Andrew Sales on Unlocking True AI Business Impact

AI is no longer a futuristic concept – it’s a present-day reality. To explore how enterprises are navigating this shift, we spoke with Andrew Sales (Chief Methodologist, Scaled Agile).

In this conversation, Andrew highlights how organizations are moving beyond the hype to tackle the real challenge: channelling widespread AI adoption into meaningful business outcomes.

Throughout the interview, he emphasizes a practical and structured approach: success depends less on tools and more on how organizations align people, processes, and strategy to unlock value and deliver measurable AI business impact. He also encourages leaders to build ‘AI intuition,’ empower change agents across teams, and focus on real results instead of just activity. Drawing from his experience, Andrew offers a clear roadmap for navigating the complexities of AI adoption at scale.

The shift in enterprise AI adoption

From your experience advising Fortune 500 companies, how has the conversation around AI adoption evolved over the past few years?

Andrew: The biggest shift is that AI is no longer a question of availability, because it’s already embedded in the productivity suites most enterprises rely on.

The conversation has moved in two clear directions. The first is governance: How do we harness, coordinate, and safely oversee AI experimentation and the urgency for adoption among our workforce? The second is strategic alignment: How do we take all the energy around AI and its potential, and connect it to real business opportunities? We’ve genuinely passed the tipping point where people want to use AI.

They’re already doing it on their own. The job for leaders now is to govern that momentum and direct it toward meaningful outcomes. It’s no longer just managing the rollout of tools.

Rise of AI-native product development

What led Scaled Agile to focus on AI-native product development, and why is it becoming essential for organizations trying to innovate?

Andrew: Over the past 15 years, we’ve developed the Scaled Agile Framework (SAFe), which has become the most widely adopted approach for enterprises scaling product development through lean and agile methods.

AI has become a natural evolution of that work. Whether organizations are building AI directly into their products or using it to accelerate their development workflows, the disruption is significant, but so is the opportunity. At the heart of lean and agile approaches are iterative cycles, fast learning loops, and incremental delivery of value. Those principles are exactly what AI can accelerate.

Helping organizations embed AI into their roles, development practices, and customer solutions is a natural extension of the transformations we’ve been enabling for years.

Why AI investments don’t always deliver ROI

Many enterprises are investing heavily in AI tools but still struggle to see meaningful ROI. Why does technology investment alone rarely translate into successful AI transformation?

Andrew: You can’t start with the technology and go looking for a business problem to solve. Instead, start with the business need, then find the technology that solves it.

The challenge right now is that AI is genuinely exciting. The outputs you can get from a simple prompt are remarkable, and we’ve all gotten a little starstruck. The technology and the challenges of scale are the secondary concerns. Most individuals are already using AI in their day-to-day work. The real question is how to harness AI at an enterprise level with the right governance to achieve consistent AI business impact. That’s a similar problem that we solved 15 years ago with SAFe.

How do you take every team’s individual agile practice and scale it meaningfully across an organization?

Starting with business problems, not technology

You often emphasize starting with the business problem rather than the technology. What does that shift look like in practice?

Andrew: We guide organizations through what we call AI value patterns, which are proven frameworks for where AI actually creates value. We’ve condensed these into five that likely cover 80% of enterprise use cases. These include:

• Knowledge and Decision Support: Ensuring everyone has organizational knowledge at their fingertips when they make decisions.
• Customer Interaction: Chatbots, AI-driven services, improved customer touchpoints.
• Workforce Automation: Less glamorous, but extraordinarily rich in unlocking value.
• Risk and Control: Using AI as a watchdog for compliance, regulation, and quality.
• Expert Productivity: AI coding and similar tools that allow teams to operate as if they had a much larger, more experienced workforce.

Start by understanding your strategy, identify which of these patterns fits, and then find the right technology for that problem. Not the other way around.

Measuring what truly matters in AI

What metrics should leaders track to ensure AI adoption is genuinely improving outcomes and not just adding complexity?

Andrew: This is the same challenge we faced with SAFe. Organizations typically default to measuring activity: How many people are trained, how many licenses have been issued, and so on. While that’s useful for tracking rollout, these are really more of vanity metrics. You need to measure outcomes that reflect real AI business impact, not just activity or adoption. We recommend organizations track metrics such as:

• Competency: Does your workforce feel they’re developing genuine mastery, and is there strategic alignment?
• Flow: Is AI actually accelerating delivery and making processes more efficient?
• Business outcomes: Are you seeing improvements in at least one of these areas – customer satisfaction, cost reduction, and/or revenue?

And finally, it’s important to track costs. When you’re running ChatGPT casually, it feels free. But at enterprise scale, multiplying AI agents across the organization can consume millions of tokens. Understanding and managing the cost of AI is a critical, often-overlooked metric.

Leadership role in driving AI business impact

What role should CTOs and product leaders play in shifting their organizations toward AI-native operating models?

Andrew: Two things. First, leaders need to develop what I call AI intuition.

That’s not necessarily deep technical expertise, but rather having a strong enough understanding of AI so that when a business opportunity comes across their desk, or a decision needs to be made, they can instinctively ask whether it feels right. Without that intuition, they’ll struggle to lead AI-native organizations. Second, they need to understand what falls to them specifically. For CTOs, that means ensuring AI systems are governed effectively, operational, scalable, and used ethically.

It’s worth noting that this is where AI differs from lean and agile transformation. Agile primarily affected product development teams, whereas AI affects everyone from legal and marketing to operations and finance. The AI-native organization isn’t a product development initiative. It’s an organization-wide transformation.

Breaking silos: Enabling cross-functional AI adoption

AI adoption often exposes silos across teams. How can leadership foster the cross-functional alignment needed for AI-driven decision-making?

Andrew: We think about workforce development in four layers. At the base, everyone needs a level of AI fluency. That’s foundational. Above that are specialists who need role-specific training, like developers using AI to write code. At the top are leaders.

But the critical layer in between is what we call AI change agents. These people aren’t a centralized team, as that model doesn’t work with AI. Instead, you need decentralized change agents embedded inside every team and department. These change agents develop strong skills and drive change across workflows, products, and operations, while remaining coordinated at the organizational level. Leaders should identify those individuals early.

Finally, find an area of the business with real problems to solve and genuine support throughout. Quick wins are highly effective at bringing along parts of the organization that are more hesitant.

Skills for the AI-native workforce

What skills will teams need to develop in order to work effectively in an AI-native environment?

Andrew: There are two categories: Technical skills and softer skills.

Technical skills are important, as people must understand how to use large language models, how to do effective prompting, how to leverage retrieval-augmented generation to incorporate organizational knowledge, and how to build agentic workflows.

But the more interesting conversation is around softer skills that enable people to apply good judgment to AI outputs. For instance, empathy is important because teams will spend more direct time with customers as AI handles more automated work. Strategic thinking is needed to know when to follow the path AI suggests and when to override it. People are moving from heads-down execution to a role in which they provide purpose and intent to AI systems, evaluate outputs, catch bias, and make key decisions.

This shift from mechanical execution to strategic judgment is one of the most exciting transformations happening in the workforce right now.

The future of AI in product development

Over the next three to five years, how do you expect AI to reshape enterprise product development lifecycles?

Andrew: I’d caution against confident predictions. Even 12 months is difficult to forecast for AI right now. But there are four shifts I think organizations will need to make.

First, we’ve traditionally focused on outcomes in product development. Going forward, we also need to focus on intent, being explicit about purpose and direction with AI systems.

Second, we need to move from iterative cycles to genuine rapid experimentation, at a pace AI enables.

Third, having solved the challenge of scaling product development through SAFe, the new frontier is scaling innovation. AI has democratized it. Everyone can suddenly access expertise in coding, marketing, and legal matters, as an example. Governing this at scale will be critical.

Finally, cross-functional teams are now “cross-functional, AI-augmented teams,” and organizations need to think carefully about what that means in practice.

Advice or mentorship

If you had one piece of advice for organizations starting their AI transformation journey today, what would it be?

Andrew: Become an adaptive, learning organization, and do it urgently. The old model of big-bang transformation, where you stop, restructure, and restart, doesn’t work when the technology is evolving as fast as AI. Training is valuable, but you also need to build in time and space for your workforce to continuously keep up to date, knowing that what they learned last month may already be changing.

Alongside that continuous learning, find a forcing function. Identify one genuine business opportunity. Don’t rely on individual experiments; find a real problem with a team behind it, technology aligned with it, change agents supporting it, and leadership sponsoring it.

Deliver something meaningful that connects directly to business outcomes. That’s the catalyst that makes the rest of the organization believe transformation is real and possible.

In brief

AI success is driven by clarity of purpose and disciplined execution, not just adoption. Organizations that anchor their efforts in real business problems, invest in building both technical capability and human judgment, and create systems for continuous learning will be better positioned to realize sustained value. The real differentiator will be how well leaders turn AI’s potential into clear, measurable results that create lasting impact.