Achieving AI-Empowered Agility

“We are living in an amazing time, but either our organizations learn to harness and control value delivery in the age AI, or it will control us.” 

– Mik Kersten, Outputs to Outcomes [1]

Definition: AI-Empowered Agility is the capability to rapidly develop and responsibly deploy AI-driven products and solutions whilst also leveraging AI to continuously enhance the speed, quality, and adaptability of existing Lean-Agile methods.

What is AI-Empowered Agility?

The exponential rise of Artificial Intelligence, particularly generative AI, presents both a challenge and an opportunity. Despite widespread adoption, a significant maturity gap exists. Approximately two-thirds of companies are stuck with isolated pilots that fail to scale. Treating AI as a one-off initiative rather than a foundational shift to a new operating model leads to fragmented data, tech debt, and workflows that are not designed for an AI-native environment.

In contrast, “Future-Built” companies [2] – the top 5% of the market – have moved beyond the pilot phase. They are ‘AI-Native,’ with their operating model architected around agentic and generative AI tools. The good news is that many organizations already have the necessary foundation: Lean and Agile methods, designed to respond to technological changes such as these. But more is needed. AI-Empowered Agility describes the way of working required to harness this specific opportunity. 

This new way of working also requires a Human-Centric AI Culture. One that recognizes that AI’s greatest value is not replacement but augmentation, enabling humans to do more creative, empathetic, and strategic work. This means integrating AI agents as full-fledged teammates with defined responsibilities and accountability loops. 

This article describes the four critical shifts and the necessary cultural evolution to achieve AI-Empowered Agility and demonstrates how to incorporate these changes into your current SAFe implementation.

What are the four critical shifts to achieve AI-Empowered Agility?

Existing Lean-Agile methods are built around some core tenets. Central to these are:

• A focus on Outcomes: which recognizes that value is delivered only when an objectively measurable outcome has been achieved. 
• A Foundation of Cross-Functional Teams: Cross-functional teams cut across traditional silos, creating an empowered, fully autonomous group that can design, build, and deploy together.
• An Iterative Process: built upon the PDCA (Plan, Do, Check, Adjust). This ensures incremental value delivery and provides a learning opportunity in each increment, improving both the product and processes. 
• Practices for Development at Scale: Frameworks like SAFe allow organizations to coordinate and integrate the work of thousands of solution builders across multiple ARTs.

These core tenets are not replaced by AI-Empowered Agility; they are amplified through four critical shifts (Figure 1).

Each of these shifts is briefly described below, and then applied to SAFe in further detail later in the article.

Shift 1: From ‘Focus on Outcomes’ to ‘Focus on Outcomes and Intent’

Intent is the purpose behind any action, distinct from outcome, which is the desired result. While achieving outcomes remains essential, this shift recognises that as more work involves formulating clear prompts for AI agents, we must precisely define the intent and purpose guiding those outcomes.

Shift 2: From ‘Cross-Functional Teams’ to ‘Cross-Functional, AI-Augmented Teams’

Successfully navigating this shift requires training the entire organization in AI fluency. Teams continue to develop specific areas of expertise while also understanding how to collaborate with AI to supplement and extend these skills into other areas.

Shift 3: From ‘Iterative Learning Cycles’ to ‘Iterative Learning and Rapid Experimentation Cycles’

AI is a productivity multiplier and an engine for accelerated learning. It creates opportunities to drastically shorten the “Do” (execution) and “Check” (data analysis) phases of the Plan-Do-Check-Adjust cycle, allowing humans to devote more time to the high-value “Plan” (strategic alignment) and “Adjust” (adaptive decision-making) phases.

Shift 4: From ‘Development at Scale’ to ‘Development and Innovation at Scale’

The general availability and accessibility of AI tools significantly increase the potential footprint of innovation across the organization. In turn, this necessitates clear guidelines and operational AI technology infrastructure. This approach is essential for preventing fragile, unmaintainable AI solutions and technical debt, ensuring that emergent innovation remains scalable, secure, and maintainable across the organization.

Sustained and Fostered by a Human-Centric AI Culture

An essential cultural change underpins these four shifts. A Human-Centric AI culture is a commitment to fostering creativity, even as AI handles more and more of the daily execution. It re-anchors the humans across the organization to focus on strategic delivery, ensuring the human role is the final, critical loop for value, safety, and purpose.

How does SAFe help organizations achieve AI-Empowered Agility?

Organizations practicing SAFe are well-positioned to succeed. This section describes some key actions that will support the shift towards AI-Empowered SAFe.

Shift 1: Focus on Outcome and Intent

Outcomes are a critical part of the SAFe operating model, ensuring that outputs are aligned with real value delivery at every level. As AI agents take over more of the heavy lifting in developing working code, documents, or other assets, the focus needs to shift to clearly defining the intent that guides AI outputs towards these outcomes.

Action 1: Incorporate Intent in Requirements

Understanding intent is critical for any AI application to move beyond generic responses and become a ‘thought partner’ that understands the questions asked of it. All SAFe requirements, from User Stories to Epics, include an expected outcome, but these must now be supplemented with a clear intent. An example of a feature is shown in Figure 2.

That intent frames the prompt that the human team members provide AI. With a well-formed prompt, as shown below, the AI can help achieve the desired outcome.

Action 2: Focus Human Effort on Strategy, Validation, and Value

AI automates routine, information-intensive tasks, fundamentally shifting human effort away from the execution of the “how” and towards the strategic definition of the “why.” As such, teams should reallocate capacity to higher-value work such as complex architectural design, deeper customer engagement, and the exploration of new business opportunities.

Some of the most valuable human skills are deep business understanding and customer empathy, combined with AI fluency to write effective prompts and critically judge the AI’s output. All roles have a responsibility to ensure AI-generated work aligns with the vision, organisational context, and customer needs. 

Technical excellence is now measured by the rigour of validation, requiring critical thinking, ethical judgment, and prompt refinement. Retaining human judgment for critical decision-making protects the organisation’s integrity and maintains accountability, especially in regulated industries where human oversight is legally required for safety and data privacy. The quality of the output is tied to the quality of its input; speed is no substitute for quality.

Shift 2: Cross-functional, AI-Augmented Teams

The cross-functional Agile Team is the fundamental building block in SAFe. AI-Augmented Agile Teams amplify their inherent strengths, acting as a force multiplier for speed and collective intelligence. This is a necessary shift toward a new organisational norm. AI should handle the more routine aspects of execution, allowing human team members to focus on value, empathy, and social dynamics.

Action 1: Develop Organizational-Wide AI Fluency

The organization must adopt a continuous learning culture that prioritizes training for all employees on AI fluency, prompting techniques, and new collaboration methods. This systematic upskilling is critical for developing AI-Empowered T-shaped skills that combine deep domain expertise and a broad understanding of AI. 

Technology and its applications evolve at an unprecedented velocity. Developing AI-Empowered Agility requires a continuous commitment to learning. This approach is more iterative than traditional, one-time training. Implement “Upskill Relentlessly” programs and encourage teams to ask, “How can AI help?” at the start of every task.

All roles across the Framework must evolve to harness AI. For example, product roles shift their focus from writing detailed specifications to defining and amplifying intent. They frame design initiatives as testable hypotheses. Their aim is to validate AI-generated outputs while AI automates tasks such as Epic-to-Feature decomposition. AI also helps generate acceptance criteria. Figure 8 illustrates just how much ‘a day in the life of a Product Owner’ changes with AI.

A Day in the Life of a Product Owner

8:00 am	Customer Research While you slept, AI scanned 1,000 support tickets and 5 competitor releases. Start your day with a “Morning Brief” on the top 3 unaddressed pain points.
11:00 am	Backlog Management Instead of a 4-hour refinement session, you use a pre-trained prompt to turn a Feature into 6 stories with full acceptance criteria in 10 minutes.
3:00 pm	Roadmap Planning A stakeholder asks, “What if we pivot?” You run a pattern-recognition simulation to show how that shift impacts dependencies across 5 teams instantly.

Figure 3. A day in the life of a Product Owner

Action 2: Design AI-Empowered Workflows

This approach integrates AI agents as full-fledged teammates, focusing human efforts away from execution and toward the high-value work of strategic definition, ethical judgment, and creative interpretation. As AI fluency grows and a continuous learning culture is fostered, teams are equipped to create increasingly AI-empowered workflows.

To illustrate how fundamentally workflows change, consider this example of one iteration in an AI-Empowered Agile Team developing a “Car Consumer Lifestyle Selection” feature for a vehicle shopping app, as described in more detail below.

The workflow begins with the Product Owner defining the Intent, which clarifies the Why. In this case, “customers want to find lifestyle-aligned cars based on speed, gas savings, family size, off-roading, etc.”

The AI then generates technical user stories and BDD Acceptance Criteria, which the Human team members review and refine. AI is then used to draft initial code, mockups, and automated tests. Humans focus on validating and refining this output, including checking for accessibility and bias. 

The AI simulates 10,000 synthetic user tests and conducts instant code reviews. The Human team checks for critical edge cases and uses ethical judgment to correct the AI’s logic where needed (e.g., preventing a 2-seater sports car recommendation for a “Family Safety” user). 

The AI instantly verifies code against security and privacy regulations in the pipeline, with the Product Owner retaining final Release on Demand authority for a live A/B test. The AI synthesizes real-time data from this test, leaving the Human team free to focus on strategic pivots.

Action 3: Optimize Around Smaller Agile Teams

Partnering with AI agents enables smaller teams to manage high work throughput without the excessive communication overhead common in larger teams. [4] Team sizes of 5-7 individuals are now commonly observed. A caution: teams that are too small (3 or fewer humans) might lack the diverse perspectives and skills needed to address complex problems.

The change also reinforces the effectiveness of long-lived, stable teams to maintain deep trust and psychological safety. Unstable or constantly churning teams cannot build the collective trust necessary to effectively govern and critique AI output, which is essential for maximizing the quality of ideas.

Action 4: Provide AI Coaching through RTEs and Scrum Masters

Critical coaching roles, specifically the Scrum Master and Release Train Engineer (RTE), must evolve and focus on guiding best practices for human-AI collaboration. They must become experts in the effective management of AI agents, advanced prompting techniques, and responsible AI practices.

This specialized coaching ensures that AI’s speed does not compromise accountability or quality. By promoting daily tactics for healthy Human-in-the-Loop (HITL) engagement, these AI coaches enable teams to successfully integrate AI.

Shift 3: Iterative Learning and Rapid Experimentation Cycles

SAFe embeds PDCA and feedback loops throughout the Framework. The shift to iterative learning and rapid experimentation cycles amplifies this existing cornerstone. AI offloads task-based execution and data analysis, which in turn means that learning loops can be more intentional and frequent.

Action 1: Accelerate Feedback Loops 

AI accelerates execution and analysis of the ‘Do’ and ‘Check’ phases of the PDCA cycle, as shown in Figure 3, enabling the team to receive a working product or detailed performance data significantly faster. For each story or feature, the team can explore multiple design options in the time it would typically take to implement one. This ‘experimentation’ mindset is critical to AI-Empowered Agility, especially when AI underpins the products themselves.

The immediate impact is that the team’s focus shifts to the high-value “Plan” (strategic alignment) and “Adjust” (adaptive decision-making) phases. The team gains more cognitive bandwidth to focus on intent, context, validation, and pivoting quickly based on verified data. 

Additionally, AI-Empowered SAFe teams can incorporate multiple on-demand learning moments throughout each iteration to improve processes since AI can synthesize data on team sentiment, flow time, and working patterns. They do not wait for the end of iteration reviews or demos. Events still serve as a cadenced time to reflect with intention, but are not used to delay feedback. 

Action 2: Drive Fact-Based and Predictive Improvements

AI enhances fact-based improvement by analyzing massive amounts of real-time and historical data to identify root causes, test hypotheses, and suggest effective countermeasures, supporting SAFe Principle #2 Apply Systems Thinking. This shift helps human teams become exponentially better problem-solvers. Agile Teams and ARTs must incorporate these approaches into Team Retrospectives and the Inspect and Adapt event.

Additionally, since AI agents can continuously track market data, user sentiment, and emerging trends, this enables Agile Teams to move beyond analyzing past failures to proactively predict where risks, bottlenecks, or outright failures might occur. 

Action 3: Human-in-the-Loop AI-augmented Continuous Integration

As more continuous integration activities involve AI in one form or another, some important changes are required to the Continuous Delivery Pipeline, as shown in Figure 4.

Firstly, the ART should deploy AI agents to monitor quality and compliance within the Continuous Delivery Pipeline (CDP). These agents are tasked with continuously and automatically verifying every new code increment and deployed artifact against established security policies, ethical standards, and regulatory compliance rules.

Automated governance protects the Continuous Delivery Pipeline by providing instant, real-time sign-off on compliance artifacts. This is essential for highly regulated industries. Without it, the speed of AI-driven feature development would be constantly bottlenecked by manual gate reviews. By building quality and ethical requirements into the system from the start, the AI agents allow the CDP to maintain velocity without compromising necessary controls. This approach promotes rapid, decentralized innovation with consistent quality across the organization.

Decentralizing innovation means establishing a robust system for human team members to manage AI agents within the Continuous Delivery Pipeline (CDP).

• Prompt/Agent Branching allows teams to customize a shared agent’s configuration for quick, focused problem-solving.
• Agentic sync is used by human team members to merge successful changes and lessons learned back into base agents, promoting organizational-wide reuse. Merge practices are AI-Empowered, with automated quality checks that test agent updates against shared standards before merging.

For a scaled organization, improving the CDP remains vital to drive rapid, decentralized innovation while maintaining consistent quality and compliance across all teams.

Shift 4: Development and Innovation at Scale

The proven patterns in SAFe address the challenges of product development at scale, and innovation is at the heart of the Framework. As AI democratizes innovation, it enables workers across the organization to explore and potentially build new functionality. This scaling of innovation, if not managed appropriately, can lead to fragile, unmaintainable AI-generated assets, creating technical debt. Managed well, it will drive a surge of innovation across all departments, significantly accelerating value delivery. 

Action 1: Unlock Organizational-Wide Innovation through AI Usage

Teams and ARTs should use AI as a creative partner to build better products faster. AI tools can instantly analyze thousands of customer reviews, survey responses, and support tickets to determine user needs and sentiment. Instead of spending weeks on early designs, teams can use AI to generate mockups and run A/B tests. By letting AI handle repetitive tasks and organize feedback, teams free up time to focus on true innovation, ensuring they create high-quality products that people actually want. 

At the organizational level, leaders should be using AI to discover new ways to grow the business. By processing massive amounts of market data, AI can predict future trends and uncover hidden opportunities. Business Owners and Portfolio Leaders should be using AI to run advanced simulations, testing business strategies, resource plans, and roadmaps before committing to further investment. This gives the organization the confidence to safely experiment with bold new models, invest in future technologies, and stay one step ahead.

Action 2: Provide Self-Service AI Capabilities

Specialty teams, often structured as Shared Services, must evolve their mandate from providing basic services to creating self-service AI capabilities provided directly to the Agile teams that need them.

This approach is common in the creative and digital product industries, where a Design Shared Services team creates and shares AI agents that continuously monitor and correct design adherence and accessibility standards. This approach shifts the cost and complexity of AI from fragmented product teams to a highly efficient, centralized platform, fueling organizational-wide innovation. This and other examples are shown in Figure 5.

Shared Service Team	AI Self-Service Capability	Benefit
Design	AI agents that continuously monitor and correct design adherence and accessibility standards	Creates alignment and consistency and reduces cost and duplicated effort.
System Team	ML (Machine Learning) driven test generation and anomaly detection.	Speeds up continuous integration and proactive issue resolution.
Legal	AI agents that automate product compliance checks and/or initiate human reviews	Reduces bureaucratic bottlenecks and accelerates approvals.
Procurement	Automated partner evaluation and negotiation tools.	Drastically reduces time-to-source and improves commercial outcomes.

Figure 7. Examples of Shared Services providing AI capabilities

Action 3: Build and Govern Operational AI Technologies 

The organization must invest in robust infrastructure platforms that provide pre-approved systems, tools, APIs, and essential data services across value streams. Additionally, governance guardrails ensure that the organization’s accelerated pace of innovation—including prototypes built by non-engineers—remains secure and maintainable. Another benefit is preventing the accumulation of technical debt. The organization should proactively track AI usage costs, which can quickly spiral as more of the organization deploys increasingly complex agentic workflows.

SAFe organizations can operationalize AI by using existing structures to balance centralized governance and decentralized innovation (Figure 6). AI platforms can be utilized across many ARTs, allowing rapid integration of new AI tools and technology via established enterprise platforms [1]. AI-focused Shared Services create shared infrastructure and reusable AI components as AI agents. They enable ARTs in the effective usage of these emerging components [2]. AI and data governance integrate into existing SAFe practices, enabling centralization of AI spend and responsible AI guardians  while enabling local decentralized governance decisions for each ART or solution [3]. Local team members can utilize Communities of Practice (CoPs) to develop AI agents and platforms tailored to their needs, ARTs, or products. Innovation proven to be useful across value streams can be integrated back into the centralized platforms as needed [4]. 

What is a Human-Centric AI Culture?

---

“The future of AI is not about replacing humans; it’s about augmenting human capabilities.”

– Sundar Pichai, CEO of Google

---

Achieving the four shifts of AI-Empowered Agility requires a fundamental change in organizational culture. A Human-Centric AI culture ensures that the workforce remains adaptive and relevant as technology evolves. It recognizes and celebrates the critical role humans play in directing the AI, not just its output.

It is a dynamic, high-leverage environment in which the core challenges have shifted from execution to strategic definition. AI handles tasks that don’t require human judgment. That frees the human workforce to do what they do best: interpretation, creativity, empathy, and strategic decision-making. As AI commoditizes execution, the ability to apply these ‘Human Skills’ becomes the organization’s most important asset.

Building Trust for AI Adoption

Fear is the silent killer of achieving value with AI. If people fear replacement, they will hide data and stick to legacy processes. Organizations that treat AI solely as a cost-cutting tool see a temporary efficiency bump, then stagnation. Organizations that use AI to offload cognitive load allow humans to solve previously unsolvable problems and are ultimately the ones that succeed. 

The solution is to be explicit about how efficiency gains will be reinvested e.g., “We aren’t cutting the team; we are expecting the team to double our output with the same headcount”. This approach helps to build psychological safety and trust. The organization that gets this right outperforms the one that treats AI as a headcount-reduction tool.

AI-Empowered Agility is the operating model for the Age of AI. It is a systemic capability, built on the foundation of SAFe, achieved via a strategic shift from merely doing AI to becoming AI-Native. This unleashes human agency and competitive differentiation through four shifts: shifting human effort from execution to intent and strategy, accelerating learning cycles, unlocking innovation at scale, and augmenting human capabilities. This evolution is sustained by a Human-Centric AI Culture, ensuring that human judgment remains the final, non-negotiable loop for value, safety, and purpose.

---

References

[1] Kersten, Mik. Output to Outcomes. IT Revolution. 2026.

[2] BCG. “The Widening AI Value Gap.” https://www.bcg.com/publications/2025/are-you-generating-value-from-ai-the-widening-gap

[3] Publitas. “The Strategic Impact of Generative AI on Spanish Retail: Lessons from Carrefour.” https://www.publitas.com/blog/generative-ai-in-retail-carrefour-ai-rollout

[4] Medium. “Building Scalable and Cost Effective Voice Agents, A Platform Based Blueprint.” https://medium.com/cvs-health-tech-blog/building-scalable-and-cost-effective-voice-agents-a-platform-based-blueprint-fae6ee5881c9

[5]Harvard Business School. “When AI Joins the Team Better Ideas Surface.” https://www.library.hbs.edu/working-knowledge/when-ai-joins-the-team-better-ideas-surface

[6]IKEA. “How IKEA is Approaching AI for the Benefit of All.” https://www.ingka.com/newsroom/how-ikea-is-approaching-ai-for-the-benefit-of-all/