What Does AI Native SAFe Actually Mean for Your Organisation

Last night I joined Andrew Sales and the Scaled Agile team for the first in a series of webinars on AI-Native SAFe. I'm forever being asked for my views on the framework's evolution, so I figured I'd share my musings here. This isn't a comprehensive summary, just my take on the key messages.

Only 9% of Executives Are Seeing Value from GenAI Investments

The webinar opened with a finding from the 2026 Return on AI Institute study: only 9% of executives globally report seeing value from their generative AI investments. I'm not surprised by the number. It mirrors what we're seeing in the field. Organisations tend to fall into one of three patterns. Some give everyone access to Copilot, Claude, ChatGPT or Gemini, run a one-hour how-to session, and let people loose. Others issue Microsoft Copilot chat as the default and require a business case for anything beyond that. And some ban GenAI entirely (while everyone uses it anyway on their personal devices). At the individual level, the most prevalent use case seems to be coders using GitHub Copilot or equivalent, with most non-coders being self-professed dabblers.

The RoAI research attributes the lack of measurable value partly to the absence of a standard framework. This aligns with our observation that organisations with mature SAFe LPM practices are better at measuring the ROI on AI investments. SAFe has given them the habit of validating hypotheses before committing, and that discipline is being applied to their AI initiatives.

Not a Layer, a Reimagining

Andrew was clear that AI-Native SAFe is not AI layered onto existing SAFe. It's built on Lean and Agile foundations but recast from the ground up for the AI-Native organisation.

I think this is the logical next step, given that AI-Empowered SAFe has been the emerging storyline for about a year. That was the layering-on approach, and it made sense (it's what people knew, and it gave organisations somewhere to start). But there's real value in reimagining SAFe for an organisation where AI is genuinely native to how work gets done.

The thing I'm most glad about is that Scaled Agile hasn't forced everyone to make the leap at once. Core SAFe remains fully supported and is not going away. AI-Native SAFe sits alongside it as the evolution organisations adopt at their own pace. For most organisations I work with, that's exactly the right framing: something realistic to use today, and something to aspire to.

The Four Shifts to Achieve AI-Empowered Agility

Andrew recapped the four critical shifts required to achieve AI-Empowered Agility first shared at the SAFe Summit Amsterdam in March. I see these as helpful for organisations seeking to understand what needs to change for them to become AI-Native given their current state.

A New Role: AI Value Architect

This new role is focused on ensuring AI efforts translate into measurable results and provides coaching on AI adoption and fluency. This was flagged as a potential career path for SAFe Scrum Masters and SPCs. I find this genuinely interesting, and I'm curious to see the next level of detail. (Andrew said that this would be covered in a future webinar.)

New ART Events

Three new ART events evolve existing ones: PI Outcome Planning replaces PI Planning, Customer Demos replaces System Demo, and Sense and Respond is a new end-of-PI analytics event powered by AI. (I love the language of Customer Demos!) 

And an Evolved Continuous Innovation and Delivery Pipeline

The Continuous Delivery Pipeline becomes the Continuous Innovation and Delivery Pipeline and now spans discovery through go-to-market. I think including GTM in the pipeline will be a valuable enabler of flow. I haven't seen the details yet, but the direction makes sense.

Curated Data is the Foundation

Andrew flagged curated data as the single biggest predictor of AI success. My SAFe life started in the data domain. I spent over a decade in data warehousing, where curated, trusted, structured data was the foundation of everything. The rise of data lakes expanded what organisations could store and analyse, including unstructured and semi-structured data. AI-Native SAFe takes that a step further: fit-for-purpose data is the real issue now, and what "fit for purpose" means depends entirely on the use case. 

Stay Tuned

If you missed last night's session, the recording is available on the AI-Native SAFe Resource Hub. As is the link to register for the next webinar on June 30, "From Outputs to Outcomes with AI-Native SAFe”. Moving from outputs to outcomes has always been harder in practice than it sounds; it will be interesting to see how AI-Native SAFe approaches this challenge.

Update: I've written up the sessions that followed.

• Session two, From Outputs to Outcomes: What AI-Native SAFe Changes: the five-stage outcome cycle, OKRs and portfolio funding.
• Session three, So What is an AI-Native ART?: teams, roles and the validation bottleneck.
• Session four, What Happened to PI Planning in AI-Native SAFe?: PI Outcome Planning and Sense and Respond.