The five standard grip points
Human Grip is the set of responsibilities where human judgment is what creates the value. A team names them, writes them down and keeps a hand on them, so they never quietly slip away.
- Customer contact and problem discovery
- People talk to users and decide which problem is worth solving. UX researchers work with the Intent Owner on this discovery. AI prepares the conversation and digests what was said.
- Irreversible decisions
- People decide on architecture, data models, public APIs and pricing. AI researches the options and advises.
- The release decision
- A named person weighs the risks and signs off each release. AI prepares the release summary and risk list.
- Ethics, privacy and security trade-offs
- People weigh ethics, privacy and security trade-offs. AI flags risks and supplies context.
- Mentoring and team rituals
- People mentor each other, lead retros and hold one-on-ones. They share knowledge person to person.
Each team adds its own grip points to this list, for example a domain where trust weighs extra heavily.
Optional grip point: Experience judgment
Whether it feels right for the user is decided by a person. AI generates variants and checks consistency.
Add this as a separate grip point when your team needs to name and review that responsibility explicitly.
Write down your Human Grip
Use this template for each grip point. The release decision is filled in as an example. Replace the owner with a person's name for each release.
The Line Keeper maintains the list. At every Kite Retro, the team checks the owners, AI's role and next review dates.
| Grip point | Why human | Owner | AI's role | Next review |
|---|---|---|---|---|
| The release decision | Accountability cannot be delegated | Named person per release | Prepares the release summary and risk list | Kite Retro |
Protecting craft
Anyone judging AI work must keep mastering the craft themselves. Four agreements make that happen.
Explain-back
Nobody approves something they cannot explain.
Hand flight
A fixed weekly block in which Engineers build, debug or design without AI, so their own judgment stays sharp.
Learning path for juniors
Juniors grow in three steps towards judging agent work on their own. See the learning path below.
Tail rotation
Everyone works on verification from time to time, because that is where you learn most about how the system really works.
The learning path for juniors
AI takes over much of the practice work juniors used to learn the trade from. The learning path gives that learning a fixed place, in three steps.
Shadowing
The junior works alongside a senior on Guided items. The senior steers the agents; at every step the junior explains what the agent did and whether it is right.
Steering under supervision
The junior writes Kite Briefs and steers agents on Free and Guided work. The senior mainly judges the junior's review: what did they spot, and what did they miss?
Checking independently
The junior becomes a reviewer in the Guided zone once explain-back succeeds consistently, their items come back no more often than average, and they find deliberately planted defects in a practice PR.
The Held zone stays with experienced people. Juniors do join those discussions, so they learn those trade-offs too.