1) Same SKU, four model styles
Scenario: Ceramic mug 4.2×3.6×4.0 in, 14 oz, ships parcel, current pack 10×8×6 with bubble, damage ~1.8%.
Ask each model for: three outers, cushioning, DIM posture, test plan, pack time.
| Lens | What to score |
|---|---|
| Measurement obedience | Did it use your inches? |
| Carrier humility | Did it admit divisor/rate uncertainty? |
| Ops realism | Pack time & substitutions |
| Risk honesty | Corner crush, glaze chipping, void migration |
| Actionability | Can a lead run a test tomorrow? |
How to run monthly: keep a frozen prompt + frozen SKU card; only models change. Log disagreements — that is the curriculum.
2) Chat cartonization vs WMS cartonization
| Chat AI | WMS/TMS engine | |
|---|---|---|
| Best at | Designing rules, exception packs | Assigning boxes to live orders |
| Data need | You paste | Integrated order/SKU data |
| Risk | Elegant wrongness | Opaque defaults |
| Use when | < few hundred orders/day design work | Mixed-SKU volume & labor targets |
Verdict: use chat to author policy; use engines to execute policy.
3) Stock packaging vs custom print
| Factor | Stock + label | Custom print |
|---|---|---|
| Time to first ship | Days | Weeks+ |
| DIM control | Excellent if sized well | Only if size locked first |
| Brand lift | Moderate | High when earned |
| Obsolescence risk | Low | High if SKU dims change |
| When to choose | Unstable sizes, testing | Stable A-movers with ASP support |
4) Paper void vs air pillows vs molded pulp
| Method | Strength | Weakness | AI help |
|---|---|---|---|
| Kraft paper | Simple recycling story | Labor / overfill temptation | Draft pack density rules |
| Air pillows | Fast, light | Plastic perception, deflation | Specify film type & when banned |
| Molded pulp | Product-shaped protection | Tooling / storage | Decide when volume justifies |
5) Poly mailer vs paper mailer vs light shipper
| Signal | Lean poly | Lean paper mailer | Lean shipper |
|---|---|---|---|
| Soft apparel | Often | If plastic-free mandate | Rare |
| Scratch-sensitive hard good | Risky | Risky | Better |
| Rainy last-mile | Strong | Check coatings | Strong |
| Perceived premium | Low | Medium | High if printed |
AI should force you to state the mandate (plastic-free vs lowest damage vs lowest DIM) before picking.
6) In-house pack vs 3PL pack cards
In-house: optimize seconds and tribal tricks.
3PL: optimize clarity, allowed substitutions, audit photos.
If your AI pack card assumes a special tape gun only your garage has, it will fail at the 3PL.