Side by side
The Scrum column separates the official Guide from trainer guidance. AI-Native SAFe is compared with its current public guidance, including team-level flow and risk controls. These are different combinations of practices, not claims that one method owns an idea.
On a narrow screen, scroll the table sideways to compare all three.
| Design choice | Scrum, with AI | AI-Native SAFe | The Kite Method |
|---|---|---|---|
| Built for | Teams typically of ten or fewer; the Guide also addresses organisations. [1] | Organisations coordinating teams through ARTs. AI-Native Teams typically have three to seven members. [4] | Teams of 2 to 5 by default, where agents do most of the building, from one team to a whole organisation. Team sizes |
| How it scales | Teams on one product share its goal, backlog, Product Owner and Definition of Done; the Guide leaves wider coordination open. [1] | ART coordination, with governance at portfolio level. [5] [7] | By adding small teams that share architecture rules and verification. Roles and teams |
| Starting point | The official 2020 Guide, with separate AI guidance from trainers. [2] [3] | A new operating model introduced in 2026 alongside Core SAFe. [8] | Written for AI-assisted delivery from the first page. |
| Rhythm | Sprints of a month or less. Delivery can happen before a Sprint ends. [1] | Continuous team flow, with ART-level Planning Intervals, PI Outcome Planning and Sense and Respond each iteration. [4] [6] | No delivery time box. Every item runs its own loop: Frame, Fly, Check, Learn. |
| How human control is set | Three accountabilities and product quality standards; teams choose working techniques. [1] | Risk management, automated rules and detection, and human oversight, supported by portfolio governance. [7] | By risk, per item: reversibility and impact route it to Free, Guided or Held. |
| Craft | Knowledge loss is a risk named in trainer guidance. [3] | A human-centric culture, critical thinking and human review of outputs. [4] | Built into the rules: explain-back, hand flight and a written Human Grip. |
| Size of the method | A short Guide, with separate companion publications. The 2025 Expansion Pack has a January 2026 successor, Scrum Guide Expanded. [9] | An enterprise operating model with framework guidance, training and certification. [8] | Seven method and comparison pages, free to read: six core chapters plus this comparison. |
Six design choices
1. Control per item, by risk
Scrum defines accountabilities and quality commitments. SAFe adds risk governance, automated checks and human oversight. [1] [7] Kite makes one particular choice explicit: every item is routed by reversibility and impact. A dependency update and a pricing change do not get the same treatment, and neither is decided by habit. See Flight zones.
2. A loop per item, not a time box
Scrum permits release within a Sprint, and AI-Native SAFe uses continuous team flow alongside ART cadence. [1] [4] Kite removes the delivery time box altogether. Each item runs its own loop, and the plan is a queue of Kite Briefs in priority order. See The Kite Loop.
3. Team size follows review capacity
AI supplies the build capacity, so a team does not need to grow for volume. What runs out is attention: how much one person can check and still understand. That is why the default Kite team has 2 to 5 people, and why you split a team before you stretch it. See Roles and teams.
4. Verification is the throttle
In Kite, automated verification does more than catch errors. It decides how fast you may go. When tests, evals and checks cover the risk, an item can move one zone lighter. Investing in the Tail is the shortest route to more AI speed. See The method.
5. Craft is a rule, not a hope
When agents do more of the building, juniors can lose practice and seniors can lose touch with their codebase. Kite turns that risk into rules. You approve only what you can explain. Fixed blocks of work stay hands-on. And the responsibilities that stay human are written down, so they do not quietly slip away. See Human Grip.
6. Scale by adding teams, not layers
When there is more work, Kite adds a team instead of enlarging one. Every team keeps its own roles, loop and zones. A Frame Architect guards shared architecture rules. A shared Tail provides test infrastructure and evals. The Experience Owner keeps design principles consistent across teams. A joint Kite Retro each quarter keeps zone routing consistent. Teams stay small; the organisation can be any size. See Roles and teams and Getting started.
Which one fits
Our practical recommendation depends on how you work today:
Stay with SAFe
When your organisation already runs on it and depends on its portfolio planning and funding. Pilot Kite team by team while preserving the coordination and governance you need.
Stay with Scrum
When your team runs it well, agents are still a small part of the work, and you would rather evolve your events than replace them.
Choose Kite
When agents do most of the building and reviews pile up, rework rises, and people can no longer explain what they ship. That holds for one team and for an organisation with many.
Questions we get
- Is Kite agile?
- Yes. It keeps short feedback loops, working software and people over process. It uses a loop per item instead of a delivery time box.
- Can we keep our sprints?
- During First flight, yes. Keep the sprint as a reporting rhythm while each item runs its own loop. Drop it when your forecasts from intent-to-value hold up. This is Kite's transition approach.
- Does Kite work for a whole organisation?
- Yes. Start with one pilot team and add teams as each gate is passed. Teams stay small; the organisation can be any size. At Full altitude you have your own Line Keepers and Kite coaches.
- We run SAFe. Can one team fly Kite?
- Yes, as a team-level pilot agreed with the train. Keep the train's commitments, coordination and governance. Meet them from the Brief queue while using zones and the Kite Loop inside the team. This is Kite's proposed integration, not an official SAFe adoption path.
- Why not just adapt Scrum?
- You can, and many teams will. Kite is for teams that want review capacity, risk routing and a loop per item to define their way of working.
- Coming from Scrum?
- See the translation table on The Kite Loop.
Sources
Checked on against primary sources. The comparison uses publicly accessible guidance; some detailed SAFe articles require a login. Trainer publications and the Expansion Pack are supplementary works, not a replacement for the official Scrum Guide.
- The Scrum Guide (2020) — team size, accountabilities, Sprint, releases and shared product commitments.
- Official Scrum Guide downloads — November 2020 remains the current official version.
- The Future of AI-Powered Product Development, Robbin Schuurman, Scrum.org, 30 April 2026 — a series drawing on trainer discussions about value, accountability and knowledge loss.
- AI-Native Teams, Scaled Agile — team size, continuous flow and human judgment.
- AI-Native Teams, Roles, and ARTs, Scaled Agile, 21 July 2026 — coordination across teams.
- PI Outcome Planning and Sense and Respond, Scaled Agile — the ART planning and learning cadence.
- AI Governance and Ethics, Scaled Agile — risk management and automated and human oversight.
- New Release: AI-Native SAFe, Scaled Agile, 23 June 2026 — the new operating model and continued Core SAFe support.
- Scrum Guide Expanded: publication history — the June 2025 Expansion Pack and its January 2026 successor.
SAFe is a registered trademark of Scaled Agile, Inc. The Kite Method is not affiliated with Scaled Agile, Scrum.org or the authors of the Scrum Guide.