told the team that sales would stop using the old CRM and the spreadsheets, effective immediately. Instead we're building our own system that fits our workflows, together with the people who use it.

Some things are gone today. Nobody enters vacation and sick days by hand anymore, and nobody routinely checks working hours manually. When we win a deal, the system creates the invoice draft itself. Inquiries land in one inbox list instead of in several places. If something is missing from billing, a warning comes up before someone happens to notice. Clients and contracts sit in one dataset that our tools share. Start with the data that several workflows need.

Chapter 4 of 6

## We're not all the way out.

A few spreadsheets are still running. Not every workflow is worth rebuilding, and some spreadsheets do their job. What mattered was getting clients and contracts out of spreadsheets, because cancellations, renewals and invoices depend on them. Clear out the spreadsheets whose errors cost money first.

Chapter 5 of 6

## What the decision costs.

If you build it, you maintain it. Every tool needs someone responsible for it, two years from now as well. It only pays off if someone knows the business well enough to know what the system has to do. No license comes with that. If that knowledge isn't in-house, buying is the better choice. Before your first custom system, settle who will look after it two years from now.

Chapter 6 of 6

## Further reading

  * [Build vs. buy AI: only the build got cheaper](https://carl-wuerfel.de/en/notes/buy-or-build-ai)
  * [Single source of truth](https://carl-wuerfel.de/en/glossary/single-source-of-truth)
  * [AI knowledge management](https://carl-wuerfel.de/en/ai-knowledge-management)

## Read next.

  * [Note**Using AI too little, trusting it too much: what the GoTo study shows**](https://carl-wuerfel.de/en/notes/using-ai-too-little-trusting-it-too-much)
  * [ Note**AI investment: buying is faster than changing how work gets done**](https://carl-wuerfel.de/en/notes/ceos-invest-in-ai-faster-than-they-change-work)
  * [ Note**AI answers by language: the market in the query decides**](https://carl-wuerfel.de/en/notes/ai-visibility-by-language-market)

**Carl-Uwe Würfel** COO & Chief AI Officer, SUMAX

[Questions about this piece](https://carl-wuerfel.de/en/contact)
