In short
The Kite Method is a way of working for software teams that combines the speed of AI with human direction. At its core: humans decide what’s worth building and whether it’s right. AI builds it, fast.
The method gives teams three things:
- The four flight zones: a way to route every work item to the right level of human involvement.
- The Kite Loop: a work rhythm that replaces the sprint.
- Human Anchors: a fixed set of responsibilities that deliberately stay human.
It is built for software companies and product teams that already use AI agents or want to scale them up. Many find that speed without direction leads to rework, review fatigue, or a loss of craft. The promise is to ship faster without the team losing control, judgment, or the joy of the work.
Why now
Writing code is no longer the bottleneck in software development. The bottleneck has moved to two places: knowing exactly what you want, and being able to establish that what you got is right.
Existing methods such as Scrum and SAFe assume that human build capacity is the scarce resource. Sprint planning, story points and velocity therefore measure something that is no longer scarce. They do not steer on what is scarce now: clear intent, reliable verification and human attention.
Teams that add AI without changing how they work run into three problems:
- Review pressure explodes. Pull requests pile up and people start approving without really looking.
- Rework rises. Agents build at lightning speed, but given a vague brief they build the wrong thing at lightning speed too.
- Craft erodes. Juniors stop learning the trade and seniors lose touch with their own codebase.
The opposite extreme does not work either. Teams that throttle AI back to autocomplete get overtaken by competitors who dare to go further.
The Kite Method sits in between. Where it can, AI gets full speed. Where it must, a human holds the line.
The core idea: line and wind
A kite climbs because two forces work together: the wind that lifts it and the line that keeps it on course. The line is not a brake: it is the tension on the line that makes a kite climb. In The Kite Method, AI is the wind and the human is the line.
Five principles make this concrete:
- Intent before executionNo agent starts without a Frame written by a human: what, why, and when is it good enough.
- Control follows risk, not habitHow much human oversight a task gets depends on how reversible it is and how many people it affects. How we always did it does not count.
- Judge outcomes, not linesHumans test behaviour against acceptance criteria. Line-by-line review happens only in the zones that call for it.
- Every failure becomes a ruleWhatever goes wrong lands as a test, a guardrail or extra context for the agents, so the same mistake does not come back.
- Accountability cannot be delegatedEvery release has a named human behind it, even when AI did the work.
The four parts of the kite
Each part of a kite stands for a part of your way of working, with a fixed owner. A team that neglects one part notices immediately: without a Frame you build the wrong thing fast, without a Tail quality wobbles.
| Part | On the kite | In your team | Owner | Concrete artifact |
|---|---|---|---|---|
| Frame | The spars that give it shape | Intent, boundaries, acceptance criteria, architecture rules | Human | Kite Brief per work item, architecture principles |
| Sail | Catches the wind | AI execution: code, tests, documentation, variants | AI agents, steered by a human | Agent runs, pull requests, Flight Log |
| Tail | Keeps it stable | Automated verification: tests, evals, checks, monitoring | Human designs, AI maintains | Test suites, evals, CI gates |
| Line | Connects it to the ground | Human judgment, customer contact, final accountability | Human | Release decision, Human Anchor list |
The Kite Brief
The Kite Brief is the Frame's most important artifact. Every work item gets one before it takes off. A good Brief fits on one screen and contains five fields: the goal, when it is done, what is out of scope, the flight zone and the risks.
- Goal
- Which problem we solve, and for whom. Finance users can export invoices to CSV, so month-end reconciliation no longer means copying rows by hand.
- Done when
- Testable acceptance criteria. Every invoice in the chosen period is in the export. Totals match the ledger to the cent.
- Out of scope
- What the agent must not touch. The invoice data model, PDF generation, permissions.
- Zone
- Free, Guided, Paired or Held. Guided: customers see the result.
- Risks
- What could go wrong and how you would notice. Wrong totals reach finance. A reconciliation test against ledger fixtures catches it.
The Tail sets how high you can fly
The stronger the automated verification, the more work can move to a lighter zone. Investing in the Tail is therefore the fastest route to more AI speed.