05
Weight what matters, then let the scores fall out
Each option gets a 1–5 score on ten criteria, once for 2022 and once for today. You set how much each criterion matters for the decision in front of you. Pick a preset, adjust the weights, and switch eras to see the ranking change. Your weights are saved in the decision console at the bottom of every page.
| Criterion | Weight | Build | Buy | LLM |
|---|---|---|---|---|
Time to market How fast users get value ✦ Coding agents move Build from quarters to weeks. LLM prototypes ship the same day. | 3 | |||
Low upfront cost Cash spent before it works ✦ Engineering hours get cheaper per feature. Prompting costs almost nothing up front. | 3 | |||
Low cost at scale Unit cost at 100× volume ✦ Token prices keep falling. Seat and transaction fees don't. | 3 | |||
Fit & control Matches your process exactly ✦ Vendors add AI configurability, and LLMs follow instructions much better. | 3 | |||
Differentiation Competitors can't copy it ✦ Unchanged per option, but it matters more now. Raise its weight. | 3 | |||
Reliability & determinism Same input → same output ✦ Structured outputs and evals close part of the gap, not all of it. | 3 | |||
Low maintenance burden Ongoing work after launch ✦ Agents take over upgrades and migrations. Models improve without your help. | 3 | |||
Low lock-in Easy to switch later ✦ Models are more interchangeable now, and agents write migration code. | 3 | |||
Handles messy input Text, docs, images, edge cases ✦ Everyone can embed a model now, but LLM-native designs still handle it best. | 3 | |||
Compliance & audit Explain every decision ✦ Traces help a little. Regulators still want deterministic paths. | 3 |
Green ring = gained a point compared with the other era. Red ring = lost one.
Weighted score
Buildleads82+16 vs 2022
Buy72+8 vs 2022
LLM72+28 vs 2022
Outline = same weights in the other era. 100 = perfect on every weighted criterion.
Reading it
Every criterion equal. Build leads by 10 points.
- why it mattersTry Balanced and switch eras. In 2022 Build and Buy were neck and neck, with LLM more than 20 points behind. In 2026 LLM gains the most and draws level with Buy.
- why it mattersTry Regulated fintech. Determinism and audit still favour Build and Buy, and AI hasn't changed that. It does make Build much cheaper to reach.
- why it mattersDifferentiation scores the same in both eras. AI changes how much weight you should give it: when everything else is cheap, differentiation is what's left to compete on.