The 4D Framework
An essay on choosing the right tools in the AI era.
A structured way to decide when to build, what to build with, and when to leave things alone.
If you only have a minute.
Most wrong tool decisions are actually wrong build decisions. Three gates filter them out before any tool gets picked.
Once a build is justified, four dimensions — Repetition, Complexity, Risk, Builder Skill — map the task to the right level of tooling.
A break-even formula tells you whether the build pays itself off — including the cost of AI errors. Below it, you are subtracting value.
The framework runs in two modes. Strategic for planning, Reality for the deadline. Most decisions are made in the second mode — that is where it has to work.
The Silent Regression
A process that looks modernised but performs worse than the version it replaced. The errors arrive faster, with more authority, and with less of the hesitation that flags an unusual case. If nobody measures, the decline is invisible — until a customer notices.
Three gates. Most "wrong tool" problems die here.
Most of what looks like tool-selection failure is actually a build decision that should never have happened. Three simple checks catch them — before any tool gets picked.
Is the solution already out there as off-the-shelf software or a built-in AI feature in your core platform?
Will your primary platform ship this feature within six months? Check roadmaps, betas, hackathon demos.
Does the expected value exceed the build cost — in both money and opportunity cost?
Four dimensions map the task to the right level of tooling.
Once a build is justified, these four questions determine whether you reach for Google Sheets, Make, Claude Code, or something in between.
How often does this task run? A one-off is a 1. Hourly is a 5.
Example: 'Draft this one press release' → 1. 'Daily lead enrichment' → 4.
How structured and stable is the work? Copy-paste is 1. Multi-system edge-cases are 5.
Example: 'Compute column total' → 1. 'Classify free-form customer feedback' → 4.
What happens when it fails? Internal sandbox is 1. Customer-facing or compliance is 5.
Example: 'Internal dashboard' → 2. 'Automated contract signing' → 5.
Who's actually going to build this? Business user is 1. Senior developer is 5.
Colour-coded on the cube. The best tool is the one the available builder can use.
One formula separates tools that compound from tools that become debt.
Most teams decide by gut feeling. This framework gives you a number. Below it, building is subtraction. Above it, compounding value.
- 20 customer-forms a month × 15 min each
- Build: 4 h × 80 €/h = 320 €
- Save: 0.25 h × 30 €/h = 7.50 €/form
- Break-even at ~46 uses → 2.3 months
- Lifetime: 2 years → several thousand € saved
- Hackathon app: HubSpot account summariser
- Break-even math said: pays off in 1 week
- Reality: HubSpot shipped the same feature natively 3 weeks later
- The math was right. Gate 2 (“Coming?”) was skipped.
- Break-even economics are necessary, not sufficient.
The essay — sixteen pages, one afternoon.
Everything above, in depth. Worked examples, counter-examples, and the empirical backbone from Harvard's Jagged Technological Frontier, McKinsey's Agents, Robots, and Us, and Deloitte's 2026 Human Capital Trends.
16 pages · editorial prose · no jargon · free, no email required
- ·The silent regression — why AI can make things worse
- ·Three gates that catch bad build decisions early
- ·Four dimensions + the break-even math you skip
- ·Reality Mode vs Strategic Mode — real-world decisions
Why this exists.
I kept explaining this in meetings.
Every week a new AI tool promises to change everything. Every month a dozen quietly disappear. Teams keep picking the wrong ones — and most of the time it's not the tool that was wrong, it's the decision to build in the first place.
So I wrote down the decision process I was repeating in my own head. Three gates. Four dimensions. One formula. If it helps someone avoid a pointless Lovable build or a premature Make-scenario — good. If it starts an argument — even better.
Senior Automation Manager focused on AI enablement and process automation. Background in revenue operations and the service industry. Based in Baden-Baden.