# What we’ll need to consider

> The responsibilities that arrive with generative making.

Chapter 4 of Design After AI · Notes · v0.1 · Updated 7 Oct 2026
Source: https://designafterai.com/what-well-need-to-consider

Speed changes what you are responsible for. When a team can make in an afternoon what used to take a quarter, the questions that used to be someone else's arrive at the designer's desk.

## Where every part of a design came from

_11 Sept 2026 · Prediction · near future · medium confidence_
Disciplines: product, architecture
Link: https://designafterai.com/what-well-need-to-consider#provenance

Which parts were generated, from what, and who checked them. Provenance becomes something clients, regulators and users ask for, and something the design process has to record as it goes rather than reconstruct afterwards.

## When the system acts on someone’s behalf

_18 Sept 2026 · Observed_
Disciplines: product
Link: https://designafterai.com/what-well-need-to-consider#acting-on-someones-behalf

Delegation needs designed boundaries: what an agent may do unasked, what needs a confirmation, and how anything can be undone. Consent stops being a checkbox and becomes an interaction that runs through the whole product.

## The pull towards the average

_25 Sept 2026 · Observed_
Disciplines: product, hardware, architecture
Link: https://designafterai.com/what-well-need-to-consider#the-pull-toward-average

Models trained on everything converge on the median. Left alone, the work drifts towards the most plausible answer, so distinctiveness has to be a deliberate decision rather than a by-product of effort.

## The cost of each generation

_1 Oct 2026 · Prediction · far future · low confidence_
Disciplines: engineering, architecture, hardware
Link: https://designafterai.com/what-well-need-to-consider#the-cost-of-generation

Every variation has an energy cost, and every rendered option that becomes a physical sample has a material one. Designing how much to generate becomes part of designing responsibly.

## How the system fails, in public

_6 Oct 2026 · Observed_
Disciplines: product, engineering
Link: https://designafterai.com/what-well-need-to-consider#failure-in-public

Generative products are wrong in new and surprising ways. Designing the failure (what the person sees, how they recover, how they report it) matters as much as designing the success.

Have your say: reply to any idea on X. The best replies are added beside the idea they answer.

## Version history

- v0.1 · 7 Oct 2026 · First notes gathered. (current: https://designafterai.com/what-well-need-to-consider.md)