The AI+Scrum Supplementary Guide

— Universal Empowerment Edition · Applicable to All Product Forms

Foreword: The Core Positioning of This Guide

This guide serves as an AI-integrated enhancement and supplement to the 2020 Official Scrum Guide. It does not replace, subvert, or alter the native Scrum framework. This guide is not limited to AI-native products, nor does it exclude standard software/hardware, purely hardware, purely software, or physical products without AI modules. It applies to all: Scrum teams utilizing AI tools and capabilities to empower the full lifecycle of product development.

Core Theses:

1. The essence of Scrum is an empiricism-based, experiment-driven, and self-adaptive framework for complex problems. It focuses on the value closed-loop of “Vision – Iteration – Experimentation – Feedback – Adaptation,” which is independent of the product form, technical medium, or whether it includes AI features.
2. The positioning of AI in Scrum is as a team productivity empowerment tool, an experimentation accelerator, and a feedback enhancer. It is neither the core of the product nor an element for reconstructing the framework.
3. AI does not change the rules of Scrum; it merely radically compresses the cost of trial and error, shortens iteration cycles, and accelerates inspection and adaptation. This makes Scrum a more fitting work paradigm in the era of high uncertainty brought about by AI.
4. Whether a team delivers hardware, software, physical products, pure service products, and regardless of whether these products possess AI features, as long as the team uses AI to assist in requirements, design, development, testing, validation, or retrospectives, this guide applies.

I. The Five Universal Core Principles of AI+Scrum (Applicable to All Scenarios)

Building upon the five native Scrum values (Commitment, Focus, Openness, Respect, and Courage), these newly added principles are exclusive to the era of AI empowerment and are applicable to all product types and AI use cases:

1. The Principle of Accelerated Experimentation

Scrum relies on continuous experimentation to validate solutions to complex problems. The core value of AI lies in reducing experimentation costs, increasing experimentation density, and accelerating validation frequency. Teams are no longer constrained by human or time costs when iterating through trial and error. They can complete multi-dimensional hypothesis validations within the same Sprint, adapting to the developmental uncertainties of various complex software and hardware products.

2. The Principle of Human-Machine Collaboration

Humans define value, goals, hypotheses, standards, and decisions; AI handles execution, assistance, generation, organization, and analytical tasks. This applies to all teams: whether it is requirement grooming, hardware structural design assistance, software code generation, copywriting for prototypes, test cases, data analysis, or retrospective summaries, this division of labor logic is uniformly followed.

3. The Principle of Transparent Efficiency

Utilize AI to solidify work processes, accumulate iteration data, and organize artifact information. This addresses the shortcomings of traditional agile teams regarding opaque processes and a lack of accumulated assets. It strengthens the “Transparency” pillar of Scrum empiricism, unconstrained by product form.

4. The Principle of High-Frequency Feedback

AI can rapidly accomplish data statistics, issue summarization, user feedback analysis, and defect retrospectives. This helps teams completely break free from the traditional R&D shortcomings of phased, low-frequency, and lagging feedback. It enables high-frequency, real-time inspection and feedback throughout the entire iteration process, allowing for more timely and precise error correction and self-adaptive adjustments for all types of software and hardware products.

5. The Principle of Controllable Adaptation

AI is an assistive tool; the team retains full authority over all business, product, and quality decisions. It strictly prohibits AI from dictating product direction or replacing team judgment. This principle adapts to all commercial product R&D scenarios, avoiding the risks of tool abuse.

II. Supplementary Definitions for the Three Roles Empowered by AI (Universal for All Products)

No new roles are added, and the rights and responsibilities of the native roles remain unchanged. We only supplement the new responsibilities under the empowerment of AI tools, which are universally applicable to both software and hardware teams.

1. Product Owner (PO) AI Empowerment Responsibilities

Native core responsibilities remain unchanged: maximizing product value and managing the Product Backlog. New AI-empowered tasks:

• Utilizing AI to quickly organize market requirements, user feedback, and competitor information to assist in determining product direction (applicable to hardware/software/physical products).
• Leveraging AI to break down large requirements, detail user scenarios, and supplement acceptance criteria, thereby reducing requirement ambiguity.
• Quickly adjusting product priorities based on AI data analysis of iteration results and the degree of user value achievement.
• Defining the hypotheses to be validated for each iteration, ensuring AI serves the experimental goals rather than merely acting as an efficiency tool for administrative tasks.

2. Scrum Master (SM) AI Empowerment Responsibilities

Native core responsibilities remain unchanged: team effectiveness, rule implementation, impediment removal, and team empowerment. New AI-empowered tasks:

• Helping the team establish standardized AI usage protocols: which scenarios permit AI usage, which decisions prohibit AI substitution, and how AI outputs should be manually verified.
• Eliminating team anti-patterns: laziness induced by AI, thoughtless piling up of AI usage, over-reliance on tools, and the abandonment of independent thinking and business validation.
• Utilizing AI to track team iteration data, blocking issues, and cycle waste to assist the team in continuous improvement.
• Guiding the team to adapt to the AI agile rhythm of “high-frequency experimentation and rapid feedback,” adjusting to the different R&D rhythms of software and hardware.

3. Developers AI Empowerment Responsibilities

Native core responsibilities remain unchanged: self-organizing to deliver usable Increments and sharing accountability for the iteration results. New AI-empowered tasks:

• Software teams: AI-assisted coding, self-testing, test case generation, troubleshooting, and technical solution comparison.
• Hardware teams: AI-assisted structural design, solution simulation, parameter verification, issue retrospectives, and test data organization.
• General capabilities: Utilizing AI to handle repetitive, administrative, and organizational tasks, freeing human effort for core work such as experimental validation, value judgment, and innovative optimization.
• Performing manual verification, practical validation, and quality control on all AI-generated outputs to ensure the Increment meets the DoD standards.

III. Universal AI Empowerment Practices for the Five Scrum Events (Adaptable to All Scenarios)

All native events, timeboxes, and core goals remain entirely unchanged. Only the AI efficiency enhancement practice rules are supplemented, with no limitations on product forms.

1. The Sprint (The Core Container)

AI does not change the iterative nature, timebox rules, or increment delivery rules of the Sprint. New adaptation logic: AI drastically reduces the administrative costs within an iteration, granting the team ample time to focus on hypothesis validation, user feedback, and product adaptation and optimization. For highly uncertain software and hardware products, AI can further shorten the trial-and-error cycle to adapt to faster market changes.

2. Sprint Planning

AI assists in requirement breakdown, task decomposition, risk identification, and workload estimation. The core focus of the meeting returns to the essence of Scrum: determining the Sprint Goal, clarifying the core hypotheses to be validated in this round, and aligning on the delivered value. Whether hardware or software, Sprint Planning no longer consumes vast amounts of time on mechanical breakdown tasks.

3. Daily Scrum

AI automatically synchronizes task progress, work updates, and routine issue summaries, reducing the internal friction of reporting-style meetings. The Daily Scrum focuses on its core: impediment resolution, progress deviation, and self-adaptive adjustments. Newly added universal inspection items: Was the use of AI tools effective today? Is there an over-reliance? Do AI outputs need correction?

4. Sprint Review

Retaining the native review core: inspecting the Increment, gathering feedback, and adapting the Product Backlog. Newly added universal inspection dimensions: Did the AI tools genuinely improve efficiency in this iteration? Did they reduce the cost of trial and error? Did they help validate product hypotheses faster? The focus is not on AI capabilities, but on the product value and iteration efficiency gains brought by AI.

5. Sprint Retrospective

Permanently adding three universal retrospective dimensions (applicable to all teams):

• Efficiency Level: Did AI reduce repetitive work, shorten the iteration cycle, and decrease R&D waste?
• Quality Level: Did AI outputs introduce deviations, defects, or cognitive laziness?
• Improvement Level: How can the usage of AI be optimized in the next round so that the tools better serve product iteration and experimental validation?

IV. Universal AI Supplementary Specifications for the Three Artifacts (Adaptable to All Products)

1. Product Backlog

Native definition remains unchanged: an emergent, ordered, and continuously optimized list of product requirements. Supplement: AI is permitted to assist in grooming, clustering, refining, and ordering, but prioritization, value judgment, and requirement trade-offs must be manually decided by the PO. All backlog items for software and hardware products must clearly define the “hypotheses to be validated,” with AI acting solely as a refinement tool.

2. Sprint Backlog

Newly added universal attributes: Recording the AI empowerment scenarios, AI-assisted tasks, and the scope of tool usage for the current iteration. Ensure the team clearly distinguishes between: human decision-making work and AI-assisted execution work, avoiding loss of process control due to tool abuse.

3. Increment

Native definition remains unchanged: a usable, completed, potentially releasable product increment. Supplemented universal rules: AI does not change the quality standards of the Increment. All AI-assisted contents, solutions, code, designs, and data must strictly comply 100% with the team’s DoD (Definition of Done) and can only be incorporated into the Increment after manual verification.

V. Universal DoD Supplements (Mandatory Adaptation for All AI-Empowered Teams)

Regardless of software or hardware, and regardless of whether the product contains AI modules, these newly added unified completion standards apply:

• All AI-generated content has undergone manual review, verification, and correction.
• The use of AI tools has not substituted any decision-making; core product decisions have all been manually executed by the team.
• The efficiencies, issues, and experiences from the AI empowerment in this iteration have been consolidated for reuse in subsequent iterations.

VI. Universal AI+Scrum Anti-Patterns (A Guide for Teams to Avoid Pitfalls)

Addressing the universal misconceptions of all AI-empowered agile teams, irrespective of product form:

1. Putting the Cart Before the Horse with Tools: Overly focusing on the novelties of AI tools while ignoring Scrum’s core closed-loop of experimentation, validation, and value.
2. AI Substituting Decision-Making: Allowing AI to determine product priorities, technical solutions, and user value, resulting in the loss of the team’s subjective judgment.
3. Waterfall Abuse of AI: Using AI to do a one-time comprehensive plan, abandoning the Scrum core of small-step iterations and continuous experimentation.
4. Enhancing Efficiency Without Validation: AI accelerates work output but fails to accelerate the validation of product hypotheses and value.
5. Unregulated Tool Abuse: The team lacks unified AI usage standards, leading to outputs that are uncontrollable, untraceable, and impossible to review retrospectively.

VII. Core Summary of the Guide

Scrum remains unchanged, while AI accelerates it; the framework remains unchanged, while efficiency is upgraded; the core of experimentation remains unchanged, while the feedback loop becomes faster. AI does not create a new Scrum; AI merely empowers orthodox Scrum. Whether a team is developing hardware, software, physical products, non-AI products, or AI-native products, AI is a tool that serves empiricism-based iteration, not a disruptor. In an AI era characterized by extreme uncertainty, orthodox Scrum + AI empowerment is the optimal solution for complex product R&D.

About the Author:

Eric Liao, Founder and CEO of Scrum.CN, President of Scrum.org China, and Senior Scrum Consultant with 20 years of practical Scrum experience.